Deep Neural Network Configuration Sensitivity Analysis in Wind Power Forecasting

Igor Kuzle, Josip Đaković, Bojan Franc, Yongqian Liu

Journal of Energy - Energija · 16 August 2022

In one sentence. Deep recurrent networks trained on Croatian wind farm output and NWP forecasts are compared with operational tools, with a focus on configuration sensitivity.

Abstract

The trend toward increasing integration of wind farms into the power system is a challenge for transmission and distribution system operators and electricity market operators. The variability of electricity generation from wind farms increases the requirements for the flexibility needed for the reliable and stable operation of the power system. Operating a power system with a high share of renewables requires advanced generation and consumption forecasting methods to ensure the reliable and economical operation of the system. Installed wind power capacities require advanced techniques to monitor and control such data-rich power systems. The rapid development of advanced artificial neural networks and data processing capabilities offers numerous potential applications. The effectiveness of advanced deep recurrent neural networks with long-term memory is constantly being demonstrated for learning complex temporal sequence-to-sequence dependencies. This paper presents the application of deep learning methods to wind power production forecasting. The models are trained using historical wind farm generation measurements and NWP weather forecasts for the areas of Croatian wind farms. Furthermore, a comparison of the accuracy of the proposed models with currently used forecasting tools is presented.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.