基於短期輻照度預測之智能太陽光電儲能系統控制策略

Abstract

近年來民眾環保意識上漲，因此再生能源的相關議題也逐漸浮上檯面，各國政府也極力推展再生能源發展計畫，但其功率輸出受到一天中的時間與氣候條件限制，造成供電間歇性的問題，而這項特性對電力行業提出了巨大的挑戰，因為它們必須維持電力的供應與需求平衡關係，以確保電力系統的穩定性和可靠性，而目前的主流解決方案有兩種，即是精確的再生能源發電預測與併入儲能系統平衡電網功率波動。因此本研究提出了一個基於短期太陽能輻照度預測的儲能系統控制決策，而該策略分為兩個部分，首先本研究提出了三種不同的預測模型比較，分別為Feedforward Neural Networks (FFNN)、Long Short-Term Memory (LSTM) 與 Gated Recurrent Unit (GRU)，並將資料集分為5分鐘與10分鐘間隔，用以訓練兩種不同間隔的預測模型，該模型以過去30分鐘的特徵資料做為訓練模型輸入，最後再將性能表現最好的兩種預測模型進行混合權重搭配，提出一個15分鐘的預測模型，並將其預測結果搭配第二階段基於模糊邏輯的儲能系統控制策略，其考慮了儲能系統的充電狀態、微電網的淨功率以及未來輻照度的變化，且策略的主要目標為減少再生能源發電所造成的不穩定性影響，並最大程度地降低與主電網間的功率波動，進而提高光伏系統發電的利用效率，並同時降低運營成本，更重要的是所有的決策控制皆維持在儲能系統之安全充電範圍內，所提出的策略將在實時數位模擬器進行實際模擬驗證以評估其效能與可靠性。

This study proposes a fuzzy logic control strategy (FLCS) for an energy storage system (ESS) based on short-term irradiance prediction. The power output of photovoltaic (PV) systems is intermittent, which posts a great challenge to electric power industries, because they have to balance the energy supply and demand to ensure the stability and reliability of a power system. The control strategy is divided into two parts. First, a solar irradiance prediction model is proposed based on FFNN, LSTM, and GRU models, which uses PV model parameters and features of satellite cloud images as the model inputs. The important features of satellite cloud images are selected by the Principal Component Analysis, and the filtered features are used to train the prediction model. Finally, the best prediction model is used to combine with the FLCS which takes a number of factors into consideration, including the state of charge of the ESS, the microgrid net power, and the change of the future irradiance. The goals of the control strategy are to reduce the impact of instability on renewable energy (RE) generation and minimize the grid power profile fluctuations. Finally, irradiance data are simulated by Real Time Digital Simulator (RTDS) to evaluate which control strategies can yield best performance. The results show that the proposed FLCS can control the ESS to balance the power of the grid no matter in the case of sunny and cloudy days, and more importantly, all decisions are maintained within a safe SOC range.