Short-term electric load forecasting using echo state networks and PCA decomposition

In this paper we approach the problem of forecasting a time-series of electrical load measured on the ACEA power grid, the company managing the electricity distribution in the city of Rome – Italy, with an Echo State Network considering two different leading times of 10 minutes and 1 day. We use a standard approach for predicting the load in the next 10 minutes while, for a forecast horizon of one day, we represent the data with a high dimensional multi-variate time-series, where the number of variables is equal to the quantity of measurements registered in a day. Through the orthogonal transformation returned by PCA decomposition, we reduce the dimensionality of the TS to a lower number k of distinct variables: this allow us to cast the original prediction problem in k different 1-step ahead predictions. The overall forecast can be effectively managed by k distinct prediction models, whose outputs are combined together to obtain the final result. We employ a genetic algorithm for tuning the parameters of the Echo State Network and we compare its prediction accuracy with a standard ARIMA model.