@3 . if . E’! E w’ 12 G)s Htidsten .rD x‘i .se of r ‘I ii > ~ale11 1 a () .4 og- I .— DISCLAIMER Portions of this document may be illegible in electronic image products. Images are produced from the best available original document. Nonlinear Dielectric Response of Water Treed XLPE Cable Insulation by Sverre Hvidsten A dissertation submitted to the Norwegian University of Science and Technology Department of Electrical Power Engineering in partial fi.dfilment of the requirements for the degree of Doctor Ingeni@r July 1999 This document has used FrameMaker@ as editor. The font is Times New Roman. Grapher@ from Golden Software has been used for graphical representation. ISBN 82-471 -0433-4 1999:63 ... PREFACE m PREFACE This thesis, submitted for the degree of doctor of engineering (dr.ing.), is restricted to work performed at the Norwegian University of Science and Technology (NTNU) dur- ing the years 1994-1999 under the supervision of Professor Erling Ildstad. I am especially grateful to my supervisor and Dr.Ing. Rolf Hegerberg at SINTEF Energy Research (SEfAS) for their guidance and inspiring discussions. I would also like to thank Oddvar Landr@ at NTNU for his help with the construction of equipment, and to all my friends and colleagues at the Department of Electrical Power Engineering at NTNU and the Materials Science group at SEfAS for their help and assistance. I am also grateful for having the pleasure of working with Bjorn Holmgren and Peter Werelius at the Royal Institute of Technology (KTH) at Stockholm. Finally I would like to express my deepest gratitude and love to my wife K&hild and our children for their patience and support during the work. Trondheim, June 1999 Sverre Hvidsten ABSTRACT v ABSTRACT Condition assessment of XLPE power cables is becoming increasingly important for the utilities, due to a large number of old cables in service with high probability of failure caused by water tree degradation. The commercial available techniques are generally based upon measurements of the dielectric response, either by time @olarisation/depo- larisation current or return voltage) or frequency domain measurements (&’,&‘‘). Recent- ly it has been found that a high number of water trees in XLPE insulated cables causes the dielectric response to increase more than linearly with increasing test voltage. This nonlinear feature of water tree degraded XLPE insulation has been suggested to be of a great importance, both for diagnostic purposes, and for fundamental understanding of the water tree phenomenon itself. The main purpose of this thesis have been to study the nonlinear feature of the dielectric response measured on watertreed XLPE insulation. This has been performed by dielec- tric response measurements in both time and frequency domain, numerical calculations of losses of simplified water tree models, and fiially water content and water permeation measurements on single water tiees. The dielectric response measurements were performed on service aged cable samples and laboratory aged Rogowski type objects. The main reason for performing laboratory ageing was to facilitate diagnostic testing as a function of ageing time of samples con- taining mainly vented water trees. A new method, based upon inserting NaCl particles at the interface between the upper semiconductive screen and the insulation, was found to successfully enhance initiation and growth of vented water trees. AC breakdown strength testing show that it is the vented water trees that reduce the breakdown level of both the laboratory aged test objects and service aged cable samples. Vented water treeing was found to cause the dielectric response to become nonlinear at a relatively low voltage level. However, the measured frequency domain dielectric re- sponse was larger, and found to be more nonlinear than values measured in time domain. This thesis describes a new mechanism for the nonlinear dielectric response. It is as- sumed that at low or no applied electric stress the water treed region is characterised by sphericrd micro voids filled with liquid water separated by channels of crazed insulation. vi ABSTRACT The effect of increasing the test voltage is to cause Maxwell mechanical tensile stresses strong enough to open up the crazing zones and elongate the water droplets into the me- chanically weak crazing zones. Finite Element Method (FEM) calculations show that the effect of the re-opening of crazing zones by an increased test voltage, strongly increases the dielectric loss of the water treed insulation. This is qualitatively in good agreement with the experimental re- sults obtained on water treed insulation, where increasing the test voltage above a certain value caused the losses to increase. The typical frequency independent dielectric re- sponse of water treed insulation can, however, not be explained by this model. Numerical calculations of losses, indicated that the mechanism of voltage assisted in- gress of water is more likely in treed regions with rather low contents of water. The mi- I cro-FTTR measurements of single vented water trees indicated that such regions were likely to be present 3-400pm within the tree tip, and close to the insulation screen. The process of refilling water into water tree structures is likely to be associated with a hysteresis effect. When removing (or reducing) the electric field, mechanical relaxation causes the channel to collapse and to slowly recover its former structure. Dielectric re- sponse measurements showed that a hysteresis was typically present when the response was nonlinear. CONTENTS CONTENTS ... Preface m Abstract v 1. Introduction 1 1.1 Background 1 1.2 Main purpose of work 3 1.3 Description of chapters 3 2. Theory 5 2.1 Introduction 5 2.2 Theory of watertree initiation and growth 5 2.2.1 Typical features of water trees 5 2.2.2 Structure of water trees 8 2.2.3 Electro-mechanical theory of watertreeing 9 2.2.4 Electrochemical theory of watertreeing 10 2.3 Dielectric response and relations between time and frequency domain measurements 12 2.3.1 The dielectric response function 12 2.3.2 Measurements of the dielectric respouse in the time domain 12 2.3.3 Measurements of the dielectric response in the frequency domain14 2.3.4 Relations between time and frequency domain 15 2.4 Mechanism causing a nonlinear dielectric response 18 2.4.1 Effect of application of electric fields 18 2.4.2 Hypothesis for increased dielectric losses 18 CONTENTS 3. Diagnostic equipment and test procedures 21 3.1 Introduction 21 3.2 Equipment for time domain measurements 21 3.3 Equipment for frequency domain measurements 24 3.4 Guarding and preparation of test objects 25 3.4.1 Rogowski test samples 25 3.4.2 Cable samples 27 3.4.3 Thermal treatment 27 3.5. Numerical evaluation of data 28 3.5.1 Methods for estimating of the degree of nonlinearity 28 3.5.2 Fourier transformation of time domain measurements 29 3.6 Test procedures 30 3.6.1 Measurements in time domain 30 3.6.2 Measurements in frequency domain 30 4. Characterisation of the service aged cable samples and laboratory test objects 31 4.1 Introduction 31 4.2 Experimental methods 31 4.2.1 Manufacturing of the rogowski type test object 31 4.2.2 Type of cables 35 4.2.3 Water tree analysis 35 4.2.4 AC breakdown strength testing 36 4.3 Service aged cable samples 38 4.3.1 Results from water tree analysis 38 4.3.2 Results from AC breakdown strength analysis 41 4.4 Laboratory aged test objects 42 4.4.1 Test condition and ageing procedure 42 4.4.2 Results from water tree analysis 42 4.4.3 Results from AC breakdown strength analysis 48 4.5 Discussion 49 coNms 4.6 Conclusions 49 5. Measurements of the time and the frequency domain dielectric response 51 5.1 Introduction 51 5.2 Time domain measurements 51 5.2.1 Time dependence 51 5.2.2 Voltage dependence 54 5.2.3 Sensitivity analysis 56 5.2.4 Water and thermal treatment 56 5.3 Frequency domain measurements 58 5.3.1 Frequency dependence 58 5.3.2 Voltage dependence 58 5.3.3 Hysteresis effect 64 5.4 Relation between time and frequency domain dielectric responses 64 5.4.1 Magnitude of the dielectric response 64 5.4.2 Degree of nonlinearity 66 5.5 Discussion 68 5.6 Conclusions 69 6. Computer simulation of proposed mechanism for nonlinear response 71 6.1 Introduction 71 6.2 Meted of calculation 71 6.2.1 The water tree model 71 6.2.2 Method of electric field and loss calculation 73 6.2.3 Calculation of electrostrictive pressure 75 6.3 Results from numerical calculations 75 6.3.1 Electric field distribution 75 6.3.2 Dielectric loss 78 6.3.3 Sensitivity analysis 81 6.3.4. Increase of temperature 81 6.3.5 Results from analytic calculations of electristrictive pressure 83 CONTENTS 6.4 Discussion 84 6.5 Conclusion 85 7. Content of water within water trees 87 7.1 Introduction 87 7.2 Experimental methods 87 7.2.1 Equipment for measurements of water content 87 7.2.3 Measurements of water permeation 90 7.3 Experimental results 92 7.3.1 Condition and amount of water in vented water trees 92 7.3.2 Distribution of water close to a water tree tip 93 7.3.3 Effect of eleetric field upon the water permeation rate 96 7.4 Discussion 100 7.5 Conclusion 101 8. Discussion 104 9 Conclusions lW Appendices 111 References 122 CHAPTER 1 INTRODUCTION 1.1. BACKGROUND In the Nordic countries, crosslinked polyethylene (XLPE) insulated power cables have been used for more than 30 years.