Buoyancy Regulation Strategy for Underwater Profiler Based

Journal of Marine Science and Engineering

Article Buoyancy Regulation Strategy for Underwater Proﬁler Based on Adaptive Genetic Algorithm

Hui Zhi 1 , Puzhe Zhou 1 , Yanhu Chen 1,* , Xiaoyan Zhao 1, Yuandong Hong 1, Mingwei Lin 1 and Canjun Yang 1,2

1 The State Key Laboratory of Fluid Power and Mechatronic Systems, Zhejiang University, Hangzhou 310027, China; [email protected] (H.Z.); [email protected] (P.Z.); [email protected] (X.Z.); [email protected] (Y.H.); [email protected] (M.L.); [email protected] (C.Y.) 2 Ningbo Research Institute, Zhejiang University, Hangzhou 315000, China * Correspondence: [email protected]

Abstract: Considering the energy limitations of underwater vehicles, a strategy for energy saving is proposed. In the proposed buoyancy regulation strategy, oil of the buoyancy regulation system is pumped out several times at different depths instead of all at once. A balance between energy and time is achieved by assigning suitable weights, and the optimised depth which can be obtained from the pressure sensor is used as the judgement threshold based on the adaptive genetic algorithm. Through the numerical simulation using sea trial data, the inﬂuence of weight selection on energy and time is explored, and the frequency of oil draining for the vehicle to ascend is optimised. Simulation results show that the proposed buoyancy regulation strategy can save energy effectively when the frequency of oil draining is 4 times within depths of 0–500 m. Finally, trials were performed in Qiandao Lake and verify the contradictory relationship between energy and time.

Keywords: energy saving; buoyancy regulation strategy; adaptive genetic algorithm

Citation: Zhi, H.; Zhou, P.; Chen, Y.; Zhao, X.; Hong, Y.; Lin, M.; Yang, C. 1. Introduction Buoyancy Regulation Strategy for There are abundant marine resources on the Earth [1,2]. With the development of Underwater Profiler Based on underwater technology, autonomous underwater vehicles [3], gliders [4,5], and buoys [6] Adaptive Genetic Algorithm. J. Mar. Sci. Eng. 2021, 9, 53. have been developed for ocean observation. Energy is a crucial matter in the design process https://doi.org/10.3390/jmse9010053 of underwater vehicles. Most underwater vehicles use lithium batteries, which have a high energy density, as an energy source [7,8]. More importantly, some underwater vehicles, Received: 1 December 2020 such as underwater gliders, use lithium batteries as weights for attitude adjustments [9] to Accepted: 4 January 2021 utilise the weight and space of the battery. Published: 6 January 2021 In general, underwater gliders and buoys are equipped with a buoyancy regulation system. Some buoyancy regulation systems increase their weight by pumping seawater into Publisher’s Note: MDPI stays neu- the body [10], and some increase their volume by pumping oil from an inner cylinder to an tral with regard to jurisdictional clai- external bladder [11]. The two methods have one thing in common—both require a motor ms in published maps and institutio- to drive the pump, and the energy required for this motor accounts for most of the overall nal affiliations. energy consumption. Yazji et al. adjusted the volume of the vehicle by utilising electrolysis and reverse electrolysis of polymer electrolyte membrane fuel cell [12]. Um et al. changed the overall volume of the buoyancy regulation system by electrolysing water and then inflating an artificial bladder with the generated gas to increase buoyancy [13]. Although Copyright: © 2021 by the authors. Li- censee MDPI, Basel, Switzerland. this method of changing the volume of gas is effective in a diving environment, depth This article is an open access article control is difficult once the vehicle reaches the deep-water area because of the greater distributed under the terms and con- compressibility of the air. Therefore, the method of changing the volume by using hydraulic ditions of the Creative Commons At- oil has been widely used. tribution (CC BY) license (https:// Figure1 shows the typical working process of an underwater profiler. When the creativecommons.org/licenses/by/ profiler prepares to descend, it pumps the oil from its outer oil bladder into its inner 4.0/). cylinder. When the profiler prepares to ascend, the reverse process is performed to increase

J. Mar. Sci. Eng. 2021, 9, 53. https://doi.org/10.3390/jmse9010053 https://www.mdpi.com/journal/jmse J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 2 of 17

J. Mar. Sci. Eng. 2021, 9, 53 2 of 16

cylinder. When the profiler prepares to ascend, the reverse process is performed to in- thecrease buoyancy. the buoyancy. Generally, Generally, this process this is process performed is performed all at once. all As at the once. depth As increases,the depth the in- powercreases, consumption the power ofconsumption the motor of of the the buoyancy motor of regulation the buoyancy system regulation becomes largersystem [ 14be-]. Somecomes researchers larger [14]. have Some explored researchers buoyancy have regulation explored strategiesbuoyancy to regulation save the energy strategies of the to profilersave the during energy the of workingthe profiler process. during Petzrick the work eting al. process. [15] proposed Petzrick pumping et al. [15] oil proposed in small incrementspumping oil and in maintainingsmall increments the total and ascent maintaining time within the total 20 h. ascent Mu et time al. [ 16within] proposed 20 h. Mu an overallet al. [16] energy proposed model an for overall the floating energy process mode basedl for the on floating the multiple process quantitative based on regulation the multi- strategy;ple quantitative through regulation numerical strategy; calculation, through they concluded numerical thatcalculation, the system they power concluded consump- that tionthe system is the lowest power when consumption the buoyancy is the regulation lowest when system the ofbuoyancy a 4000-m-level regulation submersible system of is a adjusted4000-m-level 16 times submersible during the is floatingadjusted process. 16 times Liu during et al. [the17] proposedfloating process. a ‘stage Liu quantitative et al. [17] oilproposed draining a control‘stage quantitative mode’, and theoil systemdraining parameters, control mode’, including and oilthe discharge system parameters, resolution, judgementincluding oil threshold discharge of the resolution, floating speed,judgemen andt frequency threshold ofof oil the draining, floating were speed, optimised. and fre- Thesequency strategies of oil draining, and methods were optimised. are instructive. These Itstrategies is necessary and tomethods establish are accurate instructive. kine- It maticis necessary models to and establish select appropriateaccurate kinemati judgementc models thresholds, and select especially appropriate considering judgement the interferencethresholds, ofespecially ocean currents. considering In addition, the interference it is meaningful of ocean to currents. explore the In contradictoryaddition, it is relationshipmeaningful betweento explore energy the contradictory and time during rela ascenttionship for between the development energy and of appropriatetime during missionascent for assignments. the development of appropriate mission assignments.

FigureFigure 1.1.Schematic Schematic diagramdiagram ofof thethe workingworking processprocess of of an an underwater underwater proﬁler. profiler.

InIn thisthis article,article, wewe proposepropose aa buoyancybuoyancy regulationregulation strategystrategy withwith depthdepth asas thethe judge-judge- mentment threshold.threshold. Our strategy is is based based on on the the adaptive adaptive genetic genetic algorithm algorithm (AGA) (AGA) which which is isvalidated validated using using a ahybrid hybrid underwater underwater profil proﬁlerer (HUP) (HUP) [[18]18] namednamed ZhejiangZhejiang UniversityUniversity HUPHUP (ZJU–HUP)(ZJU–HUP) inin QiandaoQiandao Lake,Lake, China.China. TheThe remainderremainder ofof thisthis articlearticle isis organisedorganised asas follows.follows. SectionSection2 presents2 presents the the principle principle of a buoyancyof a buoyancy regulation regulation system system and the and ascending the as- kinematiccending kinematic model. Section model.3 presentsSection 3 the presents operation the operation process of process the AGA. of Sectionthe AGA.4 presents Section 4 thepresents results the of numericalresults of simulationsnumerical simulation within a depths within of 0–500 a depth m using of 0–500 sea trial m using data obtained sea trial indata July obtained 2017. Finally, in July Section 2017. Finally,5 presents Section the lake 5 presents trial and the the lake corresponding trial and the results.corresponding results. 2. Buoyancy Regulation System and Kinematic Model 2.1.2. Buoyancy Buoyancy Regulation System System and Kinematic Model 2.1. BuoyancyThe HUP Regulation designed System by Zhejiang University is a buoyancy-driven vehicle that can realise small-scale ocean observations based on the station-keeping method [18]. It uses a The HUP designed by Zhejiang University is a buoyancy-driven vehicle that can bi-directional gear pump to pump oil—oil is pumped from the internal hydraulic cylinder torealise the external small-scale bladder ocean to observations increase buoyancy based on and the conversely, station-keeping from themethod external [18]. bladderIt uses a tobi-directional the internal gear hydraulic pump cylinder to pump to oil—oil decrease is pumped buoyancy. from The the main internal speciﬁcations hydraulic of cylin- the ZJU–HUPder to the are external listed inbladder Table1. to increase buoyancy and conversely, from the external

J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 3 of 17

J. Mar. Sci. Eng. 2021, 9, 53 3 of 16 bladder to the internal hydraulic cylinder to decrease buoyancy. The main specifications of the ZJU–HUP are listed in Table 1.

TableTable 1. 1.Main Main speciﬁcations specifications of of Zhejiang Zhejiang University University hybrid hybrid underwater underwater proﬁler profiler (ZJU–HUP). (ZJU–HUP).

FeatureFeature Description Description Hull lengthHull length 2300 mm 2300 mm WeightWeight 82 kg 82 kg BuoyancyBuoyancy regulation regulation Gear pump Gear pump MaximumMaximum Volume Volume of the of bladder the bladder 1.8 L 1.8 L

AsAs shownshown inin FigureFigure2 ,2, when when the the HUP HUP starts starts to to dive dive into into the the deep-water, deep-water, the the motor motor isis turnedturned onon clockwiseclockwise toto drivedrive thethe bi-directional-gearbi-directional-gear pumppump toto letlet thethe oiloil flowflow fromfrom thethe hydraulichydraulic cylindercylinder to to the the external external bladder bladder through through the the check check valve. valve. Note Note that that the solenoidthe sole- valvenoid valve of the of system the system is in the is in position the position of the one-wayof the one-way valve whenvalve thewhen solenoid the solenoid valve isvalve not energised.is not energised. Considerable Considerable electrical electrical energy energy can be can saved be assaved there as is there no need is no to need turn onto turn the solenoidon the solenoid valve in valve the diving in the process.diving process. Meanwhile, Meanwhile, the equipment the equipment inside the inside cavity the is cavity safer becauseis safer thebecause check the valve check prevents valve theprevents backflow the ofbackflow the oil in of the the external oil in the bladder external despite bladder the highdespite pressure the high on thepressure external on bladder the external in the bladder deep-water in the region. deep-water When theregion. HUP When prepares the toHUP ascend prepares to the to water ascend surface, to the the water solenoid surface, valve the is solenoid turned on valve to let is the turned oil in on the to internal let the cylinderoil in the flow internal to the cylinder external flow bladder to the to external increase bladder buoyancy. to increase buoyancy.

FigureFigure 2.2.Schematic Schematic diagramdiagram ofof thethe buoyancybuoyancy regulationregulation system. system.

SinceSince thethe oceanocean densitydensity variesvaries withwith depth,depth, thethe volumevolume ofofthe theexternal externalbladder bladderof ofthe the proﬁlerprofiler is is suitably suitably changed changed for for the the vessel vessel to to ascend ascend or or descend. descend. When When the the proﬁler profiler stays stays in ain static a static state state at a at depth a depthhx, aℎ balance, a balance is attained is attained between between gravity gravity and and buoyancy: buoyancy:

𝜌𝑔𝑉 =𝑚𝑔 (1) ρhx gVhx = mg (1)

While the profiler moves to another depth ℎ, this balance is achieved in the same While the proﬁler moves to another depth h , this balance is achieved in the manner: y same manner: 𝜌 𝑔𝑉 =𝑚𝑔 ρhy gV hy = mg (2)(2)

wherewhereρ h𝜌xand andρh y𝜌are the are densitiesthe densities of the of seawater the seawater at depth at depthhx and ℎhy .and Further, ℎ. Further,Vhx and V𝑉hy are the drained water volume of the underwater vehicle at the corresponding depth. The and 𝑉 are the drained water volume of the underwater vehicle at the corresponding differencedepth. The in difference the volume in ofthe the volume external of bladderthe external that bladder is to be achievedthat is to frombe achieved depth h fromx to depth hy can be deduced as follows: depth ℎ to depth ℎ can be deduced as follows:

m m = V − V = − ∆Vhxhy hx hy (3) ρhx ρhy

J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 4 of 17 J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 4 of 17

𝑚 𝑚 ΔV =𝑉 −𝑉 = 𝑚 − 𝑚 (3) J. Mar. Sci. Eng. 2021, 9, 53 𝜌 𝜌 4 of 16 ΔV =𝑉 −𝑉= − (3) 𝜌 𝜌 To obtain the relationship between the motor power, speed of oil discharge, and To obtain the relationship between the motor power, speed of oil discharge, and depth, an experiment was performed based on the schematic diagram shown in Figure depth,To obtainan experiment the relationship was performed between the motorbased power,on the speed schematic of oil discharge, diagramand shown depth, in Figure 3. In the experiment, a relief valve was used to simulate the external pressure acting on an3. In experiment the experiment, was performed a relief based valve on was the schematicused to simulate diagram the shown external in Figure pressure3. In the acting on theexperiment, pump. First, a relief the valve relief was valve used towas simulate set at thedifferent external pressures pressure acting (from on 0 theMPa pump. to 5 MPa), the pump. First, the relief valve was set at different pressures (from 0 MPa to 5 MPa), andFirst, then, the relief the motor valve was was set turned at different on to pressures pump the (from oil 0from MPa tothe 5 MPa),hydraulic and then,cylinder the to the and then, the motor was turned on to pump the oil from the hydraulic cylinder to the externalmotor was container. turned on The to pump time thetaken oil fromfor oil the di hydraulicscharge cylinderand the to capacity the external of the container. external con- external container. The time taken for oil discharge and the capacity of the external con- tainerThe time were taken recorded for oil discharge to calculate and the the capacity oil discharge of the external speed. container A galvanometer were recorded based to on the tainer were recorded to calculate the oil discharge speed. A galvanometer based on the Hallcalculate principle the oil was discharge used speed.to measure A galvanometer the current based of the on motor. the Hall Then, principle the wasrelations used to between measureHall principle the current was ofused the to motor. measure Then, the the current relations of between the motor. motor Then, power, the speed relations of oil between motor power, speed of oil discharge and depth were obtained indirectly. dischargemotor power, and depth speed were of oil obtained discharge indirectly. and depth were obtained indirectly.

Hydraulic cylinderHydraulic cylinder

Motor Solenoid Motor valve Gear pump Solenoid valve Gear pump Relief valve Relief valve Motor Solenoid Relief valve Motor Solenoid valve Relief valve valve M Gear pump M Gear pump Hydraulic cylinderHydraulic cylinder

Figure 3. Schematic diagram of the experimental buoyancy regulation system. Figure 3.3.Schematic Schematic diagram diagram of theof the experimental experimental buoyancy buoyancy regulation regulation system. system. The experimental results are shown in Figure 4. The variation of motor power with TheThe experimental experimental results results are are shown shown in Figure in Figu4. There 4. variation The variation of motor of power motor with power with depthdepth can bebe describeddescribed by by a first-ordera first-order polynomial polynomial fitting: fitting: depth can be described by a first-order polynomial fitting: 𝑝 =𝑝ℎ+𝑝 (4) ph = p1h + p2 (4) 𝑝 =𝑝ℎ+𝑝 (4) where the fitted coefficients 𝑝 and 𝑝 are 0.032 and 19, respectively. The variation in where thethe fitted fitted coefficients coefficientsp1 and 𝑝 pand2 are 𝑝 0.032 are and 0.032 19, respectively.and 19, respectively. The variation The in variation the in the oil discharge speed with depth can be described by a third-order polynomial fitting: oilthe discharge oil discharge speed speed with depth with candepth be describedcan be described by a third-order by a third-order polynomial polynomial fitting: fitting: 𝑉 =𝑞 ℎ +𝑞 ℎ +𝑞 ℎ+𝑞 (5) V = q h3 + q h2 +q h+ q (5) o 𝑉1 =𝑞ℎ2 +𝑞ℎ3 +𝑞4ℎ+𝑞 (5) where the fitted coefficients 𝑞, 𝑞, 𝑞 and 𝑞are −1.2 10 , 1.6 10 , −0.31 and where thethe fitted fitted coefficients coefficientsq , q𝑞, q, 𝑞and, 𝑞q areand− 𝑞1.2are× 10 −1.2−7, 1.6 × 1010−, 41.6, −0.31 10and, 190−0.31, and 190, respectively. The root 1mean2 3 squared 4 error (RMSE) of the third-order polynomial respectively.190, respectively. The root The mean root squared mean error squared (RMSE) erro ofr the (RMSE) third-order of the polynomial third-order fitting polynomial is fitting is 1.7 which means that the error is small and can be accepted. 1.7fitting which is 1.7 means which that means the error that is smallthe error and canis small be accepted. and can be accepted.

FigureFigure 4. Motor powerpower and and speed speed of oilof oil discharge discha curvesrge curves at different at different depths. depths. Figure 4. Motor power and speed of oil discharge curves at different depths.

J. Mar. Sci. Eng. 2021, 9, 53 5 of 16

2.2. Kinematic Model During ascent, the HUP is subjected to gravity, drag force, and buoyancy. The drag force changes with the speed of the HUP as follows:

1 F = Cv2Sρ (6) f 2 where C represents the resistance coefficient; v, is the speed of the HUP; S, the cross section in the radial direction of the aluminium alloy cavity; and ρ, the density of the water outside the HUP. During ascent, the speed of the HUP changes as it is accelerated. Hence, the ascent is divided into small intervals for computational convenience. A small interval can be seen as a model of motion with constant acceleration in a sufficiently small time. The final speed of the first interval is calculated and used as the initial velocity for the next interval. Finally, the velocity distribution can be obtained through iterations. Therefore, the kinematic equation of one small interval is set up as follows:

ma = ρg(V + Vout) − Ff − mg (7)

where Vout represents the volume of the external bladder, and V represents the volume of the HUP, excluding the external bladder. According to the integral expression of acceleration:

dv a = (8) dt The speed of the HUP in the small interval during ascent is shown as follows: r a1 t √ v = tan h + C1 a1a2 (9) a2 2m

a1 = 2ρg(V + Vout ) − 2mg (10)

a2 = C·S (11)

where C1 represents the integral constant. The integral expression of velocity is as follows:

dh v = (12) dt The movement equation of the displacement and time is established by substituting Equation (9) into Equation (12):

r √ t a1 2·m· log tan h a1a2· C1 + 2m + 1 h = t· − + C2 (13) a2 a2

where C2 represents the integral constant. Using Equation (9) and Equation (13), the velocity and displacement of each subinterval are calculated and then the total displacement is obtained by summing up the displacements of all subintervals. The time of motion in the whole interval is the product of the number of subintervals and the step length.

3. Buoyancy Regulation Strategy Based on the kinematics model established in Section2, the optimal depth at which the oil is pumped needs to be solved. Since the kinematics model is nonlinear, it is difﬁcult to use conventional methods. Currently, intelligent optimisation algorithms are widely used. Among them, the AGA is an intelligent optimisation method in which the probability of crossover and mutation changes with the value of the objective function; J. Mar. Sci. Eng. 2021, 9, 53 6 of 16

AGA can effectively prevent the search process from converging to the local optimum value. The AGA was applied to a variety of problems such as job-shop problems [19], optical metasurface design [20], and sensor acquisition frequency adjustment [21], and satisfactory optimisation results were obtained. To optimise the depth accurately and to prevent convergence to the local optimal value, the AGA was applied to the buoyancy regulation strategy.

3.1. Optimisation Principle and Process In the AGA, the roulette method is used to select better individuals to prevent the loss of individuals with excellent genes. Gene recombination and mutation are important means of biological evolution. Similarly, two-parent individuals exchange partial structures to generate new individuals to improve the global search ability of the AGA. The individuals in the population can be mutated by changing the value at a certain gene position. Mutation may have two effects: one is to improve the local search ability of the AGA, and another is to maintain population diversity. In the population of a certain generation, different individuals have different fitness values. A high fitness value indicates a better individual, and such an individual should be retained. So this individual should have a small crossover probability and mutation probability. Generally, the difference between the maximum fitness value and the average fitness value in a generation is defined as Fdi f f for convenience. This value indicates the degree of convergence of the population. When the algorithm gradually converges, it may jump into the local optimal solution as Fdi f f decreases gradually. In this situation, the crossover probability and the mutation probability should be increased to enable the system to jump out of the local optimal solution. On the other hand, in the population of a certain generation, excellent individuals should have small crossover probability and mutation probability, and poor individuals should have large crossover probability and mutation probability to accelerate the elimination process. Therefore, the crossover probability Pc and mutation probability Pm [22] are expressed as follows:

 0 −  fmax− f 0  k1· , f ≥ f = Fdi f f Pc − (14)  0  k3 , f < f

 −  fmax− f  k2· , f ≥ f = Fdi f f Pm − (15)   k4 , f < f

where f represents the average fitness value of the population; fmax is the maximum fitness 0 value of the population; f and f are the fitness value; k1, k2, k3, k4 ≤ 1.0.

3.2. Construction of the Fitness Function The fitness function is used to describe the adaptability of each individual in the environment, and it determines whether an individual can be retained. Individuals with high fitness values are more likely to be inherited to the next generation. For the buoyancy regulation strategy described herein, the minimum value of the fitness function should be calculated. This function has two parts, namely, energy and time. It can be expressed as follows: k ! k !! f (x) = C − · P ·t + · (t + t ) max ωE ∑ h1 oi ωT ∑ i oi (16) i=1 i=1

where Cmax is a positive number used to make the ﬁtness function positive. ωE is the weight of energy and varies from 0 to 1. It represents the proportion of motor energy consumed of the buoyancy regulation device. ωT is the weight of time representing the overall time of the ﬂoating movement. The sum of ωE and ωT is 1. Furthermore, k is the

frequency of oil drainage. Phi is the power consumption of the motor when pumping oil J. Mar. Sci. Eng. 2021, 9, 53 7 of 16

at depth hi, and toi is the corresponding time. The ﬂoating time between two adjacent depth values is represented by ti, and it can be obtained by multiplying the number of subintervals by the step size.

3.3. Adaptive Genetic Algorithm Process J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEWThe application of the AGA for the buoyancy regulation strategy involves the follow-8 of 17

ing steps and the corresponding ﬂowchart is shown in Figure5.

Start AGA Generating the initial population

Parameter initialization Selection operation

Select the values of energy Crossover operation weight and time weight

Mutation operation Select the frequency of oil draining N Parameter Reach maximum selection iteration? Y

Y Have next frequency plan?

N Select the best solution of oil draining

Finish

FigureFigure 5.5. ImplementationImplementation processprocess ofof thethe buoyancybuoyancy regulationregulation strategy.strategy.

•4. NumericalStep 1: Parameter Simulations initialisation. First, perform experiments on the buoyancy regu- 4.1. Discussionlation system on Weights to obtain of Time the relationshipand Power between the motor power and depth and the relationship between the oil draining rate and depth of the pump. Then, collect In general, underwater gliders or profilers pump all oil to their external bladder at the temperature, salinity, and depth information when the profiler moves from the once to increase buoyancy. This means that the weight of energy is 1, and that of time sea surface to the target depth to fit the relationship between the density and depth weight is 0. In July 2017, a ZJU-HUP profile measurement experiment was conducted in according to the EOS-80 equation [23]. the South China Sea to verify its performance. According to the temperature and salt • Step 2: Select the values of power, weight and time weight. Based on the task re- data, the density for depths varying from 0 m to 553 m was obtained. The vertical pro- quirements, if the requirements for low energy consumption are stringent and the files are shown in Figure 6. time to perform the detection task is relatively sufficient, then a larger ωE is selected; if less time is available to perform the detection task, a larger ωT, that is, a smaller ωE is selected. • Step 3: Select the oil draining frequency. Different frequency values should be set in the AGA to compare energy and time at different oil draining frequencies.

J. Mar. Sci. Eng. 2021, 9, 53 8 of 16

• Step 4: Generate the initial population. In the AGA, randomly generated depth is used as individuals. For frequency k, generate (k − 1) individuals randomly, that is, (k − 1) depth values, to generate a population consisting of (k − 1) depth values. The reason for setting the number of individuals in the population as 1 less than the frequency of oil draining is that the first depth value is fixed. • Step 5: Execute the AGA. A new population is obtained through three operations: selection operation, crossover operation, and mutation operation. The number of population is set to 10, and the number of generation is set to 50. The value of k1 and k3 is 1. The value of k2 and k4 is 0.1. • Step 6: Determine whether the maximum number of iterations is reached. The number of iterations should be sufficiently large to allow results to converge. • Step 7: When a plan is completed, record the optimal energy and time. Change the frequency of oil draining until all the steps are executed. Finally, the frequency of oil draining corresponding to the smallest fitness value can be used as the execution solution.

4. Numerical Simulations 4.1. Discussion on Weights of Time and Power In general, underwater gliders or profilers pump all oil to their external bladder at once to increase buoyancy. This means that the weight of energy is 1, and that of time weight is 0. In July 2017, a ZJU-HUP profile measurement experiment was conducted in the South China Sea to verify its performance. According to the temperature and salt data, J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 9 of 17 the density for depths varying from 0 m to 553 m was obtained. The vertical profiles are shown in Figure6.

Figure 6. Depth profiles of temperature, salinity, and density in the South China Sea. Figure 6. Depth profiles of temperature, salinity, and density in the South China Sea. As seen from Figure6, the temperature in the ocean decreases as depth increases. The As seen from Figure 6, the temperature in the ocean decreases as depth increases. temperature at the ocean surface is about 30 ◦C, while it is only about 8 ◦C at a depth of The temperature at the ocean surface is about 30 °C, while it is only about 8 °C at a depth 500 m. However, the salinity first increases and then decreases, and its maximum value ofis 500 observed m. However, at a depth the of salinity 100 m. Bothfirst temperatureincreases and and then salinity decreases, affect and density. its maximum Therefore, valueaccording is observed to the EOS-80at a depth equation of 100 [23 m.], theBoth depth temperature profile of and density salinity can affect be fitted. density. The Therefore,temperature according and salinity to the remain EOS-80 nearly equation constant [23], in the the depth depth profile range of dens 0–22ity m, can and be hence, fitted.the The density temperature remains nearly and salinity constant remain in this nearly depth constant range. in the depth range of 0–22 m, and hence, the density remains nearly constant in this depth range. To gain insight into the relationship between energy and time under different weight conditions, a numerical simulation was performed using data of the aforementioned sea trial. Since the sum of energy weight and time weight is 1, numerical simulation could be performed only under different energy conditions or time conditions. The three-dimensional diagram of the relationship between energy and time under different power weight conditions is shown in Figure 7, and the corresponding two-dimensional diagram is shown in Figure 8.

J. Mar. Sci. Eng. 2021, 9, 53 9 of 16

To gain insight into the relationship between energy and time under different weight conditions, a numerical simulation was performed using data of the aforementioned sea trial. Since the sum of energy weight and time weight is 1, numerical simulation could be performed only under different energy conditions or time conditions. The three- dimensional diagram of the relationship between energy and time under different power J. Mar.J. Mar. Sci. Eng.Sci. Eng. 2021 2021, 9, x, FOR9, x FOR PEER PEER REVIEW REVIEW 10 of10 17 of 17 weight conditions is shown in Figure7, and the corresponding two-dimensional diagram is shown in Figure8.

FigureFigureFigure 7. Three-dimensional 7.7. Three-dimensionalThree-dimensional graph graphgraph of energy ofof energyenergy and andand time timeti variationme variationvariation with withwith different differentdifferent weights. weights.weights.

FigureFigureFigure 8. Two-dimensional 8.8. Two-dimensionalTwo-dimensional graph graphgraph of energy ofof energyenergy and andand time timetime variation variationvariation with withwith different differentdifferent weights weightsweights (top: (top: energy energy varyingvaryingvarying with withwith different differentdifferent weights; weights;weights; bottom: bottom:bottom: time timetime varying varyingvarying with withwith different differentdifferent weights).weights). weights).

As Asseen seen from from the theresults results of theof thenumeri numerical calsimulation, simulation, the theenergy energy consumed consumed gradu- gradu- allyally decreased, decreased, but butthe thetime time spent spent increased increased with with an increasean increase in energy in energy weight. weight. This This also also suggestssuggests that that to reduceto reduce energy, energy, the thetime time spent spent will will be longer.be longer. That That is, energyis, energy and and time time are areinversely inversely related. related. Different Different weights weights should should be selectedbe selected for forspecific specific tasks. tasks. If a If suffi- a suffi- cientlyciently long long time time is available is available or ifor the if theenergy energy possessed possessed by theby thedevice device is relatively is relatively less, less, thenthen energy energy should should be assignedbe assigned a greater a greater weight. weight. Conversely, Conversely, a larger a larger value value can can be se-be se- lectedlected for forthe theweight weight of the of thetime. time.

J. Mar. Sci. Eng. 2021, 9, 53 10 of 16

As seen from the results of the numerical simulation, the energy consumed gradually decreased, but the time spent increased with an increase in energy weight. This also suggests that to reduce energy, the time spent will be longer. That is, energy and time are J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 11 of 17 inversely related. Different weights should be selected for specific tasks. If a sufficiently long time is available or if the energy possessed by the device is relatively less, then energy should be assigned a greater weight. Conversely, a larger value can be selected for the 4.2.weight Optimisation of the time. of Oil-Draining Frequency 4.2.After Optimisation determining of Oil-Draining the weights, Frequency the frequency of oil draining is selected. When the AGA isAfter utilised determining for optimisation, the weights, different the frequency frequencies of oil draining correspond is selected. to different When theoptimal fitnessAGA values. is utilised When for optimisation,the energy weight different is frequencies0.5, that is, correspond the time weight to different is also optimal 0.5, a nu- mericalfitness simulation values. When is the performed energy weight for is different 0.5, that is, oil-draining the time weight frequencies. is also 0.5, a numericalThe result is shownsimulation in Figure is performed 9a. The convergence for different oil-draining curve of the frequencies. fitness value The is result shown is shown in Figure in 9b whenFigure an9 oil-draininga . The convergence frequency curve of of4 times. the fitness As seen value from is shown Figure in Figure9a, the9 bvalue when of an the fit- nessoil-draining function frequencyfirst increases of 4 times. and then As seen decreases from Figure with9a, the the increase value of theof the fitness oil-draining function fre- first increases and then decreases with the increase of the oil-draining frequency. This quency. This also means that the sum of energy and time first decreases and then in- also means that the sum of energy and time first decreases and then increases. When the creases.frequency When of oilthe draining frequency is 4 times,of oil thedraining energy is and 4 times, time spent the areenergy the minimum. and time Thisspent also are the minimum.means that This we shouldalso means turn on that the we motor should of the tu buoyancyrn on the regulation motor of system the buoyancy at four different regulation systemdepths at tofour pump different oil tothe depths external to pump bladder. oil The to the volume external of the bladder. oil required The betweenvolume twoof the oil requiredadjacent between depths shouldtwo adjacent be calculated depths according should tobe Equation calculated (3). according to Equation (3).

(a) (b)

FigureFigure 9. (a) 9. The(a) Thefitness ﬁtness value value against against th thee oil-draining oil-draining frequency; ( b()b Convergence) Convergence curve curve of the of ﬁtnessthe fitness value. value.

5. Lake Trial Results and Discussion 5. Lake Trial Results and Discussion TheThe above above discussion shows shows that that an inverse an inve relationrse relation exists between exists energybetween and time.energy If and time.a larger If a larger weight weight is assigned is assigned to energy, to the ener energygy, consumptionthe energy consumption of the device will of bethe reduced. device will be However,reduced. theHowever, time spent the in time ascending spent willin ascend increase.ing Similarly, will increase. if time isSimilarly, assigned if a largertime is as- signedweight, a larger the ascent weight, will bethe faster. ascent A fieldwill experimentbe faster. A was field conducted experiment in Qiandao was conducted Lake in in QiandaoSeptember Lake 2020 in forSeptember verifying the2020 buoyancy for verifying regulation the strategy buoyancy (see Figureregulation 10). strategy (see Figure 10).The depth of the site of the trial is 47.6 m. The depth profiles of temperature and salinity are shown in Figure 11. The temperature at the water surface was 28.8 ◦C, and that at a depth of 47.6 m was 12.5 ◦C. Salinity did not change significantly with depth as Qiandao Lake is a freshwater lake.

Figure 10. Lake trial conducted at Qiandao Lake.

J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 11 of 17

4.2. Optimisation of Oil-Draining Frequency After determining the weights, the frequency of oil draining is selected. When the AGA is utilised for optimisation, different frequencies correspond to different optimal fitness values. When the energy weight is 0.5, that is, the time weight is also 0.5, a numerical simulation is performed for different oil-draining frequencies. The result is shown in Figure 9a. The convergence curve of the fitness value is shown in Figure 9b when an oil-draining frequency of 4 times. As seen from Figure 9a, the value of the fitness function first increases and then decreases with the increase of the oil-draining frequency. This also means that the sum of energy and time first decreases and then increases. When the frequency of oil draining is 4 times, the energy and time spent are the minimum. This also means that we should turn on the motor of the buoyancy regulation system at four different depths to pump oil to the external bladder. The volume of the oil required between two adjacent depths should be calculated according to Equation (3).

(a) (b) Figure 9. (a) The fitness value against the oil-draining frequency; (b) Convergence curve of the fitness value.

5. Lake Trial Results and Discussion The above discussion shows that an inverse relation exists between energy and time. If a larger weight is assigned to energy, the energy consumption of the device will be reduced. However, the time spent in ascending will increase. Similarly, if time is as- J. Mar. Sci. Eng. 2021, 9, 53 signed a larger weight, the ascent will be faster. A field experiment was conducted11 of 16in Qiandao Lake in September 2020 for verifying the buoyancy regulation strategy (see Figure 10).

J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 12 of 17

The depth of the site of the trial is 47.6 m. The depth profiles of temperature and salinity are shown in Figure 11. The temperature at the water surface was 28.8 °C, and that at a depth of 47.6 m was 12.5 °C. Salinity did not change significantly with depth as Qi- FigureFigure 10.10. LakeLake trialtrial conductedconducted atat QiandaoQiandao Lake.Lake. andao Lake is a freshwater lake.

FigureFigure 11. 11. DepthDepth profiles profiles of of temperature, temperature, salinity salinity and and density density in in Qiandao Qiandao Lake. Lake. The main control chip of the ZJU-HUP is MSP430. It is responsible for the execution of The main control chip of the ZJU-HUP is MSP430. It is responsible for the execution the related actions, for example, turning on the motor and sending GPS (global positioning of the related actions, for example, turning on the motor and sending GPS (global posi- system) information. The sensor management board collects information from various tioning system) information. The sensor management board collects information from sensors. A circuit board based on Arduino was added to the ZJU-HUP in this experiment. various sensors. A circuit board based on Arduino was added to the ZJU-HUP in this This circuit board can receive sensor data from the master chip and store it into an SD experiment. This circuit board can receive sensor data from the master chip and store it (secure digital) card. It is also connected to two Hall elements that measure the current of into an SD (secure digital) card. It is also connected to two Hall elements that measure the buoyancy regulation system. the current of the buoyancy regulation system. The buoyancy regulation strategy, after AGA optimisation, is described in Table2. TheThe strategy buoyancy needs regulation to account strategy, for an initial after optimised AGA optimisation, depth of 40 is m described and a final in optimised Table 2. Thedepth strategy of 5 m. needs If the oilto account is pumped for to an the initial external optimised bladder depth once, theof 40 HUP m turnsand a on final the motoropti- misedof the depth buoyancy of 5 regulationm. If the oil system is pumped at a depth to the of 40external m to ascend bladder to once, a depth the of HUP 5 m allturns at once. on theIf themotor oil isof pumped the buoyancy to the externalregulation bladder system twice, at a thedepth HUP of partially40 m to ascend pumps to oil a outdepth first of at 5 a mdepth all at of once. 40 m toIf reachthe oil the is nextpumped depth to of the 9 m, external and then bladder at 9 m, ittwice, pumps the the HUP rest partially of the oil. pumps oil out first at a depth of 40 m to reach the next depth of 9 m, and then at 9 m, it pumps the rest of the oil. If the oil is pumped to the external bladder three times, the HUP pumps oil at three depths (40, 12, and 7 m).

Table 2. Optimisation results based on adaptive genetic algorithm.

First Depth Oil Draining Frequency Second Depth (m) Third Depth (m) Energy (J) Time (s) (m) Once 40 - - 1682.8 217 Twice 40 9 - 1677 346.27 Three times 40 12 7 1636.6 394.56

J. Mar. Sci. Eng. 2021, 9, 53 12 of 16

If the oil is pumped to the external bladder three times, the HUP pumps oil at three depths (40, 12, and 7 m).

Table 2. Optimisation results based on adaptive genetic algorithm.

Oil Draining First Depth Second Depth Third Depth Energy (J) Time (s) Frequency (m) (m) (m) J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 13 of 17 Once 40 - - 1682.8 217 Twice 40 9 - 1677 346.27 Three times 40 12 7 1636.6 394.56 The cable displacement sensor is a device that can output the position of the piston in a cylinder. Using this sensor, the amount of oil in the external bladder can be deter- The cable displacement sensor is a device that can output the position of the piston in mined by setting the position of the piston. The cable displacement sensor output volt- a cylinder. Using this sensor, the amount of oil in the external bladder can be determined age signal and the target voltage value can be calculated according to the amount of oil by setting the position of the piston. The cable displacement sensor output voltage signal required. The piston positions at different stages during the overall process are listed in and the target voltage value can be calculated according to the amount of oil required. The Table 3. piston positions at different stages during the overall process are listed in Table3. Table 3. The value of the cable displacement sensor corresponding to the depth interval. Table 3. The value of the cable displacement sensor corresponding to the depth interval. Initial Position First Position Third Position Oil Draining Frequency Second Position (mV) Oil Draining Initial Position(mV) First Position(mV) Second Position Third Position(mV) FrequencyOnce (mV) 1033 (mV)1165 (mV)- (mV) - OnceTwice 10331033 11651161 1165 - - - TwiceThree times 1033 1033 1161 1106 1165 1163 -1165 Three times 1033 1106 1163 1165 The value of the motor current during the trial is shown in Figure 12. The entire process is divided into three periods. The value of the cable displacement sensor is The value of the motor current during the trial is shown in Figure 12. The entire shown in Figure 13; the insets in the figure show a magnified view of the starting posi- process is divided into three periods. The value of the cable displacement sensor is shown tions for the buoyancy regulation strategy. When the buoyancy regulation system starts in Figure 13; the insets in the figure show a magnified view of the starting positions for the pumpingbuoyancy oil regulation from the strategy. external When bladder the (① buoyancy), the HUP regulation begins to system descend starts while pumping simulta- oil neouslyfrom the recording external bladder temperature, ( 1 ), the pressure, HUP begins time, to and descend current while data. simultaneously When a depth recording of 38 m istemperature, reached, the pressure, HUP starts time, pumping and current oil into data. the When external a depth bladder of 38 to m isincrease reached, the the overall HUP volumestarts pumping (②), so oilthat into it slows the external down gradually bladder to an increased then thebegins overall to ascend. volume The ( 2 ),depth so that at whichit slows the down HUP gradually begins to and pump then oil begins out is to ca ascend.lculated The based depth on at the which kinematics the HUP and begins dynamics;to pump calculations oil out is calculated show that based the onmaximu the kinematicsm depth that and the dynamics; HUP can calculations reach is 42 show m. Whenthat the the maximum HUP returns depth to thata distance the HUP of 2 can m fr reachom the is 42greatest m. When depth, the it HUPfirst stops returns for to 1 as anddistance then ofstarts 2 m to from implement the greatest the buoyancy depth, it first regulation stops for strategy 1 s and (③ then). starts to implement the buoyancy regulation strategy ( 3 ).

(a) (b) (c)

Figure 12. (a) Value ofof currentcurrent whenwhen thethe oiloil isis draining draining once; once; ( b(b)) Value Value of of current current when when the the oil oil is is draining draining twice; twice; (c )(c Value) Value of current when the oil is draining threeof times. current when the oil is draining three times

J. Mar.J. Mar. Sci. Sci. Eng. Eng.2021 2021, 9,, 53 9, x FOR PEER REVIEW 13 of14 16of 17

(a) (b) (c)

FigureFigure 13. (13.a) Value(a) Value of the of the cable cable displacement displacement sensor sens whenor when the the oil oil is drainingis draining once; once; (b )(b Value) Value of of the the cable cable displacement displacement sensorsensor when when the oil the is drainingoil is draining twice; twice; (c) Value (c) Value of the of cable the displacementcable displacement sensor sensor when when the oil the isdraining oil is draining three times.three times

TwoTwo phenomena phenomena were were noted noted in the in experiment. the experiment. First, duringFirst, theduring experiment the experiment in which in thewhich oil was the drained oil was thrice, drained the outputthrice, datathe output of the pressuredata of the sensor pressure underwent sensor a fluctuationunderwent a greaterfluctuation than 2 greater m, which than caused 2 m, thewhich buoyancy caused regulation the buoyancy strategy regulation to start strategy earlier during to start the ear- descent.lier during However, the descent. this phenomenon However, didthis not phen affectomenon the time did and not energy affect the calculations. time and Next,energy thecalculations. depths optimised Next, the by thedepths AGA optimised are listed inby Tablethe AGA2. However, are listed owing in Table to the 2. complex However, underwaterowing to the terrain complex of Qiandao underwater Lake, securityterrain of time Qiandao protection Lake, is security necessary time between protection two is consecutivenecessary depthsbetween and two the consecutive protection wasdepths triggered and the because protection the ascentwas triggered time between because the the twoascent depths time exceeded between 1 the min. two So depths the buoyancy exceeded regulation 1 min. So strategythe buoyancy was not regulation executed strate- at thegy set was depth. not executed Therefore, at thisthe set experiment depth. Therefor is a preliminarye, this experiment verification is a ofpreliminary the buoyancy verifi- regulationcation of strategy,the buoyancy but the regula contradictorytion strategy, relationship but the betweencontradictory energy relationship and time canbetween be obtained.energy and Then, time when can the be HUPobtained. is launched Then, when to perform the HUP tasks, is overalllaunched time to protectionperform tasks, is necessary.overall time If the protection operation is timenecessary. of HUP If the exceeds operation the predetermined time of HUP exceeds time, the the external predeter- bladdermined will time, be the filled external with oil, bladder and then, will the be HUPfilled will with start oil, ascending. and then, Therefore,the HUP will while start the as- HUPcending. was pumping Therefore, oil outwhile during the HUP the three was oil-drainingpumping oil steps, out overallduring timethe three protection oil-draining was triggered,steps, overall and as time a result, protection the final was oil triggered, volume in and the thirdas a result, experiment the final was oil more volume than thatin the inthird each ofexperiment the first two was experiments, more than that and in the each final of depth the first achieved two experiments, by the HUP and float the was final smaller.depth However,achieved by according the HUP to float the buoyancywas smaller. regulation However, strategy, according the minimumto the buoyancy depth atreg- whichulation the strategy, HUP can the reach minimum is fixed. Therefore,depth at whic accordingh the HUP to the can minimum reach is depth fixed. the Therefore, HUP reachedaccording in the to first the twominimum experiments, depth thethe timeHUP when reached the buoyancyin the first regulation two experiments, strategy the of the time thirdwhen experiment the buoyancy ends can regulation be inferred. strategy Thus, of the the range third of theexperiment buoyancy ends regulation can be strategy inferred. ofThus, the three the experimentsrange of the canbuoyancy be obtained. regulation Since st therategy changes of the in three the depthexperiments in the firstcan twobe ob- experimentstained. Since are the similar, changes the in overall the depth process in the of first the buoyancytwo experiments regulation are similar, strategy the is bothoverall representedprocess of asthe Interval buoyancy I. The regulati processon ofstrategy the third is both experiment represented is represented as Interval as I. Interval The process II. of Thethe third optimisation experiment results is represented listed in Table as Interval3 show thatII. the target position of the first oil pumpingThe is optimisation close to that of results the second listed oilin pumping.Table 3 show This that makes the thetarget depth position change of ofthe the first oil oil drainingpumping twice is close similar to tothat that of ofthe the second oil draining oil pumping. once. From This Figuremakes 14 the, we depth can seechange that of the the time spent is 236 s in interval I and 258 s in interval II. The third experiment requires 9.32% oil draining twice similar to that of the oil draining once. From Figure 14, we can see that more time than the first two experiments. the time spent is 236 s in interval I and 258 s in interval II. The third experiment requires To analyse the current data more comprehensively, box diagrams were plotted based 9.32% more time than the first two experiments. on the current value. The power of the motor changed when pumping oil at different To analyse the current data more comprehensively, box diagrams were plotted depths, and as the depth increased, the energy consumed of the motor increased too. based on the current value. The power of the motor changed when pumping oil at dif- By calculating the energy for interval I and interval II, we can draw the conclusion that the ferent depths, and as the depth increased, the energy consumed of the motor increased energy consumed of oil draining twice is 0.26% less than that of oil draining once and the too. By calculating the energy for interval I and interval II, we can draw the conclusion energy consumed of oil draining three times is 2.72% less than that of oil draining once. that the energy consumed of oil draining twice is 0.26% less than that of oil draining once and the energy consumed of oil draining three times is 2.72% less than that of oil draining once.

J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 15 of 17 J. Mar. Sci. Eng. 2021, 9, 53x FOR PEER REVIEW 1415 of 1617

3838

Ⅱ Ⅱ

Ⅰ Ⅰ

202 438 460 202 438 460 Figure 14. Depth variation during the experiment. Figure 14.14.Depth Depth variation variation during during the the experiment. experiment. The simulation results are compared with the experimental results, as shown in FigureTheThe 15. simulation simulationIt can be observed results results are that are compared the compared simula withtion with theresults experimentalthe are experimental close to results, the experimentresults, as shown as re- inshown in Figuresults, but 1515. .there It It can can are be bealso observed observed differences. that that the For simulation the the simula energy resultstion in Figure areresults close 15a, are to th the eclose value experiment to of the simulation experiment results, re- isbut greater thereare than also the differences. value of Forthe theexperiment. energy in FigureThis may 15a, be the due value to ofthe simulation inconsistency is greater be- sults, but there are also differences. For the energy in Figure 15a, the value of simulation tweenthan the pressure value of and the experiment.depth caused This by maythe syst be dueem toerror the of inconsistency the pressure between sensor. pressure For the istimeand greater depthin Figure causedthan 15b, the by the value the error system of between the error experiment.of the the simulation pressure This sensor. valuemay beand For due thethe time toexperiment the in Figureinconsistency value 15b, be- tweenmaythe error be pressurecaused between by and thethe simulationexperimentaldepth caused value setup. by and the Be thecause syst experimentem there error is valueno of depth the may pressure control be caused in sensor. the by theex- For the timeperiment,experimental in Figure the ZJU-HUP setup. 15b, Because the does error not there startbetween is noto move depth the from controlsimulation the in depth the value experiment, of 40 andm, but the themore experiment ZJU-HUP than 40 value maym.does This notbe will caused start make to moveby the the speed from experimental greater the depth than of setup. 0 40 at m, the butBe depthcause more of there than 40 m, 40 is which m.no This depth will will lead control make to a there- in the experiment,ductionspeed greater in thethe thanfloating ZJU-HUP 0 at thetime. depthdoes ofnot 40 start m, which to move will leadfrom to the a reduction depth of in 40 the m, ﬂoating but more time. than 40 m. This will make the speed greater than 0 at the depth of 40 m, which will lead to a reduction in the floating time.

(a) (b)

FigureFigure 15. 15.Comparison Comparison of of simulation simulation resultsresults andand experimentexperiment results. ( (aa)) Energy Energy comparison comparison of of simulation simulation results results and and experiment results; (b) Time comparison of simulation results and experiment results experiment results; (b) Time comparison of simulation results and experiment results. (a) (b) 6. Conclusions Conclusions Figure 15. Comparison of simulation results and experiment results. (a) Energy comparison of simulation results and InIn this paper, a buoyancy regulation stra strategytegy for for underwater underwater profilers profilers is is proposed. proposed. experiment results; (b) Time comparison of simulation results and experiment results InIn this strategy,strategy, oiloil is is pumped pumped out out several several times times instead instead of of all al atl once.at once. First, First, the the relationship relation- shipbetween between the volume the volume of the of external the external bladder bladder and theand density the density difference difference was established, was estab- 6.lished,and Conclusions that and between that between the motor the powermotor power and the and depth the wasdepth experimentally was experimentally obtained. obtained. Then, Then,a floatingIn a thisfloating kinematics paper, kinematics a model buoyancy ofmodel the regulation profilerof the prof was strailer established, tegywas established,for andunderwater the depthand the profilers of depth oil draining ofis oilproposed. Indrainingwas this optimised strategy, was optimised based oil onis pumped thebased AGA. on Next,outthe AGA.several the buoyancy Next, times the instead regulation buoyancy of strategy regulationall at once. was strategy numericallyFirst, the was relation- numerically simulated using sea trial data for depths of 0–500 m. The fitness function shipsimulated between using the sea volume trial data of for the depths external of 0–500 bladder m. The and fitness the function density was difference the smallest was estab- was the smallest when the oil is pumped out four times under the condition that the lished,when the and oil that is pumped between out the four motor times underpower the and condition the depth that was the weights experimentally assigned toobtained. weightspower and assigned time were to power both 0.5.and Finally,time were the both trial 0.5. was Finally, performed the trial at depths was performed of 0–40 m inat Then, a floating kinematics model of the profiler was established, and the depth of oil depthsQiandao of Lake.0–40 Duringm in Qiandao the ascent, Lake. the During ascent th timee ascent, of oil drainingthe ascent thrice time was of oil 9.32% draining more draining was optimised based on the AGA. Next, the buoyancy regulation strategy was numerically simulated using sea trial data for depths of 0–500 m. The fitness function was the smallest when the oil is pumped out four times under the condition that the

weights assigned to power and time were both 0.5. Finally, the trial was performed at depths of 0–40 m in Qiandao Lake. During the ascent, the ascent time of oil draining

J. Mar. Sci. Eng. 2021, 9, 53 15 of 16

than the ascent time of oil draining once, but the corresponding energy consumed was less by 2.72%. As the depth of the water area in this experiment is relatively shallow, the reduction of energy and the increase in time are not obvious. In addition, an appropriate depth control method should be studied and applied. Therefore, in subsequent research, we will verify the buoyancy regulation strategy with depth control in the sea trial to observe more signiﬁcant numerical changes.

Author Contributions: Conceptualization, H.Z., P.Z., and Y.C.; Data curation, H.Z. and M.L.; Formal analysis, Y.H.; Funding acquisition, C.Y.; Methodology, H.Z.; Resources, Y.C.; Software, X.Z.; Writing— original draft, H.Z.; Writing—review and editing, H.Z. All authors have read and agreed to the published version of the manuscript. Funding: This work was funded by National Natural Science Foundation of China (No. 51979246) and the Natural Science Foundation of Zhejiang Province of China (No. LY18E090003) Data Availability Statement: The data presented in this study are available on request from the corresponding author. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References 1. Guillerme, J.-B.; Couteau, C.; Coiffard, L. Applications for Marine Resources in Cosmetics. Cosmetics 2017, 4, 35. [CrossRef] 2. Copping, A.E.; Hemery, L.G.; Overhus, D.M.; Garavelli, L.; Freeman, M.C.; Whiting, J.M.; Gorton, A.M.; Farr, H.K.; Rose, D.J.; Tugade, L.G. Potential Environmental Effects of Marine Renewable Energy Development—The State of the Science. J. Mar. Sci. Eng. 2020, 8, 879. [CrossRef] 3. Fiorelli, E.; Leonard, N.E.; Bhatta, P.; Paley, D.A.; Bachmayer, R.; Fratantoni, D.M. Multi-AUV Control and Adaptive Sampling in Monterey Bay. IEEE J. Ocean. Eng. 2006, 31, 14. [CrossRef] 4. Yu, J.; Zhang, A.; Jin, W.; Chen, Q.; Tian, Y.; Liu, C. Development and experiments of the Sea-Wing underwater glider. China Ocean Eng. 2011, 25, 721–736. [CrossRef] 5. Webb, D.C.; Simonetti, P.J.; Jones, C.P. SLOCUM: An underwater glider propelled by environmental energy. IEEE J. Ocean. Eng. 2001, 26, 447–452. [CrossRef] 6. Roemmich, D.; Johnson, G.; Riser, S.; Davis, R.; Gilson, J.; Owens, W.B.; Garzoli, S.; Schmid, C.; Ignaszewski, M. The Argo Program: Observing the global ocean with profiling floats. Oceanography 2009, 22, 34–43. [CrossRef] 7. Asakawa, K.; Nakamura, M.; Kobayashi, T.; Watanabe, Y.; Hyakudome, T.; Ito, Y.; Kojima, J. Design Concept of Tsukuyomi —Underwater Glider Prototype for Virtual Mooring. In Proceedings of the OCEANS 2011 IEEE—Spain, Santander, Spain, 6–9 June 2011; pp. 1–5. 8. Cao, J.; Lu, D.; Li, D.; Zeng, Z.; Yao, B.; Lian, L. Smartfloat: A Multimodal Underwater Vehicle Combining Float and Glider Capabilities. IEEE Access 2019, 7, 77825–77838. [CrossRef] 9. Asakawa, K.; Kobayashi, T.; Nakamura, M.; Watanabe, Y.; Hyakudome, T.; Itoh, Y.; Kojima, J. Results of the First Sea-Test of Tsukuyomi: A Prototype of Underwater Gliders for Virtual Mooring. In Proceedings of the 2012 Oceans, Hampton Roads, VA, USA, 14–19 October 2012; pp. 1–5. 10. Alexander, W. Design of a Low-cost Open-source Underwater Glider. engrXiv 2018.[CrossRef] 11. Jones, C.P. Slocum Glider Persistent Oceanography. In Proceedings of the 2012 IEEE/OES Autonomous Underwater Vehicles (AUV), Southampton, UK, 24–27 September 2012; pp. 1–6. 12. Yazji, J.; Zaidi, H.; Torres, L.T.; Leroy, C.; Keow, A.; Chen, Z. A Novel Buoyancy Control Device Using Reversible PEM Fuel Cells. In Proceedings of the ASME 2019 Dynamic Systems and Control Conference, Park City, UT, USA, 8–11 October 2019; p. V003T17A006. 13. Um, T.I.; Chen, Z.; Bart-Smith, H. A novel electroactive polymer buoyancy control device for bio-inspired underwater vehicles. In Proceedings of the 2011 IEEE International Conference on Robotics and Automation, Shanghai, China, 9–13 May 2011; pp. 172–177. [CrossRef] 14. Si, W.; Xue, Y.; Liu, Y.; Li, Z.; Xue, G. Energy Consumption Modeling and Sensitivity Analysis for Deep-Argo Otarriinae Profiling Float. Math. Problems Eng. 2020, 2020, 1–14. [CrossRef] 15. Petzrick, E.; Truman, J.; Fargher, H. Profiling from 6000 meter with the APEX-Deep float. In Proceedings of the 2013 OCEANS— San Diego, San Diego, CA, USA, 23–27 September 2013; pp. 1–3. [CrossRef] 16. MU, W.; ZOU, Z.; Zhenxing, H. A Control Strategy with Low Power Consumption for Buoyancy Regulation System of Sub- mersibles. J. Xi’an Jiaotong Univ. 2018, 52, 44–49. [CrossRef] 17. Liu, M.; Yang, S.; Li, H.; Xu, J.; Li, X. Energy Consumption Analysis and Optimization of the Deep-Sea Self-Sustaining Profile Buoy. Energies 2019, 12, 2316. [CrossRef] 18. Zhou, P.; Yang, C.; Wu, S.; Zhu, Y. Designated Area Persistent Monitoring Strategies for Hybrid Underwater Profilers. IEEE J. Ocean. Eng. 2020, 45, 1322–1336. [CrossRef] J. Mar. Sci. Eng. 2021, 9, 53 16 of 16

19. Azzouz, A.; Ennigrou, M.; Said, L.B. Solving ﬂexible job-shop problem with sequence dependent setup time and learning effects using an adaptive genetic algorithm. Int. J. Comput. Intell. Stud. 2020, 9, 18–32. [CrossRef] 20. Jafar-Zanjani, S.; Inampudi, S.; Mosallaei, H. Adaptive Genetic Algorithm for Optical Metasurfaces Design. Sci. Rep. 2018, 8, 11040. [CrossRef][PubMed] 21. Chen, F.; Xu, S.; Zhao, Y.; Zhang, H. An Adaptive Genetic Algorithm of Adjusting Sensor Acquisition Frequency. Sensors 2020, 20, 990. [CrossRef][PubMed] 22. Srinivas, M.; Patnaik, L.M. Adaptive probabilities of crossover and mutation in genetic algorithms. IEEE Trans. Syst. Man Cybern. 1994, 24, 656–667. [CrossRef] 23. Isdale, J.D.; Morris, R. Physical properties of sea water solutions: Density. Desalination 1972, 10, 329–339. [CrossRef]