Neural Networks Application for Power Consumption Planning of the Water Supply Facilities

Abstract

The issue of constructing mathematical dependency of power consumption of the water supply facility on relevant variables that have an influence on the efficiency of power consumption is considered. The expediency of mathematical modelling based on the experimental data obtained from the system of monitoring the efficiency of water supply process and the application of methods of data mining are substantiated. The power consumption model is constructed using a multilayer perceptron neural network. The procedure for identifying cyclic changes in the water supply process is applied to eliminate anomalous data. The set of models is trained to create a neural network model. The choice of better neural network architecture is made based on the productivity and indicators of the network errors within training, test and control subsamples using morphological criterion. The constructed model allows planning power consumption provided the known planned values of technological parameters and energy performance indicators. This model is suitable for adequately determining the energy baseline normalized to the relevant variables.