End-to-end probabilistic forecasting of electricity price via convolutional neural network and label distribution learning

Abstract

With the advancement of power market reforms, electricity price prediction has attracted increasing attention. This paper proposes a novel probabilistic forecasting approach based on deep neural network for electricity prices. Firstly, reasonable price distributions are constructed from historical data based on the nearest neighbors. Then, a deep convolutional neural network(DCNN) is employed to extract high-level features. Meanwhile, these features are fed to label distribution learning forests (LDLFs) to generate probabilistic forecasts. The proposed framework, dubbed DCNN–LDLFs, can jointly learn the price distributions. The DCNN–LDLFs provides three types of forecasts, including deterministic forecasts, prediction intervals (PIs), and probability density functions. Unlike most popular models, the DCNN–LDLFs can be trained in an end-to-end manner, which has the opportunity to obtain a globally optimal solution. The case study on Singapore shows that the proposed method provides superior forecasts over the existing approaches.

Keywords

• Probabilistic forecasting
• End-to-end training
• Deep convolutional neural network
• Label distribution learning forests
• Electricity price