摘要
为了降低可再生能源机组波动性对电网运行的影响,基于模型预测的思想对当前应用于功率预测的自适应小波网络(AWNN)参数迭代求解方法进行了优化。使用与差分进化相融合的改进粒子群算法PSO-DE来替代传统梯度下降算法,优化了AWNN网络的迭代方式。PSO-DE算法一方面借助差分进化算法中的遗传、变异及交叉机制提升了粒子种群间的信息流通效率;另一方面则通过惯性权重因子和束缚机制将粒子束缚在指定区间内波动,从而避免了算法在优化求解时陷入局部最优解的情况。基于广东某地区的光伏发电数据集进行了算法仿真,结果表明在引入PSO-DE算法后模型的主要性能指标显著提升,有效提高了可再生能源的功率预测精度。
In order to reduce the impact of renewable energy unit volatility on power grid operation,the current Adaptive Wavelet Neural Network(AWNN)parameter iterative solution method applied to power prediction is optimized based on the idea of model prediction.The improved particle swarm optimization PSO-DE combined with differential evolution is used to replace the traditional gradient descent algorithm to optimize the iterative mode of AWNN network.PSO-DE algorithm improves the efficiency of information flow among particle populations with the help of genetic,mutation and crossover mechanisms in differential evolution algorithm;On the other hand,the particles are bound to fluctuate in the specified interval through inertia weight factor and binding mechanism,so as to avoid falling into the local optimal solution in the optimization solution.The algorithm is simulated based on the photovoltaic power generation data set in a region of Guangdong.The results show that the main performance indexes of the model are significantly improved after the introduction of PSO-DE algorithm,and the power prediction accuracy of renewable energy is effectively improved.
作者
刘岩
李雨森
张夕佳
黄红伟
毛文照
LIU Yan;LI Yusen;ZHANG Xijia;HUANG Hongwei;MAO Wenzhao(Shenzhen Power Supply Company,Shenzhen 518000,China;Beijing QU Creative Technology Co.,Ltd.,Beijing 100102,China)
出处
《电子设计工程》
2023年第9期146-150,共5页
Electronic Design Engineering
基金
南方电网技改项目(090000GS62191470)。
