基于WNR-CLSSA-LSTM的短期电力负荷预测

Short-term power load prediction based on WNR-CLSSA-LSTM

准确的短期电力负荷预测(STPLF)在智能电网的日常运行中起着重要作用。为了更有效地预测短期负荷,本文提出了一种基于小波去噪、改进樽海鞘群优化算法(SSA)和长短期记忆网络(LSTM)的预测方法。首先,通过小波去噪降噪;其次,种群的初始位置采用混沌初始化Cubic策略,并将莱维飞行策略引入樽海鞘群领导者和跟随者的位置更新中,接着将跟随者的更新公式引入每个维度的最优适应度位置维度,加快收敛速度;然后,利用改进的SSA算法优化LSTM模型的参数,得到STPLF结果。通过实验比较改进的CLSSA-LSTM与GA-LSTM、PSO-LSTM、SSA-LSTM和单一的LSTM,结果表明,改进的CLSSA-LSTM预测效果优于其他算法优化的LSTM。同时,将CLSSA-LSTM模型与不同的预测模型PSO-SVR、GA-BP对比,均有不错的表现。因此本文提出的预测模型是一种有效的STPLF工具。

Accurate short-term power load prediction(STPLF)plays an important role in the daily operation of smart grids.In order to more effectively predict short-term load,this paper proposes a hybrid deep learning method based on the wavelet noise reduction,improvement of the Salp Swarm Algorithm(SSA)and long-term memory network(LSTM).First of all,the original data removes some noise through the method of wavelet noise reduction;secondly,the initial position of the population is used to initialize the chaos Cubic strategy,at the same time add the Levy flight strategy to position update of the leader and the follower,and the follower's update formula is introduced to the optimal adaptability position dimension of each dimension,so that the algorithm is accelerated to convergence;then the improved SSA algorithm optimizes the parameters of the LSTM model.The optimized LSTM is applied to short-term power load predictions to obtain the real STPLF results.Through the experimental comparison,the improved CLSSA-LSTM is compared with the GA-LSTM,PSO-LSTM,SSA-LSTM and a single LSTM.The experimental results show that the improved CLSSA-LSTM predictive effect is superior to other optimized LSTMs.Meanwhile,compared with different predicted models PSO-SVR and GA-BP,CLSSA-LSTM model has a good performance.Therefore,the hybrid depth learning method proposed in this article is an effective STPLF tool.