風機動態增益輔助頻率控制器於系統故障時的應用

Abstract

本論文主要目的在於設計雙饋式感應風力發電機之類神經網路輔助頻率控制器，以改善電力系統於穩態運轉時遭受到一大型發電廠跳機之頻率響應，並提高系統頻率之最低點，增加整體電力系統的穩定度，避免造成用戶停電。 首先推導電力系統頻率控制非線性數學模型，再將其線性化進行小訊號分析，並利用參與率與特徵值靈敏度釐清系統狀態變數、系統特徵值與輔助頻率控制器參數之間的關係。接著利用非線性數學模型分析在不同跳機容量、風速及風機容量下之最佳動態增益參數。以這些當做類神經網路輔助頻率控制器之訓練樣本資料，再將訓練完的類神經網路輔助頻率控制器應用於風機上，與固定增益('K' _'PD' )控制器比較兩者間的系統動態響應，以驗證類神經網路輔助頻率控制器在系統遭遇跳機事故後，可以獲得較佳的頻率響應，提高系統的穩定度。 本論文藉由MATLAB®/Simulink軟體進行模擬，並以台灣西部地區電網之非線性數學模型為例，驗證所提出之類神經網路輔助頻率控制器的有效性。

To improve the dynamic frequency response for a power system subject to a generator trip, a supplementary frequency controller for a wind farm with doubly fed induction generator (DFIG) is designed using artificial neural network (ANN) in this thesis. The proposed controller can also improve system frequency nadir and stability. Service interruption due to generator trip can there be avoided. A nonlinear mathematical model for frequency control of the power system is first derived. The nonlinear model is then linearized for small signal analysis. Participation factors and eigenvalue sensitivity are analyzed in order to clarify the relationship between system state variables, eigenvalues and parameters of supplementary frequency controller. The optimal dynamic gains for the supplementary frequency controller under various operating conditions such as outage capacities, wind speeds, and wind farm capacities are obtained using nonlinear model simulations. There dynamic gains for different operating conditions are employed as the training patterns for the ANN. After the ANN has been trained using there training patterns, it is tested with both case within the training set and cases outside the training set. It is concluded that the proposed ANN based frequency controller can yield better dynamic frequency response than the fixed gain controller under various operating conditions. MATLAB®/Simulink is employed for digital simulations and the nonlinear mathematical model of the power grid in western Taiwan is taken as an example to verify the effectiveness of the proposed neural network supplementary frequency controller.