Prediction of Load Capacity in Microgrid by Multiple Regression Method

Abstract

The article presents the results of load power forecasting in Microgrid systems by multiple regression with a forecast range of one day. energy sources, as well as tools for storage, redundancy and load management. The design and construction of such systems is cost-effective, as these systems are powered by renewable energy sources, which is attractive due to subsidies and discounts on energy distribution - the so-called "green tariff". depends on weather conditions, such as temperature, pressure, humidity, wind speed and direction, cloudiness, etc., the task of predicting the load capacity depending on environmental parameters is relevant. Therefore, a forecast model of load capacity based on environmental data is developed and its software implementation is given. The daily curves of changes in load power with a discreteness of one hour are presented. Daily curves of load capacity changes on weekdays and weekends are also provided. A free resource has been selected to download the environmental database. A specific day is set for load forecasting. Hourly values ​​of environmental data (temperature, pressure, humidity) for a given day are given. The criteria for finding such days according to the environmental data are selected and the allowable percentage difference of mathematical expectation and variance of the relevant data is established. The parameters of mathematical expectation and variance of a given day are calculated. The statistical dependence between load data and environmental data is calculated. Regressive equations of the found similar days are constructed, on the basis of which the regressive forecast equation of loading capacity for days ahead is received. The daily curve of the forecasted load is presented and the comparative schedule of the forecasted with the real value of the load is constructed. The accuracy of the prediction is estimated using the average absolute error of MAPE. The algorithm and results of work of the developed program on which search of a similar day and calculation of forecast value for forecasting of power of loading for days ahead are represented are resulted.