基于DA-RKELM算法的光伏发电功率预测方法

针对光伏发电功率具有的波动性和随机性等特点造成的电网安全问题,提出了一种基于蜻蜓算法优化的正则核极限学习机光伏发电功率预测方法。通过相关性分析确定影响光伏发电功率的关键影响因子,构建光伏发电功率预测模型;利用蜻蜓算法获取网络最优的权重和阈值,在标准极限学习基础上引入正则化函数和核函数以避免传统梯度下降法造成的过拟合问题,增强模型空间映射能力;仿真实验表明,与DA-ELM、PSO-ELM以及标准-DA-ELM模型相比,DA-RKELM预测模型能达到更高的预测精度,更贴近光伏发电的实际运行功率。

A Photovoltaic Power Forecasting Method Based on DA-RKELM Algorithm

Aiming at the power grid safety problems caused by the fluctuation and randomness of photo-voltaic power generation, a method for predicting photo-voltaic power generation of a regular nuclear limit learning machine based on the optimization of a dragonfly algorithm was proposed. Through correlation analysis, the key factors affecting the photo-voltaic power generation are determined, and the photo-voltaic power prediction model is constructed. Dragonfly algorithm is used to obtain the optimal weight and threshold value of the network, and regularization function and kernel function are introduced based on the standard limit learning machine to avoid the over fitting problem caused by the traditional gradient descent method and enhance the spatial mapping ability of the model. Simulation experiments show that compared with DA-ELM, PSO-ELM and GA- ELM models, the DA-RKELM prediction model achieve higher prediction accuracy, closer to the actual operating power of photo-voltaic power generation.