基于时空相关性的光伏发电功率爬坡预测方法

Prediction Method of Photovoltaic Power Ramp Based on Spatiotemporal Correlation

传统的光伏发电功率预测方法爬坡预测可靠性较低,准确性不高。于是提出一种时空相关性的光伏发电功率爬坡预测方法。在典型日理想光伏发电出力归一化曲线提取基础上,采用线性插值方法生成光伏发电理想出力归一化曲线。通过蒙特卡洛法生成光伏发电随机分量,结合光伏发电与随机分量生成光伏发电序列;通过偏移爬坡率及变量状态划分方法构建信度网络节点变量和各节点变量的状态集;利用贪婪搜索算法从已有变量的状态集中获取最优信度网络结构后,进行光伏发电序列学习,完成光伏发电功率爬坡事件预测。实验结果表明,上述方法可有效完成光伏发电序列生成,并且爬坡预测可靠性较高,可实现多种气象条件下的光伏发电功率爬坡预测。

Traditionally,the photovoltaic power prediction method has low reliability and accuracy.Therefore,a time-space correlation method for photovoltaic power ramp prediction is presented in this work.According to the typical normalized curve of daily ideal photovoltaic power output,a linear interpolation method was introduced to generate the normalized curve of ideal photovoltaic power output.Based on Monte Carlo method,the random components of photovoltaic power generation were generated.Photovoltaic power generation and random components were used to generate the sequence of photovoltaic power generation.The node variables and the state set of each node variable in the reliability network were established by the method of shift climbing rate and variable state partition.Greedy search algorithm was utilized to obtain the optimal reliability of the network structure,and then photovoltaic power generation sequence learning was implemented to complete the prediction of photovoltaic power ramp events.The results show that the method has high prediction reliability,and can effectively complete the generation of photovoltaic power generation sequence and the climbing prediction of photovoltaic power generation under various meteorological conditions.