基于高斯变异粒子群优化的短期负荷预测

Short-Term Load Forecasting Based on PSO with Gaussian Variation

针对现有短时预测方法精度不高及电网负荷数据不确定性变化的问题,提出一种基于高斯变异粒子群优化(GPSO)的长短时记忆神经网络(LSTM)负荷预测模型,实现对短时负荷数据的高精度预测。方案首先对负荷序列数据进行预处理,提升数据之间的相关性。进一步引入非线性惯性权重加速粒子收敛速度,同时结合自适应高斯变异操作减小粒子陷入局部最优的风险,从而提升了PSO算法的寻优能力。实验结果证明,改进的粒子群优化算法能够提升LSTM模型的预测性能,验证了提出方法的有效性。与已有的预测模型相比,GPSO-LSTM模型有着更优的预测能力。

Considering the problems of inaccurate load forecasting and uncertain load data changes, a long-short-term memory neural network(LSTM) prediction model based on particle swarm optimization with Gaussian variation(GPSO) is proposed. The proposed scheme first preprocessed the load sequence data to improve the correlation between the data. Furthermore, nonlinear inertia weight was introduced to accelerate the convergence speed of particles, and adaptive Gaussian variation operations were used to reduce the risk of particles falling into local optimum. The optimization capability of PSO algorithm was finally further improved. Experimental results demonstrate that the improved particle swarm optimization algorithm can improve the prediction performance of LSTM-based model and thus verifies the effectiveness of the proposed method. Compared with the existing prediction models, the GPSO-LSTM shows the superior prediction performance.