Transformer Condition Health Monitoring and Predictive Maintenance of HVDC Converter Stations

Introduction - Current Maintenance Practices and Industry Survey Findings

To establish a comprehensive understanding of current maintenance practices of HVDC substations, WG B4.89 conducted a questionnaire involving HVDC manufacturers, transmission system operators (TSO), and asset owners.

The survey showed that HVDC substations use a mix of traditional, time-based preventive and emerging predictive maintenance, with preventive maintenance dominating in line commutated converters (LCC) and more traditional approaches applied to VSC substations. A shift from annual to biennial outages is being considered to improve availability and reduce costs. Converter valves are the most critical components requiring continuous monitoring, together with its cooling system due to their essential role in thermal management and safe operation.

Overall, the survey highlights an industry need to transition toward condition-based and predictive maintenance, supported by enhanced monitoring capabilities, diagnostics, and data analytics.

Operation, Reliability and Failure Mechanisms

Chapter 2 describes the VSC converter station, the possible configurations (i.e. symmetric, asymmetric monopole, bipole) and applications (offshore and onshore). As throughout their service life, VSC substations operate under varying electrical, operational and environmental conditions that put stress on the equipment. The main stress factors are identified along with the VSC valves components that are prone to failure.

Chapter 3 presents the relationship between condition monitoring (CM), diagnostics, and prognostic in the context of HVDC systems. It first provides an overview of potential failure modes affecting VSC valve components, including valve submodules, capacitors, gate driver units, and cooling subsystems. These components are subjected to electrical, thermal, and mechanical stresses that may lead to two categories of failure: single-event failures (hard failures) and progressive degradation (wear-out failures).

The Figure 1 gives an overview of the information flow from power semiconductor devices and capacitors to the required variables, HI and RUL, for diagnostic and prognostic respectively.

Condition Monitoring (CM) and Health Indicators

Chapter 4 identifies the sensors within VSC valves to monitor parameters indicative of potential failures (health indicators) and reviews the main CM methods.

CM aims to assess equipment state of health (SoH) using sensors and monitoring systems that collect electrical, thermal, and environmental data. These measurements enable early detection of component degradation, abnormal operating behaviour, and wear mechanisms.

Health indicators are classified as electrical, thermal, or environmental parameters. Their effective interpretation is complex as values are affected by operating conditions such as temperature variations, load profiles, and measurement uncertainties.

For VSC valves, particular attention is given to Damage Sensitive Electrical Parameters (DSEPs). For semiconductors, DSEPs include parameters related to junction temperature such as on-state voltage, threshold voltage, thermal resistance, and power dissipation. For capacitors, equivalent series resistance (ESR) and capacitance represent key health indicators.

Diagnostic and Prognostic Techniques

Chapter 5 reviews the various diagnostic methods that evaluate equipment health conditions through the analysis of monitored parameters and generate a HI. Diagnostic techniques include signal processing, current waveform analysis, and machine learning algorithms such as neural networks. The root cause of failure of the VSC valve components are also described.

Chapter 6 provides an overview of the prognostic methods. Prognostic techniques track the health indicators, historical operational data, and additional contextual information over time to predict the RUL or to anticipate performance issues of VSC valves. Those methods typically rely on physics-of-failure models, knowledge-based approaches, and data-driven models. Several methodologies are described, including physical model-based, machine learning-based, knowledge-based, and data-driven approaches.

The effectiveness, challenges and future needs of CM, diagnostics and prognostics applied to VSC valves are identified.

The Different Maintenance Strategies

Chapter 7 describes the various maintenance strategies, the traditional maintenance approaches such as "run to failure" and "maintenance at regular time", as well as more advanced methods such as reliability-centered maintenance (RCM) and predictive maintenance. The Chapter...