基于GA-GNNM的极地光伏发电功率预测方法

为了更加准确有效地对极地光伏发电功率做出预测,提出一种基于GA-GNNM的极地光伏发电功率预测方法。首先对采集到的气候因素数据以及光伏发电数据中缺失、异常部分进行清洗归一化处理;通过最大相关最小冗余算法（MRMR）选择最佳的气候特征组合,构建多维气候特征数据集;并将其输入到K均值聚类算法中完成不同季节天气类型聚类划分,然后引入相对距离量化寒暖季不同天气类型下与预测日相似度高的发电功率;构建灰色神经融合模型（GNNM）,将灰色模型（GM）的微分方程映射入神经网络模型,并采用遗传优化算法（GA）对模型参数进行优化以避免局部最优,提高极地光伏预测算法的精度。最后,以南极恩克斯堡岛气候及光伏发电数据为例进行验证,算例分析的结果为极地第5个新站的建立奠定理论基础。

POLAR PHOTOVOLTAIC POWER FORECASTING METHOD BASED ON GA-GNNM

Photovoltaic power forecasting is one of the basics for optimal configuration and coordinated operation of the microgrid. However, due to the extreme conditions of the two seasons in the polar region： polar day and night, and more stormy snow, occasional snow fog weather etc., the polar photovoltaic power is greatly affected, with instability, uncertainty and non-linear characteristics. In order to forecasting the power of polar photovoltaic more accurately and effectively, a polar photovoltaic power forecasting method based on GA-GNNM is proposed. First, cleaning and normalizing abnormal and missing data, useing the maximum correlation minimum redundancy algorithm to select the best feature set, the K-means clustering algorithm is used to cluster the weather types of different seasons, and the relative distance is used to quantify the power with high similarity to the predicted day under different weather types. Construct a gray neural hybrid model （GNNM）, map the differential equations of the gray model into the neural network model, and use genetic optimization algorithm （GA） to optimize the model parameters to avoid local optimization and improve the accuracy of the polar energy prediction algorithm. Finally, the climate and photovoltaic power data of Enksburg Island in Antarctica are used as examples for verification. The results of the example analysis lay a theoretical foundation for the establishment of the fifth new station in the polar region.