基于云计算和智能优化SVR的光伏发电功率预测

Cloud computing and intelligent optimized SVR-based power prediction of photovoltaic power generation

为提高光伏发电功率的预测精度,针对支持向量机回归(Support Vector Regression,SVR)模型的预测结果易受其惩罚系数C、敏感损失函数的最大误差系数ε和核函数g影响的问题,提出一种基于新型智能算法-蝗虫算法优化SVR模型参数的光伏发电功率预测模型。由于光伏发电功率数据存在随机性和间隙性的特征,Multi-Agent和分布式思想被引入蝗虫算法优化SVR模型,通过将云计算的MapReduce框架和GOA-SVR结合,提出一种基于MapReduce和GOA-SVR并行化的光伏发电功率预测模型(MapReduce and GOA-SVR,MR-GOA-SVR),从而提高海量高维光伏发电数据的处理能力。将影响光伏输出功率的11个气象因素作为GOA-SVR的输入向量,光伏输出功率作为GOA-SVR的输出向量,建立GOA-SVR的光伏发电功率预测模型。研究结果表明:MR-GOA-SVR可以有效提高不同天气类型下的光伏发电功率的预测精度,具有很强的现实性和指导意义。与PSO-SVR、GA-SVR、GOA-SVR和SVR相比,MR-GOA-SVR在晴天、阴天和雨天均可以提高预测精度,且具有优异的并行性能。

In order to improve the power prediction accuracy of photovoltaic power generation,a power prediction model of photovoltaic power generation based on the parameters of SVR(support vector machine regression)model optimized by the new intelligent algorithm--grasshopper optimization algorithm is proposed herein for the problem of that the prediction result from SVR model is vulnerable to its penalty coefficient C,the maximum error coefficient of the sensitive loss functionε,and the kernel function g.Due to the characteristics of the randomness and intermittency of the power data of photovoltaic power generation,the multi-agent and distributed idea are introduced into grasshopper optimization algorithm for optimizing SVR model,and then a MapReduce and GOA-SVR parallelization-based power prediction model of photovoltaic power generation(MapReduce and GOA-SVR,MR-GOA-SVR)is proposed through the combining the MapReduce framework of cloud computing with GOA-SVR,thus the processing capacity of the massive high dimensional data of photovoltaic power generation is enhanced.By taking 11meteorological factors affecting the photovoltaic output power as the input vectors and the photovoltaic output power as the input vector of GOA-SVR,the GOA-SVR power prediction model of photovoltaic power generation is established.The study result shows that MR-GOA-SVR can effectively improve the accuracy of the power prediction of photovoltaic power generation under various types of weather with strong realistic and guiding significances.If compared with PSO-SVR,GA-SVR,GOA-SVR and SVR,the MR-GOA-SVR can improve the accuracy of the power prediction on all sunny,cloudy,and rainy days and has an excellent parallel performance.