基于PTA-MTL的园区综合能源系统多元负荷预测

Multivariate load prediction of park integrated energy system based on PTA-MTL

园区综合能源系统中多异质能流深度融合,精准的负荷预测是实现系统容量配置与优化调度的前提。提出一种基于多任务学习的多元负荷预测方法。首先,采用最大信息系数筛选重要耦合特征作为模型输入,以降低噪声;其次,构建基于改进粒子群算法优化的时间卷积网络和注意力机制相结合的多元负荷预测模型(PTA-MTL),以实现冷、热、电负荷的联合预测;最后,实验分析表明,所提模型不仅具有较高的预测精度,还具备较快的运行速度。

In the park integrated energy system,multiple heterogeneous energy flows are deeply integrated,and accurate load prediction is the premise of realizing system capacity allocation and optimal scheduling.Therefore,a multivariate load prediction method based on multi-task learning is proposed.Firstly,the maximum information coefficient is used to screen the important coupling features as the model input to reduce the noise.Then,a multivariate load prediction model(PTA-MTL)based on a combination of temporal convolutional network optimized by improved particle swarm algorithm and attention mechanism is constructed,so as to achieve the joint prediction of cold,heat and electricity load.The experimental analysis shows that the proposed model not only has high prediction accuracy,but also has fast running speed.