小波优化多任务学习的综合能源负荷预测

Integrated energy load forecasting based on multi-task learning and wavelet optimization

为实现多元负荷快速、准确短期预测,提出一种小波优化多任务学习的综合能源负荷预测方法.采用多任务学习神经网络方法,设置参数的软共享机制,不仅可以提高模型泛化能力,而且可以提高综合能源系统负荷预测方法鲁棒性.通过设置特殊的隐藏层激活函数,使得预测负荷时,保证网络解唯一性,防止局部极小值点,提高收敛速度,保证其综合能源系统电、气、热负荷预测误差具有更好的收敛性.研究结果表明:小波优化多任务学习具有有效性与可行性.研究结论小波优化能够防止局部极小值点,有助于提高预测速度.

In order to realize fast and accurate short-term prediction of multivariate load,this paper proposes a comprehensive energy load forecasting method based on wavelet optimized multi-task learning.By adopting the multi-task learning neural network method and setting the soft sharing mechanism of parameters,this method can not only improve the generalization ability of the model,but also improve the robustness of the integrated energy system load forecasting method.In addition,by setting a special hidden layer activation function,this paper not only guarantees the uniqueness of the network solution and prevents the local minimum point,but also improves the convergence speed to ensure the prediction error of the electric,gas and heat load of the integrated energy system.The simulation results show the effectiveness and feasibility of multi-task learning and wavelet optimization.This study concludes that the wavelet optimization can prevent local minimum point and improve the prediction speed.