电力市场放松管制中的位置边际定价（LMP）是来自节点（位置）的竞标竞争的结果。本文介绍了一种深度学习神经网络，用于使用其他分区价格对PJM电力市场中某个区域的特定节点进行24小时提前LMP预测。提出的方法使用卷积神经网络（CNN），这是一种深度学习的著名架构，已成功应用于图像分类和模式识别。CNN在本文中用于通过卷积，核和合并操作提取实时电价的特征，这些操作使用各个区域和位置的时空电价。使用分步手动方法确定模型的最佳参数，以最大程度地减少暴力试验和错误方法的负担。来自PJM电力市场的数据集用于仿真中，其中使用Tensorflow和Keras库应用所提出的方法。对各种方法进行了比较，结果表明，该方法在24小时超前LMP预测中具有比传统方法更高的准确性。
Locational marginal pricing (LMP) in a deregulated electricity market is a result of bidding competition from nodes (locations). This paper presents a deep learning neural network for the 24 h-ahead LMP forecasting of a specific node in a zone at the PJM power market using other zonal prices. The presented method uses a Convolutional Neural Network (CNN), which is a well-known architecture for deep learning that has been applied successfully for image classification and pattern recognition. The CNN is used herein to extract features of real-time electricity prices by convolution, kernel, and pooling operations using spatiotemporal electricity prices in various zones and locations. The optimal parameters of the model were determined using a step-by-step manual approach to minimize the burden of brute force trial and error methods. Datasets from the PJM power market are used in simulations in which the proposed method is applied using Tensorflow and Keras Library. Various approaches are compared, and the results thus obtained demonstrate that the proposed method has higher accuracy than traditional methods in 24 h-ahead LMP forecasting.
