基于自适应樽海鞘算法优化BP的风光互补并网发电功率预测

为解决风光互补并网发电功率预测问题，针对前馈(BP)神经网络容易陷入局部最优而导致预测精度降低的问题，提出了一种自适应樽海鞘算法(ASSA)优化BP神经网络的风光互补并网发电功率预测模型。首先，在标准的樽海鞘算法(SSA)中引入动态权重策略和变异算子构建ASSA。其次，引入BP神经网络算法，构建BP神经网络的风光互补并网发电功率预测模型。最后，通过ASSA算法优化BP神经网络的权值和阈值，提出ASSA-BP的风光互补并网发电功率预测模型。仿真结果表明，利用ASSA-BP模型预测发电功率数据的相对误差小于BP模型预测数据的相对误差。ASSA-BP和SSA-BP的模型平均绝对误差数值更小，ASSA-BP模型的平均绝对误差最小，ASSA-BP模型的预测稳定性最强。该预测模型较传统风光互补并网发电功率预测方法有更高的精确度。

Wind and solar complementary grid-connected power generation prediction based on BP optimized by a swarm intelligence algorithm

There is a power prediction problem of wind solar complementary grid connected power generation in that the feedforward (BP) neural network can easily fall into a local optimization, resulting in the reduction of prediction accuracy. Thus an Adaptive Salp Swarm Algorithm (ASSA) optimized BP neural network wind solar complementary grid connected power prediction model is proposed. First, dynamic weight strategy and a mutation operator are introduced into the standard Salp Swarm Algorithm (SSA) to construct the ASSA. Secondly, a BP neural network algorithm is introduced to construct the wind solar complementary grid connected power prediction model. Finally, the weight and threshold of the BP network are optimized by the ASSA algorithm, and the power prediction model is thus proposed. The simulation results show that the relative error of power generation data predicted by the ASSA-BP model is less than that predicted by the BP model. The average absolute error of ASSA-BP and SSA-BP models is smaller, and the average absolute error of ASSA-BP model is the smallest, and the prediction stability of the ASSA-BP model is the strongest. The prediction model has higher accuracy than the traditional wind solar complementary grid connected power prediction method. This work is supported by Henan Joint Key Project of National Natural Science Foundation of China (No. U1804262).