The algorithm of the simulation graph of industrial enterprises reactive load at different values of the correlation functions parameters have been developed. The simulated graph has 172 values: 162 are used for training of each of the prediction methods, and the remaining 10 for comparison with the predicted values and for calculation of errors. A two-layer network with back-propagation of a mistake, with seven neurons in the hidden layer, one input and one output were used in predicting with neural networks in Matlab package. The activation function of the first layer is Hyperbolic tangent sigmoid transfer function, and the second one is a linear function. The LewenbergMarkvart function was selected for training of the model. Predicting by using statistical methods was made by extrapolation. In this paper, predicting based on extrapolation is implemented in the Mathcad program. The problem reduces to determining the value of a parameter outside the region in which the value of this parameter is known. This function is implemented in Mathcad by the command "predict". The last of the studied methods widely used today in industrial enterprises is predicting of the mean value. Its essence lies in finding the arithmetic mean value of the reactive load for the previous day and tuning the compensating devices to the power. The conclusions of the feasibility of each of the methods of the reactive loads predicting for different values of the attenuation coefficient of the correlation function are based on the comparison of the errors of predicting methods.

Разработан алгоритм моделирования графика реактивной нагрузки промышленного предприятия при различных значениях параметров корреляционной функции; произведено прогнозирование электрических нагрузок с помощью нейронных сетей, статистических методов, а также по среднему значению потребления реактивной мощности за предыдущие сутки. На основе сравнения погрешностей методов прогнозирования сделаны выводы по целесообразности применения каждого из методов прогнозирования при различных значениях коэффициента затухания корреляционной функции.