基于PSO-DE-BP的光伏发电功率短期预测

Short Term Prediction of Photovoltaic Power Generation Based on PSO-DE-BP

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。

Abstract:

The improvement of photovoltaic power prediction accuracy is of great significance to ensure the safe and stable operation of smart grid.In order to solve the disadvantages of low prediction accuracy and slow convergence speed of traditional BP neural network, this paper proposes a short-term photovoltaic power prediction method based on particle swarm differential evolution(DE) parallel computing which optimizes BP Neural Network. First,this method analyzes the importance of the influencing factor and selects similar training sample sets through weighted Euclidean Distance.Second, the algorithm groups the Particle Swarm Optimization(PSO), and optimizesthe PSO internallyand externally via the hybrid algorithm of PSO and differential evolution,so as to ensure the PSO diversity.After that,the prediction model is established, and the PSO-DE-BP algorithm is parallelized through the memory computing platform based on spark.Finally, the model is analyzed and verified according to the prediction results of different weather types. This method has higher stability and prediction accuracy than PSO-BP and BP algorithm models.