基于LSSVR和LSTM的多模型优化集成负荷预测

针对负荷需求受多源因素影响和现有单模型预测方法精度较低的问题,提出了一种基于最小二乘支持向量回归(LSSVR)和长短期记忆循环神经网络(LSTM)的多模型优化集成负荷预测方法。首先探究负荷相关特征的特性并由互信息进行特征选择,获取最优特征集。在此基础上采用随机抽样(bootstrap)生成多个训练集,然后使用具有良好预测能力的LSSVR和LSTM模型对多个训练集分别进行预测。利用混沌粒子群优化算法(CPSO)进一步提高模型预测精度。最后,在决策阶段中使用偏最小二乘回归(PLSR)组合各个子模型的最优预测输出并提供最终预测结果。对真实电网数据进行了仿真,并与其它预测方法进行了比较。本文所提方法的应用范围广泛且预测精度提高显著。

Multi-Model Optimization Integrated Load Forecasting Method Based on LSSVR and LSTM

To solve the problems that load demand is affected by multiple factors and the accuracy of existing single model prediction method is low,a multi-model optimization integrated load prediction method based on least square support vector regression(LSSVR)and long and short memory cyclic neural network(LSTM)is proposed.Firstly,the characteristics of load-related features were explored and selected based on mutual information to obtain the best features.On this basis,multiple training sets were generated using bootstrap,and then LSSVR and LSTM models with good predictive ability were used to predict multiple training sets.Chaos particle swarm optimization(CPSO)was used to improve the prediction accuracy.Finally,partial least squares regression(PLSR)was used in the decision-making stage to combine the optimal prediction output of each sub-model and provide the final prediction results.The real grid data were simulated and compared with other prediction methods.The method proposed in this paper has a wide range of applications and a remarkable improvement in prediction accuracy.