An indirect forecasting system of the power from a solar panel array based on modified fuzzy neural network

Forecasting systems of the power from a solar panel array based on neuronets increase the efficiency of a solar plant. Therefore, these systems are relevant in accordance with item 20A of the strategy of scientific and technological development of the Russian Federation. The power from a solar panel array has complex non-linear dynamic with uncertainties due to changes in cloudiness. Thus, it is impossible to approximate this complex dynamic with classical methods with a given accuracy, while neuronets provide the required accuracy. For identification of a forecasting system of the power from a solar panel array, intelligent methods in comparison with traditional methods provide the required accuracy by contributing to the safe and effective management of electric grids that integrating solar power plants. Under uncertainties by means of recurrent neurons and the attention mechanism the effective generation and transmission of a hidden information representation as a signal of the output layer of hidden recurrent neurons of deep neural networks, on the basis of the outputs of which a modified fuzzy neural network generated the forecasted value of power from a solar panel array by the fuzzy- possible convolution algorithm. The modified fuzzy neural network effectively distinguishes from the data significant functional aspects of forecasting of the power from a solar panel array, including aspects of identifying the cloudiness of the hour. The experimental modelling results of the indirect day ahead forecasting system of the power from a solar panel array based on the modified fuzzy neural network demonstrate its robustness and a decrease in the mean square error of its forecast by an average of three and six times in comparison with recurrent neural networks and a standard model of moving average autoregression under uncertainties.