A Novel Combined Electricity Price Forecasting Method Based on Data Driven

Abstract

In the deregulated electricity market, accurate knowledge of electricity price tend helps maximize the profitability of the participants in the electricity market, so electricity price forecasts become extremely important. On the basis of not considering the situation of the electricity market itself and many factors affecting the electricity price, the historical load and electricity price are used as inputs to predict the electricity price from the perspective of data driven. First Lasso, random forests and Gradient Boosting the SVM and BP neural network model and LSTM six kinds of single algorithm electricity price models are construct, and then the combination model of Lasso, BP neural network and LSTM combining six algorithms and the combination model of BP neural network model combining three kinds of algorithm are constructed. The actual electricity price and load data from Queensland were used for simulation, and the simulation results shows that: Among the single algorithmic electricity price models, the LSTM model has the highest accuracy. BP neural network is suitable for combining the single algorithm electricity price model. The BP neural network model with the combination of the six algorithms has the highest accuracy.