A method for forecasting the energy consumption of an industrial production plant, the computing device for carrying out the process as well as associated computer program - product and the computer readable medium

摘要

A method for forecasting the energy consumption of an industrial production plant, in which for a prediction period planned production quantities in a neural network as input data and energy consumptions for the prediction period as output data from the neuronal network are determined, wherein the neural network with known historical a - and output data (his) to be trained, characterized in that for training the neural network, the number of historical input data (pm_ij) by forming weighted sums of input data is reduced, in which case for the determination of weighting factors (x_i) for the formation of the weighted sums, an optimization problem (opt) with a predetermined target condition (min) is formulated, and wherein from the known relationships, after which the output data (ev_j) from the input data (pm_ij) , first a system of equations (gls) with weighting factors (x_i) as a variable is set up, and that then the weighting factors (x_i) as long as be newly determined to this with regard to the target condition (min) for the deviations of the from the respectively last determined weighting factors (x_i) calculated output data (ev_j(x_i)) of..