Das kompakte Design von gasisolierten Systemen und die Unabhängigkeit gegenüber Umgebungsbedingungen führt zu einer idealen Eignung des Betriebsmittels für den Einsatz in modernen Energieversorgungssystemen. Ein Betrieb der Anlagen unter Gleichspannungsbelastung ist dabei aufgrund der zunehmenden räumlichen Distanz zwischen Erzeugungs- und Verbrauchszentren unumgänglich. Der sichere Betrieb über die geplante Lebensdauer von mehreren Jahrzehnten ist nur mit einer genauen Teilentladungsdiagnose, welche eine sensitive Messung und zweifelsfreie Interpretation der Ergebnisse beinhaltet, möglich. Dabei ist zu beachten, dass die von Wechselspannungsanwendungen bekannten physikalischen Zusammenhänge der Entladungsprozesse und des Einflusses von dielektrischen Grenzflächen aufgrund der veränderten Belastung mit einem zeitlich konstanten elektrischen Feld und der damit einhergehenden Raum- und Oberflächenladungsakkumulation nicht direkt übernommen werden können. Ziel dieser Arbeit ist daher die Entladungsprozesse an Defekten mit und ohne dielektrischer Grenzfläche in gasisolierten Gleichspannungssystemen zu analysieren und damit einen Beitrag für die sichere Interpretation von Teilentladungsmessungen zu leisten. Auch werden bekannte elektrische und neuartige optische Messmethoden hinsichtlich ihrer Möglichkeiten und Grenzen beim Einsatz unter Gleichspannungsbelastung untersucht. Für die experimentellen Arbeiten an drei verschiedenen Störstellen wird das schwach inhomogene Isoliersystem der Anlagen in drei Modellanordnungen nachgebildet. Die Untersuchung der ablaufenden Entladungsprozesse wird durch eine direkte Messung des Teilentladungsstroms ermöglicht. Dabei wird zwischen impulsbehafteten und impulslosen Anteilen unterschieden. Infolge von Montagefehlern oder unzureichender Materialqualität können feste, metallische Störstellen im Isoliersystem entstehen. Die experimentell betrachteten Abhängigkeiten der Entladungsprozesse von der Störstellenpolarität, dem Isoliergasdruck und der Spannungsbelastung erlauben eine Klassifizierung von vier verschiedenen Entladungsarten. Zusätzlich zu den Untersuchungen im derzeit am häufigsten verwendeten Isoliergas Schwefelhexafluorid konnte ein Vergleich der Ergebnisse mit der klimafreundlichen alternative synthetische Luft die Gemeinsamkeiten und Unterschiede bei Entladungen an festen Störstellen aufzeigen. Dabei ist insbesondere die signifikant veränderte Polaritätsabhängigkeit hervorzuheben. Der Kontakt von metallischen Partikeln mit der Feststoffisolierung kann zur Anlagerung des zuvor freibeweglichen, metallischen Defektes an der dielektrischen Grenzfläche führen. Die Ansammlung von Oberflächenladungen auf dem Feststoff beeinflusst dabei insbesondere den Entladungseinsatz. Aufgrund des zur festen metallischen Störstelle ohne Grenzfläche vergleichbaren Entladungsverhaltens im stationären Zustand ist eine Unterscheidung der Defekte anhand von Impulswiederholraten und Amplituden herausfordernd. Eine Besonderheit bei Gleichspannungsbelastung sind Entladungen, welche auf einer Gas-Feststoff-Quergrenzfläche an beschichteten Elektroden einsetzen können. Die Untersuchung der Ursachen für das Auftreten dieser Entladungen, die elektrischen und optischen Charakteristika der ablaufenden Prozesse und Strategien für die Vermeidung werden untersucht. Aus den Ergebnissen werden Prüfempfehlungen für die Teilentladungsdiagnose von gasisolierten Gleichspannungssystemen abgeleitet. Diese sind wesentlicher Bestandteil für einen zukünftigen Einsatz gasisolierter Gleichspannungssysteme in einem leistungsfähigen Elektroenergiesystem mit hoher Versorgungszuverlässigkeit. / The compact design and the independence from environmental conditions of gas-insulated systems leads to an ideal suitability of this high-voltage equipment for the use in a modern power supply system. The operation of the assets under DC voltage stress is unavoidable due to the increasing distance between the areas of power generation and consumption. The reliable operation during the estimated lifetime of several decades is only feasible with a precise partial discharge diagnosis. Hence, a sensitive measurement and a doubtless interpretation of the results is necessary. Nevertheless, it is necessary to take into account, that under alternating voltage stress established physical mechanisms of the discharge processes and the influence of dielectric interfaces cannot be adopted directly, due to the changed voltage stress with a constant electric field and the related surface and volume charge accumulation. Aim of this thesis is the analysis of defects in gas-insulated systems with and without dielectric interfaces under DC voltage stress and thereby to contribute to a reliable interpretation of partial discharge measurements. In addition, known electrical and novel optical measurement methods are investigated with respect to their capabilities and limitations when used under DC voltage stress. The experimental investigations are carried out in model electrode arrangements. The weakly inhomogeneous electrical field of the gas-insulated systems is replicated in three configurations, one for each defect investigated. The detailed analysis of the discharge processes is enabled by a direct measurement of the partial discharge currents. A distinction between impulse currents and pulseless currents is made. Due to assembly faults or insufficient material quality fixed, metallic protrusions can be created within the insulation system. The experimentally observed dependencies of the discharge processes on the polarity of the defect, the insulating gas pressure and the voltage stress permit a classification of four different types of discharge. In addition to the investigations in the most commonly used insulating gas sulphur-hexafluoride a comparison of the results with measurements in the climate-friendly alternative synthetic air are made. Derived from this, commonalities and differences in the discharge behaviour are discussed. Free moving, metallic particles can adhere to the gas-solid interface. The accumulation of surface charges at the solid insulator influences the partial discharge inception significantly. Due to the steady-state discharge behaviour, which is comparable to the fixed, metallic protrusion without contact to a dielectric interface, distinguishing between the two defects based on pulse repetition rates and amplitudes is challenging. A unique aspect under DC voltage stress are discharges at the orthogonal interface between electrode coating and insulating gas. The analysis of the causes of the occurrence of these discharges, their optical and electrical characteristics and strategies for the prevention are investigated. Derived from the results, recommendations for partial discharge diagnosis of gas-insulated DC systems are discussed. These recommendations are an essential component for the future use of this asset in a high-performance electric power system with high reliability of the power supply.

An AC system’s voltage unbalance by a fundamental frequency negative sequence system is usually the main cause for the emission of non-characteristic harmonics by current source converters as used in conventional HVDC systems. This emission takes place on both sides of each 12-pulse converter. On the DC side mainly a 2nd harmonic voltage appears driving a 2nd harmonic current. The magnitude of this harmonic current can exceed the magnitudes of the characteristic harmonics even if no low order resonance exists. Further non-characteristic harmonics generated by the converter under such unbalanced supply voltage conditions have frequencies with a frequency distance to the characteristic harmonics of 2 times the fundamental frequency. The main technical drawbacks are the unintended coupling between both AC systems and the risk of thyristor over-stresses by DC current discontinuities at low power transfer levels. On both AC sides the largest 2 non-characteristic current harmonics generated by a 12-pulse HVDC converter under unbalanced supply voltage conditions are a negative sequence system of the fundamental harmonic and a positive sequence system of the 3rd harmonic. Also on the AC sides further harmonics are emitted by the converter with a order number distance of 2 to the orders of the characteristic harmonics. However, in practical AC system operation special attention has to be paid to the 3rd harmonic distortion level, in particular when low order resonance appears between the system impedance and the impedance of the converter station AC filters. In order to avoid the above mentioned problems, large smoothing reactors and sometimes large blocking filters are installed on the DC side and the voltage distortion on the AC sides is reduced by AC filters. However, these filters require an expensive high component rating if they are tuned to the 2nd or 3rd harmonic respectively. The work shows that a modification of the valve firing can reduce the levels of the 2nd and 3rd harmonic without investment into additional primary equipment. Furthermore, this offers the chance to reduce the minimum power transfer level since also the risk of an intermittent DC current can be reduced. A corresponding algorithm and a control strategy are proposed. However, the calculation of an appropriate firing pattern requires a detailed modelling of the processes within the converters, especially the formation of the harmonics and the harmonic transfer between AC and DC sides. The work proposes a component vector model for the calculation of the harmonics. This model assumes that each harmonic consists of a first component representing the ideal conversion process, a 2nd component representing the impact of different commutation angles and in the case of the modified firing a 3rd component considering the impact of the intended non-equidistant firing. The work shows, that the harmonic component vectors resulting from voltage unbalance and from firing modulation can be treated separately and superimposed linearly. The calculation of the harmonic component vectors is performed applying the method of switching functions. For the consideration of the commutation and firing angle differences the modelling of switching functions based on differential impulses is proposed. However, especially an accurate representation of the above mentioned 2nd component vector requires a correct calculation of the commutation angles and their valve-specific differences. The investigations of this work have revealed that the conventional method of calculating the commutation angles – assuming an ideal smoothed DC current - may not produce results of sufficient accuracy. This is especially true in the case of a high ripple of the DC current, e.g. smoothed with a small smoothing reactor. A small smoothing reactor is typical for HVDC back-to-back applications. Therefore a new analytical method for the calculation of the commutation angles has been developed which in particular considers the typical pulse form of the DC current and additionally the impacts of the voltage unbalance and of the proposed modification of the firing on the ripple shape of the DC current. Moreover, as this analytical method requires the instantaneous values of the DC current at the instants of valve firing, a further analytical method for the calculation of these discrete current values has been developed. The equations are valid under the same conditions as the new ones for calculation of the commutation angles, i.e. resistive-inductive AC system fundamental frequency impedances, any degree of DC current smoothing between ideal smoothing and a ripple at the limit for current discontinuities. Symmetrical conditions, supply voltage unbalances and non-equidistant firing as proposed are applied. It is shown that, using this method, also the discrete values of the DC current at the end of the commutation intervals can be determined. In practice one of these discrete current values indicates the minimum value during one period of the fundamental frequency. This offers the chance for a more exact analytical determination of the limit for the appearance of DC current discontinuities. For typical parameters of a back-to-back installation the new methods and the new analytical equations have been compared with simulation results showing excellent correlation for typical voltage unbalances of not more than 1...2% and firing angle differences of not more than 2.5°. This verification is performed for the harmonics, the commutation angles and the discrete values of the DC current at the firing instants as well.

An AC system’s voltage unbalance by a fundamental frequency negative sequence system is usually the main cause for the emission of non-characteristic harmonics by current source converters as used in conventional HVDC systems. This emission takes place on both sides of each 12-pulse converter. On the DC side mainly a 2nd harmonic voltage appears driving a 2nd harmonic current. The magnitude of this harmonic current can exceed the magnitudes of the characteristic harmonics even if no low order resonance exists. Further non-characteristic harmonics generated by the converter under such unbalanced supply voltage conditions have frequencies with a frequency distance to the characteristic harmonics of 2 times the fundamental frequency. The main technical drawbacks are the unintended coupling between both AC systems and the risk of thyristor over-stresses by DC current discontinuities at low power transfer levels. On both AC sides the largest 2 non-characteristic current harmonics generated by a 12-pulse HVDC converter under unbalanced supply voltage conditions are a negative sequence system of the fundamental harmonic and a positive sequence system of the 3rd harmonic. Also on the AC sides further harmonics are emitted by the converter with a order number distance of 2 to the orders of the characteristic harmonics. However, in practical AC system operation special attention has to be paid to the 3rd harmonic distortion level, in particular when low order resonance appears between the system impedance and the impedance of the converter station AC filters. In order to avoid the above mentioned problems, large smoothing reactors and sometimes large blocking filters are installed on the DC side and the voltage distortion on the AC sides is reduced by AC filters. However, these filters require an expensive high component rating if they are tuned to the 2nd or 3rd harmonic respectively. The work shows that a modification of the valve firing can reduce the levels of the 2nd and 3rd harmonic without investment into additional primary equipment. Furthermore, this offers the chance to reduce the minimum power transfer level since also the risk of an intermittent DC current can be reduced. A corresponding algorithm and a control strategy are proposed. However, the calculation of an appropriate firing pattern requires a detailed modelling of the processes within the converters, especially the formation of the harmonics and the harmonic transfer between AC and DC sides. The work proposes a component vector model for the calculation of the harmonics. This model assumes that each harmonic consists of a first component representing the ideal conversion process, a 2nd component representing the impact of different commutation angles and in the case of the modified firing a 3rd component considering the impact of the intended non-equidistant firing. The work shows, that the harmonic component vectors resulting from voltage unbalance and from firing modulation can be treated separately and superimposed linearly. The calculation of the harmonic component vectors is performed applying the method of switching functions. For the consideration of the commutation and firing angle differences the modelling of switching functions based on differential impulses is proposed. However, especially an accurate representation of the above mentioned 2nd component vector requires a correct calculation of the commutation angles and their valve-specific differences. The investigations of this work have revealed that the conventional method of calculating the commutation angles – assuming an ideal smoothed DC current - may not produce results of sufficient accuracy. This is especially true in the case of a high ripple of the DC current, e.g. smoothed with a small smoothing reactor. A small smoothing reactor is typical for HVDC back-to-back applications. Therefore a new analytical method for the calculation of the commutation angles has been developed which in particular considers the typical pulse form of the DC current and additionally the impacts of the voltage unbalance and of the proposed modification of the firing on the ripple shape of the DC current. Moreover, as this analytical method requires the instantaneous values of the DC current at the instants of valve firing, a further analytical method for the calculation of these discrete current values has been developed. The equations are valid under the same conditions as the new ones for calculation of the commutation angles, i.e. resistive-inductive AC system fundamental frequency impedances, any degree of DC current smoothing between ideal smoothing and a ripple at the limit for current discontinuities. Symmetrical conditions, supply voltage unbalances and non-equidistant firing as proposed are applied. It is shown that, using this method, also the discrete values of the DC current at the end of the commutation intervals can be determined. In practice one of these discrete current values indicates the minimum value during one period of the fundamental frequency. This offers the chance for a more exact analytical determination of the limit for the appearance of DC current discontinuities. For typical parameters of a back-to-back installation the new methods and the new analytical equations have been compared with simulation results showing excellent correlation for typical voltage unbalances of not more than 1...2% and firing angle differences of not more than 2.5°. This verification is performed for the harmonics, the commutation angles and the discrete values of the DC current at the firing instants as well.:1 Einleitung und Ziel der Arbeit 1.1 Einführung in die Problematik 1.2 HGÜ-Systeme als Quelle von Strom- und Spannungsharmonischen 1.3 Netzspannungsunsymmetrien 1.4 Abgrenzung der betrachteten technischen Systeme 1.5 Beweggründe für die Betrachtung 1.6 Zielstellungen 2 Erkenntnisstand und Analyse der Aufgabenstellung 2.1 Harmonische 2.2 Aktive Kompensation von Harmonischen 2.3 Diskrete Werte des Zwischenkreisstromes am Beginn und Ende der Kommutierungsintervalle 2.4 Kommutierungswinkel 3 Grundlagen 3.1 Methodischer Ansatz 3.2 Allgemeine Voraussetzungen, Annahmen und Festlegungen 3.3 Maßgebliche Impedanzen für die Stromaufteilung 3.4 Maßgebliche Impedanz für die gleichstromseitigen Stromharmonischen 3.5 Leerlauf-Klemmenspannung des Stromrichters 3.6 Kommutierungsspannung 3.7 Nummerierungssystem der Ventile 3.8 Überlappungsformen der Kommutierungsintervalle 3.9 Komplexer Spannungsunsymmetriefaktor 3.10 Anwendung und Modifikation von Schaltfunktionen 3.11 Verifikation der Ergebnisse 4 Harmonische auf der Gleichstromseite 4.1 Bildungsgesetz 4.2 Charakteristische Harmonische 4.3 Nichtcharakteristische Harmonische infolge unsymmetrischer Netzspannungen 4.4 Nichtcharakteristische Harmonische infolge Ansteuermodifikation 5 Diskreter Wert des Zwischenkreisstromes im Zündzeitpunkt 5.1 Vorgehensweise 5.2 Lösungsansatz 5.3 Konstante Gegenspannung 5.4 Reale Gegenspannung des HGÜ-Stromrichters 5.5 Berücksichtigung von Resistanzen 5.6 Unsymmetrische Netzspannungen 5.7 Ansteuermodifikation 5.8 Unsymmetrische Netzspannungen und gleichzeitige Ansteuermodifikation 5.9 Ergebnisse 6 Kommutierungswinkel 6.1 Vorgehensweise 6.2 Konstante Gegenspannung 6.3 Reale Gegenspannung des HGÜ-Stromrichters 6.4 Berücksichtigung von Resistanzen 6.5 Unsymmetrische Netzspannungen 6.6 Ansteuermodifikation 6.7 Unsymmetrische Netzspannungen und gleichzeitige Ansteuermodifikation 6.8 Ergebnisse 7 Vertiefende Betrachtung der nichtcharakteristischen Harmonischen auf der Gleichstromseite 7.1 Vorbemerkungen 7.2 Unsymmetrische Netzspannungen 7.3 Ansteuermodifikation 7.4 Spannungsunsymmetrie und gleichzeitige Ansteuermodifikation 7.5 Ergebnisse 8 Harmonische auf der Netzseite 8.1 Bildungsgesetz 8.2 Charakteristische Harmonische 8.3 Nichtcharakteristische Harmonische 9 Betrachtungen zur aktiven Kompensation 9.1 Vorbemerkungen 9.2 Betrachtungsumfang 9.3 Grundlagen 9.4 Konzeptioneller Vorschlag für die Kompensation der 2. Stromharmonischen 9.5 Betrachtung der Drehstromseite 9.6 Vorschlag zur Weiterentwicklung des Konzeptes 9.7 Berechnungsbeispiel zur Kompensation der 2. Harmonischen im Zwischenkreis 9.8 Ergebnisse und Schlussfolgerungen 10 Zusammenfassung 11 Literatur 12 Formelzeichen und Abkürzungen 13 Anlagenverzeichnis