Session: 09-03-01 Wind Energy: Installation

Submission Number: 156283

A Framework for Reducing O&M Costs at Offshore Wind Farms 

Offshore Wind developments continue to grow globally at a rapid pace, with growth estimates of 630GW by 2050. To facilitate this rapid growth, deducing operations and maintenance costs, offers opportunity. A three-stage data-driven framework is proposed to offer a systematic approach for reducing these costs. The framework can be applied to any operational offshore asset and is presented here for the case-study of an offshore wind farm in the UK. The three stages are: data-analysis; failure prediction model development; maintenance schedule optimisation.

First an analysis of the operational data to identify the maintenance tasks that are most costly to the operator is performed. These costs arise from direct cost of maintenance and the revenue loss due to downtime. It is proposed that, given different characteristics, they should be approached differently in the context of failure predictions. It is also revealed that electrical components, particularly the power converters, are critical to the failure rate and energy losses due to maintenance at the offshore wind farm. As a result, failure predictions for power converters are chosen as the target for stage 2 of the framework.

Next, the root causes of failure and their leading indicators are identified. This process identifies the required data and defines the model architecture for failure prediction model. The model outcomes are then evaluated against the following scoring function: C=c_p(y+xz) + c_cx(1-z)  where C is the total score of the model, c_p is the cost of preventative maintenance, c_c is the cost of corrective maintenance, x is the annual failure rate of the component, y is the false positive rate of the model and z is the successful detection rate of the model. This function quantifies the benefit of the trained failure prediction model over business as usual, capturing the costs of both false positive replacements and correct replacements. Applied to the case-study wind farm, the signals required to detect failures in the power converters are identified and the insufficiencies in the SCADA data available to operators are highlighted. Failure prediction models are trained on the SCADA data to predict the power converter failures and evaluated with the scoring function. Potential cost savings if they are deployed in operation are shown.

The maintenance schedule optimisation method facilitates planning of maintenance across both tactical and operational timescales, taking inputs of maintenance jobs and failure predictions from stage 2 of the framework, along with weather forecasts, maintenance constraints and asset properties. Two mixed integer programming models are formulated, and a novel metaheuristic method is presented. When applied to the case-study wind farm, cost savings of 50% from using this method are found.

Overall, this work demonstrates a framework that leverages data available at offshore assets to reduce operations and maintenance costs and improve operational efficiencies. Data-driven techniques offer a promising route to further facilitating the deployment of offshore renewable energy assets

Presenting Author: Demitri Moros IDCORE & EDF UK

Presenting Author Biography: Demitri is a research engineer with EDF R&D Renewables in the UK and an EngD student on the IDCORE program, which is a joint CDT between the universities of Edinburgh, Exeter and Strathclyde. Demitri’s research is focused on operations and maintenance cost reductions for offshore wind farms by applying a data-driven approach to predict wind turbine failures and then optimise the maintenance scheduling.