基於短期風力預測估算最大電能擷取之大型風機迎風轉向控制估測研究

Abstract

此篇論文以2MW之大型風力發電機為研究載具，發展提升其發電效率之控制法則，提出以前饋倒傳遞類神經網路 (Neural Network)預測短期風向、風速，計算風能變化，整合風機迎風轉向系統的響應時間及推導風機轉向耗能與風能擷取分析，使迎風轉向反應時間較慢的大型風力發電機可依據短期風力風向預測資料，做適當的風機迎風轉向控制，進而達到在降低風機轉向耗能之下擷取最大風能之目的，以節省不必要的偏向運轉。本論文研究方法由長期的風力觀測資料庫，以前饋倒傳遞類神經網路進行短期風速與風向預測，並與歷史資料比較其準確性。由風速與風向預測結果推導在風能擷取與風機轉向耗能正負相抵間之最高電能擷取功率的轉向機制。研究結果顯示，本論文提出之迎風轉向控制法則在平均風速約為5.924m/s的風場下，可比隨風向迎風轉向之風機多出約1.1kWh的淨獲能。

Based on 2MW large-scale wind turbine platform, this research proposed feed-forward back-propagation neural network (NN) algorithm to predict short-term wind direction and wind speed for calculating the wind energy's variation. The estimation results of maximum power capture as well as the analysis of nacelle veering energy consumption are integrated with the response time of wind turbine yaw operation to derive wind energy gains and losses in order to make the appropriate wind turbine yaw control by predicting short-term wind data and finding out the best yaw angle to capture maximum wind energy which included the loss of consumption for turning orientation. This yaw control algorithm is expected to reduce unnecessary turning of large wind turbines which need longer time to change orientation. In additions, the prediction methods based on various structures of the feed-forward neural network from the values of meteorological variables are also described in this work. The result show that the proposed algorithm could have a surplus of about 1.1kWh net energy per five minute compare to the same type specification wind turbine that yaw directly follow with wind direction under the condition of 5.924m/s wind velocity. Finally, there is a simulated human machine interface for showing the conditions of controlled turbine facing and the real wind direction and velocity.