Learnings

Outcomes

• FMEA and gap analysis revealed continuously changing equipment failure trends with time, mostly driven by technological changes: sudden unreliability spikes were found to be generally related to early adoption of new technologies and increasing power transmission levels (infant mortality).
• The evidence at hand revealed that LCC and VSC converters are in principle highly reliable, with most of the systems’ outage time rather resulting from failures on the AC interfacing equipment. The LCC converter transformer and the STATCOM’s step up transformer have an important contribution to systems outage time. Transformer continuous health monitoring ought to be strengthened.
• No substantial evidence of ageing was found on the failure data for the electronic assets considered within the scope of the project, suggesting that early replacement may be a widespread practice in the industry. Equipment replacement may be occurring with very low risk margins.
• Keeping detailed maintenance and repair logs of system equipment is highly recommended to enable representative LCC-HVDC and STATCOM systems reliability analysis.
• STATCOM’s public failure data is almost non-existent. An initiative to record STATCOM failure and maintenance data across the GB system in a common format is recommended to provide better information to formulate future operation and maintenance policies.

Recommendations for further work

• The development and use of  digital twin models combining physics of failure, statistical and AI based models for the main system assets is recommended as an alternative to fill the ageing information void.
• The conduction of forensic analysis of failed equipment is recommended to clearly discriminate random from age-related failures. The collected information would be highly valuable for model validation under real life, practical operating conditions.
• The analysis of retired, non-failed equipment is recommended to have a critical assessment of the actual ageing state of the asset. Models fed with the collected data can help to better predict equipment lifetime under practical operating conditions and define improved risk-based maintenance policies, reducing unwanted unnecessary early equipment replacement.

Lessons Learnt

Deliverable 1A: findings of the LCC-HVDC FMEA and gap analysis.

LCC-HVDC is a mature technology with historical availability of around 94%, with forced and programmed unavailability of 1.8% and 3.5%, respectively.

Sudden jumps in unavailability observed in the historic data series were found to be mostly related to transitional periods to higher transfer capacities and voltage levels. Temporary Increases in unavailability reflected an increase due to the early stages of new technology adoption (infant mortality).

According to existing failure data, the converter transformer has been historically the main contributor to forced system unavailability. However, recent statistical failure data suggests that its participation in system outage time has been declining considerably in the past few years. Close monitoring of the converter transformer operating conditions is recommended to improve general system availability.

Physical damage to the converter valves or DC cable has the potential to result in prolonged outage times, however most of the severe failures reported for such components seem to be random in nature rather than age related.

Deliverable 1B: LCC-HVDC system reliability analysis and survival analysis of critical system components were reported.

With the help of the developed analytical reliability models, it was found that the reliability of 300-500kV systems seem to be not very different to that of 600 kV LCC-HVDC systems. Thus more broadly available failure statistics from 300-500 kV systems may be used in the analysis of emerging 600 kV LCC-HVDC links. The analysis of the historic series of the failure intensity for HVDC systems operating with AC/DC voltages in the 300-500 kV range revealed no evident sign of ageing.

Alternatives to the generation of realistic equipment PDF curves were recommended. The development and use of phenomenological or physics-of-failure lifetime models for the main system assets was suggested as a possible alternative.

The analysis of the thyristor cell failure data revealed the existence of a cyclic pattern in its failure rate related to system age, exhibiting local unreliability peaks every 12 years. The evidence suggests that periodic maintenance of ancillary systems (e.g. cooling system) influence the converter cells’ reliability.

Kaplan-Meier curves constructed for the thyristor cell survival analysis revealed a survival probability of about 90% after 40 years of operation, highlighting the components high resilience. No evidence of ageing in other key assets was found, suggesting that early replacement may be a widespread practice in industry.

Keeping detailed maintenance and repair logs of system equipment is highly recommended to enable representative LCC-HVDC systems reliability analysis.

Deliverable 2A: STATCOM FMEA and gap analysis. 

STATCOM system main components’ functions, sources of failure data and principal failure modes are described in the report. A survey conducted revealed that native reliability data for STATCOMs is extremely scarce in the public domain. The use of physics of failure models has been frequently used in academia to mitigate this deficiency. Use of failure data from similar equipment, but from a different installation type, is suggested as another alternative.

Existing aggregate failure data revealed high system availability in practical STATCOM installations, approaching in most cases 99%, with a converter reliability of about 99%. Partial and even full redundancy of critical STATCOM components is commonly offered by manufacturers. This fact can explain the high availability exhibited by the system in practice.

The survey also revealed that most of the unavailability was the result of failures on the AC side equipment. Close monitoring of such equipment is recommended.

An initiative to record STATCOM failure and maintenance data across the GB system in a common format is recommended to provide better information to formulate future operation and maintenance policies.

Deliverable 2B: Reliability analysis of the STATCOM system.

Performing representative STATCOM reliability analyses is difficult given the lack of empirical failure data. The use of failure data for the STATCOM system main components, obtained from alternative installation types, as recommended in deliverable 2A, was used to fill the information gap. MTBF and MTTR reliability parameters for the system components were calculated using this approach.

Reliability analyses conducted with a purpose-built analytical model and the calculated reliability indices revealed a theorical STATCOM availability of 99.75%, which aligns with empirical evidence. The STATCOM can be considered a very reliable system. It can be argued that the inbuilt partial and full system redundancy of the main subsystems is the main contributor to the high system availability.

Given the lack of empirical failure data from which one can derive realistic equipment PDFs, sensitivity analyses were conducted to emulate system ageing. The sensitivity analyses showed that increases in the equipment failure rate results only in a moderate system unavailability increase.

Methods to fit limited empirical data to the normal and two parameters Weibull distribution models were investigated. When limited failure data is analysed, the reliability projections should be considered with caution. The investigated models exhibited excessive sensitivity to large, sudden variations due to ‘rogue’ data points within a small dataset, having the potential to skew results to unrealistic values. In the presence of very limited failure data, engineering judgement is necessary to select the most adequate regression model for specific dataset, and to judge the representativeness of the ensuing probabilistic distribution.

Dissemination

Two papers on the LCC HVDC system reliability modelling and analysis were presented at the IET Power Electronics, Machines and Drives (PEMD) conference 2020.

Two papers on the STATCOM reliability modelling and analysis are also planned.