基于小波去噪-KPCA神经网络的光伏功率预测方法

为了提高光伏发电功率的预测精度,提出了一种基于小波去噪/聚类/核主成分分析(KPCA)神经网络的光伏发电功率预测方法。首先,应用二维小波阈值去噪法预处理光伏出力数据;然后,应用k-means算法将预测模型分为4种不同模式下的子预测模型;引入KPCA对输入空间降维重构,利用粒子群优化(PSO)神经网络算法建立基于聚类/KPCA/神经网络的光伏发电功率预测模型。采用某光伏电站的实例数据进行预测分析,结果表明该模型实现了不同模式下的光伏出力较为精准的预测,显示出良好的预测性能,验证了预测模型的可行性和有效性。

Forecasting method of photovoltaic output power based on wavelet denoising/KPCA/PSOBP

In order to improve the forecasting accuracy of photovoltaic (PV) output power, a forecasting method is proposed based on k-means, KPCA and PSO-BP. Firstly, wavelet threshold de-noising algorithm is used to pretreat PV output data. Then, the k-means clustering algorithm is applied to divide the forecasting model into four sub models under different modes. The Kernel Principal Component Analysis (KPCA) method is used to reduce the dimensionality of the input space. Neural network algorithm is optimized based on Particle Swarm Optimization (PSO). Finally, the PV output power forecasting model based on k-means /KPCA/PSO-BP is established. The example data is used to verify the forecasting model, the results show that it can forecast the PV output power accurately in different modes and have good forecasting performance.#$NLKeywords: wavelet denoising; kernel principal component analysis; particle swarm optimization; neural network; photovoltaic forecast