基于双层优化VMD-LSTM的农村超短期电力负荷预测

Ultra-short-term Power Load Forecasting Based on Two-layer Optimization VMD-LSTM

稳定的供电是农村发展建设的有力保障,而电力负荷水平是建设效果的重要衡量标准,因此建立精确的负荷预测模型可以更准确直观显现电力负荷情况,为供电公司制定决策提供有力支撑。由于LSTM负荷预测模型在数据预测方面存在收敛性差、预测精度不高等问题,为提高模型的预测精度,提出一种基于双层优化VMD-LSTM的超短期电力负荷预测方法。首先提出麻雀算法优化变分模态分解(sparrow variational mode decomposition,SVMD),通过SVMD将原始数据转化为模态分量(intrinsic mode functions,IMF);其次采用改进樽海鞘群算法(association salp swarm algorithm,ASSSA)优化LSTM模型。通过引入4种策略增强标准樽海鞘算法优化能力;最后将各模态分量分别代入到新模型并进行叠加预测。选取辽宁省某市某乡村10kV变压器真实历史负荷数据,以均方根误差(RMSE)、平均绝对误差(MAE)、平均绝对百分比误差(MAPE)、拟合度(R^(2))作为评价指标,并与其他基础预测模型进行对比,结果表明,改进后的算法在计算精度、稳定性方面均优于其他基础预测模型。

Stable power supply is a guarantee for rural development and construction,and the level of power load is an important measure of the construction effect.Therefore,establishing a precise load prediction model can more accurately and intuitively show the power load,and provide a strong support for the formulation of decision-making for power supply companies.Since the LSTM load forecasting model has problems such as poor convergence and low forecasting accuracy in data forecasting,in order to improve the forecasting accuracy of the model,an ultra-short-term power load forecasting method based on two-layer optimization VMD-LSTM is proposed.First,the sparrow algorithm is proposed to optimize the variational mode decomposition(sparrow variational mode decomposition,SVMD),and the original data is converted into modal components(intrinsic mode functions,IMF) through SVMD;secondly,the improved salp swarm algorithm(association salp swarm algorithm,ASSSA) to optimize the LSTM model.The optimization ability of the standard salp algorithm is enhanced by introducing four strategies;finally,each modal component is substituted into the new model and superimposed prediction is performed.The real historical load data of 10 kV transformer in a certain city and village in Liaoning Province is chosen,the root mean square error(RMSE),average absolute error(MAE),average absolute percentage error(MAPE),and fitting degree(R^(2)) are taken as evaluation indicators,and compared with other basic prediction models.The results show that the improved algorithm is superior to other basic prediction models in terms of calculation accuracy and stability.