基于随机森林和卷积神经网络的风-光伏-抽水蓄能电站联合优化运行

Combined Optimal Operation of Wind-photovoltaic-pumped Storage Power Station Based on Random Forest and Convolutional Neural Network

风力发电和光伏发电的输出波动性、间歇性和不确定性,将加剧风光发电大规模并网的困难。针对以上问题,综合考虑环境温度、地表水平辐射、直接辐射和散射辐射四个因素,利用随机森林模型预测光伏发电输出功率;利用卷积神经网络卷积核参数共享与短期信息提取较好的特点,根据地面风速,风机轮毂中心风速,风向和气压数据,预测风电输出功率,预测结果误差小,最后利用BP神经网络算法具有较强容错和泛化能力特点对风力发电和光伏发电输出功率联合优化,得到总输出功率以及抽水蓄能发电功率。预测得到的最终总输出功率较为稳定,预测误差在可接受范围内,大大降低了风光发电并网给系统带来的不稳定性。

The fluctuation,intermittency and uncertainty of the output of wind power and photovoltaic power generation will aggravate the difficulty of large-scale grid integration of wind and solar power generation.In view of the above problems,by comprehensively considering ambient temperature,surface level radiation,direct radiation and scattered radiation,the random forest model is used to predict the output power of photovoltaic power generation,and the convolutional neural network with convolution kernel parameter sharing and short-term information was adopted to extract good characteristics based on the ground wind speed,wind hub center wind speed,wind direction and air pressure data to predict the wind power output power,the prediction result error of which is small,and finally the BP neural network with strong fault tolerance and generalization ability characteristics was adopted to jointly optimize the output power of wind power generation and photovoltaic power generation,obtain the total output power and pumped storage power generation,and predict the final total output power is relatively stable.The results showed that the prediction error is within the acceptable range.It greatly reduces the instability brought by the wind and solar grid connection to the system.