ADAPTIVE MODELS OF ELECTRIC LOAD FORECASTING OF PRODUCTION CONSUMERS

The article investigates models and methods of electric load forecasting. It is shown that the following power consumption control are currently known: instantaneous norm; at ideal rate; management on forecast value; with use average a moving time interval ("moving window" method). it better to focus those based study estimates, which source information for decisions. main requirements real-time systems are: high accuracy operational forecasting simplicity algorithms, provides minimum solution time; work in conditions uncertain insufficient information, ensuring stability management. analysis works devoted issues processes industrial enterprises carried out. automated have specific mathematical due little nature parameter, small amount reporting statistics reliability most accurately meets such requirements, adaptive approach method design. allows solve problem adequacy object from point view realization calculations first place necessary put and, all, exponential smoothing. Exponential smoothing, considered this paper as predictive model, identify inadequacy model real process bring estimate determined basis closer one, ie reduce prediction error. However, requires time, increases increasing changes coefficients model. In regard, there regulating reaction rate predicted its coefficients. A number automatic adjustment smoothing parameter analyzed: evolutionary adaptation, using tracking signal, adapting by, optimization gradient adaptation signal simple especially valuable modeling series short history. Due simplicity, convenient where predictions made computer technology. Studies statistical data obtained various adapts step 4-6 then error does not exceed 2%. Analysis showed their efficiency good adaptability electricity consumption. greatest difficulty cases abrupt development process. Abrupt can lead violation previously existing qualitative relationships parameters projected system. If jump, very important assess whether deviation caused by an obstacle or change obstacle, must be filtered deviations current value. From fast working off value rather effective. highly susceptible interference. To eliminate circumstance, modified procedure correcting proposed. introduction logical operator, inconsistencies forecasts imposes additional restrictions values original statistics. Experimental studies