基于CNN-Attention-BiLSTM的碳化硅企业负荷预测与可调节潜力分析

Load forecasting and adjustable potential analysis of silicon carbide enterprises based on CNN-Attention-BiLSTM

中文关键词: 负荷预测  可调节潜力  碳化硅企业  注意力机制  卷积神经网络  双向长短期记忆网络

英文关键词: load forecasting  adjustable potential  silicon carbide enterprises  Attention mechanism  convolutional neural network  Bi-directional long short-term memory

中文摘要:

      工业负荷在社会用电中占比量巨大，可调节负荷资源丰富，对其可调节潜力分析势在必行。由于工业负荷变化率大，负荷尖端多，对其实时预测可调节潜力难度较大。因此，选取典型工业企业碳化硅行业进行建立可调潜力推演模型，首先，通过Person 关联分析法考虑天气特征以及电价因子对该企业负荷的影响，同时建立通过卷积神经网络（convolutional neural network，CNN）与注意力机制（Attention）处理后的双向长短期记忆网络（Bi-directional long short-term memory，BiLSTM）预测模型，利用模型预测结果对碳化硅企业进行可调节潜力挖掘。为了验证该方法的有效性，通过建立不同算法进行对比以及不同策略下的可调潜力结果，该算法显著优于其他对比算法，同时3种策略的可调潜力结果能加深电网认识该类企业的负荷特性。

英文摘要:

      Industrial load accounts for a large proportion of social electricity consumption, and the adjustable load resources are abundant,so it is imperative to analyze its adjustable potential. Due to the large change rate of industrial load and many load tips, it is difficult to predict the adjustable potential in real time. Therefore, the silicon carbide industry of a typical industrial enterprise is selected to establish an adjustable potential deduction model. First, the impact of weather characteristics and electricity price factors on the enterprise load is considered through the Person correlation analysis method. At the same time, the Bi-directional long short-term memory（BiLSTM）prediction model processed by convolutional neural network（CNN）and Attention mechanism is established. The adjustable potential of silicon carbide enterprises is explored by using the model prediction results. In order to verify the effectiveness of this method, this algorithm is significantly superior to other comparison algorithms by establishing different algorithms for comparison and the tunable potential results under different strategies. Meanwhile, the tunable potential results of the three strategies can deepen the power grid’s understanding of the load characteristics of such enterprises.