基于相似日和神经网络的光伏发电预测

Photovoltaic generation prediction based on similar days and neural network

光伏发电系统的输出功率受到季节、太阳辐射强度、温度和湿度等气象条件影响,呈现出时变性、间歇性和随机性。文章提出了基于相似日原理和改进的BP神经网络预测方法,利用光伏电站的历史气象信息建立气象特征向量,基于曼哈顿距离寻找相似日,根据给定的不同预测日选取3个相似日的输出功率作为预测模型输入,直接预测发电站的输出功率。以某光伏电站为例进行建模预测,并通过预测误差分析证明了算法的有效性。

Output power of photovoltaic （PV） power generating system has the characteristics of time- varying, intermittence and randomness due to the various meteorological factors such as season, solar radiation, temperature, humidity, etc. In this paper, a forecasting method is proposed based on the principle of similar days and BP neural network. By using historical weather information from the solar power station, meteorological feature vectors are established, and similar days are found based on Manhattan distance. According to the given different forecasting day, output power of three similar days would be chosen as inputs of the forecasting model, and then the output power of generating station can be predicted directly. A forecasting model is made based on a photovoltaic power station and the forecast error is calculated and analyzed. The results show the validity of the algorithm.