Research On Short-term Electricity Price Forecasting In Electricity Market

With the continuous advancement of the power reformation,the marketization of power is an irresistible trend in the development of the power industry.Under the deregulated power market,electricity can be freely traded in the m arket environment as a commodity.In this context,electricity prices are at the core of the market.Accurate electricity price forcasting has also become the focus of research among market participants,electricity decision makers,and power manufacturers.They use electricity price prediction to adjust production plans and make related decisions in real time to achieve optimal allocation of resources and maximization of benefits.However,electricity prices have strong instability.Because the supply and demand of electricity is affected by weather,load,fuel type,and season,which causes real-time fluctuations in electricity prices,makes forecasting difficult.Therefore,electricity price prediction is a very challenging task.The article describes the basic concepts of the electricity market,introduces the relevant theories of electricity price forecasting,the selection and preprocessing of electricity price data,and the basic model of electricity price pre diction.Based on the summary of existing electricity price prediction methods,this paper proposes a combined prediction model based on the Singular Spectrum Analysis and the combination of Long-short Term Memory networks and BP neural networks.The main innovation of this paper is to propose a new short-term electricity price hybrid forcasting model by using the Singular Spectrum Analysis(SSA)method to decompose and reconstruct the original electricity price sequence based on a combination model of Long-short Term Memory network(LSTM)and BP neural network.For the non-stationary and non-linear characteristics of the original electricity price sequence,a Singular Spectrum Analysis method was used to pre-process the original electricity price sequence.First,the original electricity price seque nce was decomposed into several sub-sequences containing a trend component,an oscillation component and a noise component,and reconstructed this several sequences,which the noise components are removed.Long-short Term Memory network is used to predict the denoised reconstructed sequence 1,and the remaining reco nstructed sequences are predicted using BP neural network.Finally,the predicted values of all reconstructed sequences are superimposed and reconstructed to obtain the actual electricity price prediction result.In this paper,the measured electricity price data of the Queensland power market in Australia was taken as an sample.The electricity price data for the four seasons of spring,summer,autumn,and winter is used to do the one-hour advanced electicity price forcasting and obtain the corresponding prediction results and prediction errors.Compared with other existing models,the experimental results show that:(1)After the original electricity price sequence is decomposed and reconstruct ed by the Singular Spectrum Analysis method,the prediction accuracy is significantly improved compared with other decomposition methods.(2)The prediction effect of the combination model of Long-short Term Memory network and BP neural network is better than that of single model,which is effective and superior in short-term electricity price prediction.(3)The proposed combination model has great advantages over other existing hybrid models in terms of prediction accuracy.