Compressed sensing and P1P in electrical consumption prediction

© 2018 IEEE. The proposal of this current work is to take advantage of statistic and periodic attributes of Electrical Energy Consumption (EEC) by using Potential Polynomials of Degree One (P1P) in estimation and forecast of parameters related with load curves obtained from different student buildings. Furthermore, considering that electrical energy measurement vectors present compress characteristics by projecting them in different vectorial spaces (called dictionaries too), this paper proposes use of the Compressed Sensing (CS) concept too. CS will be used to reduce data quantity previous model and prediction process. Identification and validation performance is analyzed by using metrics like Mean Absolute Percentage Error (MAPE) and a modified version of Mean Square Error (MSE) used in related studies. These values indicate promising results of MAPE between 60 and 120 and MSE modified version between 1% and 3% in foretold data error, which enables a guarantee with high fidelity predictions of response to the demand of electrical consumption. At the end of this document, graphics of forecast error versus: compression ratio, amount of measured data, different dictionaries and different reconstruction algorithms will be presented.