摘要
提高光伏发电功率预测精度对保障智能电网安全稳定运行有重要意义;针对传统BP神经网络存在预测精度不高且收敛速度慢的弊端,提出一种基于粒子群(PSO)差分进化(DE)并行计算优化BP神经网络的光伏发电短期预测方法;首先分析影响因素重要程度,通过带权重的欧式距离筛选相似的训练样本集;其次,对粒子群分组,通过粒子群和差分进化混合算法对粒子组内和组间优化,以保证种群多样性、提高预测稳定和精度、避免局部最优;然后,建立预测模型,通过基于spark的内存计算平台,将PSO-DE-BP算法并行优化以提高算法运行效率;最后,根据不同天气类型的预测结果对模型进行分析验证,此方法比PSO-BP、BP算法模型具有更高的稳定性和预测精度。
Improving photovoltaic power prediction accuracy is of great significance to ensure the safe and stable operation of smart power grid.Aimed at the disadvantages of low prediction accuracy and slow convergence speed of traditional BP neural network,a short-term photovoltaic power prediction method based on particle swarm optimization(PSO)differential evolution(DE)parallel computing optimization BP neural network is proposed.Firstly,the influencing factor is analyzed,and the selects similar training sample sets are selected by the weighted Euclidean distance.Secondly,the particle swarm is grouped by the hybrid algorithm based on particle swarm and differential evolution to optimize within and between particle groups,so as to ensure the PSO diversity,improve the prediction accuracy,and avoid the local optimization.Then,the prediction model is established,and the PSO-DE-BP algorithm is parallelized through the Spark-based memory computing platform.Finally,the model is analyzed and verified by the prediction results of different types of weather.This method has higher stability and prediction accuracy than PSO-BP and BP algorithms.
作者
刘春芳
王攀攀
曹菲
LIU Chunfang;WANG Panpan;CAO Fei(NARI Group Limited Company(State Grid Electric Power Research Institute Company),Nanjing 211106,China;Jiangsu Linyang Renewable Energy Technology Co.,Ltd.,Nanjing 210019,China)
出处
《计算机测量与控制》
2023年第5期180-186,共7页
Computer Measurement &Control
基金
国家电网公司总部科技项目资助(5100-202113396A)。
关键词
光伏发电预测
BP神经网络
差分进化
粒子群分组
Spark并行计算
prediction for the photovoltaic power generation
BP neural network
differential evolution
particle group
spark parallel computing
