Locational Marginal Price Forecasting Based on Deep Neural Networks and Prophet Techniques

In many of the electricity markets in North America, the electricity prices are in terms of the locational marginal price (LMP) which reflects the cost of supplying the next MWh of electricity at a bus, considering transmission constraints. Electricity price forecasting provides vital information on system conditions to the independent system operator (ISO). It indicates important signals pertaining to the need of investing in the new generation, upgrading transmission, or reducing electricity consumption. Power suppliers and consumers use the forecasted price to optimize the profit in the day-ahead market and bilateral contracts. Facility owners rely on the forecasted price to make investment decisions. Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks are applied to improve short-term (time series) LMP forecasting accuracy. Dataset from the ISO-NE power market is utilized in modeling and analysis in which the proposed techniques are applied using Matlab and Python software. Various methods are evaluated and compared, and the conclusions achieved show that LSTM has lower error rates and higher accuracy than the Prophet forecasting model in 24 h-ahead LMP forecasting.