Wind Farm Management Decision Support Systems For Short Term Horizon

Yildirim Yurusen, Nurseda
Melero Estela, Julio Javier (dir.)

Universidad de Zaragoza, 2019


Abstract: Wind energy is one of the fastest growing energy sources and its technology maturity level is already higher than the majority of other renewables. Therefore, many countries started to change their financial support policies in an unfavourable way for the wind energy. This unsubsidised new era forces the wind industry to re-visit its expenditure components and to make improvements in operating strategies in order to minimise operational and maintenance (O&M;) costs. The classical maintenance strategies focus on a year advanced programming of calendar based maintenance visits and corrective interventions. In this classical approach the maintenance programming flexibility is quite limited, since this kind of programming ignores dynamic environment of the wind farm and real time data-driven indicators. Then, downtimes, and corresponding revenue losses, due to wind turbine inaccessibility occur because wind turbines are exposed to challenging dynamic environmental conditions and located in remote areas. Low accessibility is one of the predominant problems, and remote control not always solves the problems. The cost optimal O&M; strategies for the wind energy must consider condition based maintenance and a timely programming of wind turbine visit.
Thus, an elaborate and flexible approach, which is capable of considering condition and accessibility of wind turbines using meteorological measurements and operational records is highly needed for the wind farm O&M; management. The core objective of this thesis is the investigation of decision-making processes in wind farm management, and the generation of Decision Support Systems (DSSs) for O&M; of wind farms. In order to develop practical and feasible DSSs, the research is conducted prioritising data-driven approaches. There still exist various inefficiently used data sources in an operational wind farm, therefore there is a room for an improvement to use efficiently available data. Generally, in a wind farm, two types of condition monitoring data can be collected as online inspection and offline inspection data. Online inspection data can be obtained from both condition monitoring system (CMS) and Supervisory Control and Data Acquisition (SCADA). CMS data require an additional investment in the turbines while, on the contrary, SCADA data are already available in the turbines. As a third source, offline inspection data consist of the records of all O&M; visits to the wind farm, which are available but poorly recorded. In this study, the answer for the question of how to change a classical O&M; strategy to an enhanced one using only the existing data sources without the need for an additional investment is searched.
Firstly, analysis of key factors influencing in wind farm maintenance decisions is performed. In this regard, exploratory data analysis was considered to understand the monthly seasonality and the dependencies of day ahead hourly electricity market price, which is one of the decisive parameters for the wind farm revenue. Then, the connection between wind turbine failures, atmospheric variables and downtime is studied in order to provide additional information to a maintenance team and a maintenance planner for the intervention day. For the first part, well-structured and analysed electricity market price, electricity generation and demand data are needed. Therefore, the existing databases are reviewed for the case countries and a relevant analysis period is chosen. The electricity market data can be easily interpreted as time series data. To exhibit the characteristics of different electricity markets, various time series comparison tools are combined as an analysis guideline. By using this guideline, the drivers of the electricity market price are summarised for each case country. For the second part, available atmospheric and failure data for the relevant wind turbine components are gathered and combined. Then, convenient approaches among unsupervised learning models are selected. By combining the available tools and considering the needed information level for different purposes, the failure rules of prior to failure occurrence per month, in hours and in ten minutes increments are mined.
Then, what-if analysis for revenue tracking of maintenance decisions is performed in order to generate a DSS for the evaluation of the major maintenance decisions taken in wind farms. To this purpose, the impact of country dynamics and subsidy frameworks considering the electricity market conditions are modelled. The impact of the intervention timing is analysed and the sensitivity of financial losses to environmental causes of underperformance are estimated.
Finally, generation of decision support tool for planning of a maintenance day is studied to provide a useful maintenance DSS for in situ applications. The safe working rules considering the wind speed constraints for the accessibility to the wind turbine are reviewed taking into account the turbine manufacturer's O&M; guidelines. The characteristics of the maintenance visits are summarised. Wind turbine accessibility trials using numerical weather prediction forecasting techniques for wind speed variable and synthetic forecasts for wind speed and wind gust variables are presented. An intervention decision pool considering safe working rules is generated, containing a list of plans capable of providing the optimal sequence of various tasks and ranked for revenue prioritised timing.
This work has been part of the “Advanced Wind Energy Systems Operation and Maintenance Expertise" project, a European consortium with companies, universities and research centres from the wind energy sector. Parts of this work were developed in collaboration with other fellows in the project.


Abstract (other lang.): 

Pal. clave: energia ica ; ingenieria de mantenimiento ; procesos de decision

Titulación: Programa de Doctorado en Energías renovables y Eficiencia energética
Plan(es): Plan 509

Department: Instituto de Investigación Mixto CIRCE

Nota: Presentado: 17 12 2019
Nota: Tesis-Univ. Zaragoza, Instituto de Investigación Mixto CIRCE, 2019


-



Todos los derechos reservados All rights Reserved - Todos los derechos reservados



Visitas y descargas



 Record created 2020-06-01, last modified 2021-05-20


Fulltext:
Download fulltext
PDF

Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)