Machine intelligent forecasting based penalty cost minimization in hybrid wind‐battery farms. (8th July 2021)

• Record Type: Journal Article
• Authors: Dhiman, Harsh S.; Deb, Dipankar; Muyeen, S. M.; Abraham, Ajith
• Is Part Of: International transactions on electrical energy systems. Volume 31:Number 9(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-07-08
• Subjects: battery energy storage systems (BESS) -- machine learning (ML) -- penalty cost -- wind farms -- wind forecasting; Electric power -- Periodicals; Electric power systems -- Periodicals; Electrical engineering -- Periodicals; 621.3
• DOI: 10.1002/2050-7038.13010
• Languages: English
• ISSNs: 2050-7038

Abstract

Summary: Modern‐day hybrid wind farm operation is fundamentally dependent on the accuracy of short‐term wind power forecasts. However, the inevitable error in wind power forecasting limits the power transfer capability to the utility grid, which calls for battery energy storage systems to furnish the deficit power. This manuscript addresses a wind forecasting based penalty cost minimization solution for hybrid wind‐battery farms. We choose six wind farm sites (three offshore and the other three onshore) to study machine intelligent forecasting based solutions and compare the performance of a wavelet‐Twin support vector regression (TSVR) based wind power forecasting model with ε ‐Twin support vector regression, Random forest, and Gradient boosted machines, for penalty cost minimization. We access the penalties that arise as power imbalances along with the battery system's cost. We find that TSVR based wind power forecasting method results in a minimum global operational cost for all the wind farm sites under study. Abstract : The decision making environment for a hybrid wind farm involves an appropriate operational strategy that results in a minimum cost over a period of time. A wind farm operating with its aim to deliver power to the grid often has auxiliary power support in {the} form of BESS which operates either in charging or discharging mode. For deploying the integrated solution, we consider a multi wind farm dispatch scenario where every wind farm under consideration tries to minimize its cost incurred. The Operational cost incurred to a wind farm operator is purely based on the available forecast schedules. The availability of wind power forecast for a definite time horizon is an important factor that must match with the market timing where the energy balances are cleared between 5 minutes to 6 hours. For wind generation stations participating in short‐term electricity markets, accurate forecasting method results in lower penalty costs and also auxiliary costs for the TSO.