ADAPTIVE SHORT-TERM FORECASTING OF ELECTRICITY CONSUMPTION BY AUTONOMOUS POWER SYSTEMS OF SMALL NORTHERN SETTLEMENTS BASED ON RETROSPECTIVE REGRESSION ANALYSIS METHODS

Abstract

The relevance. The construction of a problem-oriented tool for forecasting electricity consumption in small northern settlements is of paramount importance for the implementation of development plans for Arctic and Far North regions. At present, a large number of varieties of electricity consumption forecasting methods are used, including expert, statistical, artificial intelligence methods, hybrid methods, and others. As a rule, there is no universal method, equally effective (by the criterion "counting time - counting accuracy" for the main types of problems of electricity consumption forecasting. The noted circumstance requires research in the direction of creating a computational complex: identification of computational properties of electricity consumption model - construction of an adequate method of information extraction. The purpose: Development of an approach based on a retrospective regression analysis, which allows to make adaptive short-term forecasting of electricity consumption of regional isolated electrical complexes (RIEC). Methods: The approach to obtain short-term forecasts of electricity consumption of regional isolated electrical complexes is based on a retrospective regression analysis. The predictive model based on the QRETC responses is presented in the form of a linear regression with an internal set of functions forming an orthogonal and orthonormalized basis. At the same time the preliminary information received from the object - RIEC's responses, is written in the form of a system of linear algebraic equations presented in matrix form. Finding of coefficients at basis functions is carried out taking into account the method of least squares, and the solution of the received equations on the basis of Kaczmarz method. Verification of performance of the developed approach was carried out by means of the analysis of regression residuals of forecasting. Results. With the help of аdaptive short-term forecasting of electric power consumption by autonomous power systems of small northern settlements based on the methods of retrospective regression analysis the short-term forecast for the prediction interval of 30 minutes was obtained. Conclusions. The approach of adaptive short-term forecasting of power consumption by autonomous power systems of small northern settlements on the basis of regression analysis methods, which allows obtaining a short-term forecast of power consumption for the preemptive interval of 30 minutes, has been developed. Substantial advantage of the developed approach is demonstrated by the fact that when building a procedure of adaptive short-term forecasting of energy consumption on the basis of retrospective regression analysis it is rational to combine the processes of identification of coefficients at basic functions and tunability of mathematical model of non-stationary discrete stochastic process at each step. The analysis of regression residuals of ROETC response forecasting has been carried out and the performance of the developed algorithm of electricity forecasting and the adequacy of the accepted provisions in forming a priori information when implementing the approach to short-term forecasting of a stochastic process based on retrospective regression analysis has been confirmed.