Training of neural network for forecasting results of the intensification of gas inflow in conditions of information insufficiency

Decision-making process at designing actions on intensification of gas inflow is rather complicated and cannot be subject to the precise mathematical description. In this connection the information system of support of decision-making with application of neural networks is developed. The developed neural network has only known entrance vectors for training. The process of training consists in adjusting weights of synapses. Adjustment of synapses can be made only on the basis of the information accessible in neuron, that is its condition and already available weight factors. To check efficiency of the given method of training the forecasting calculation of seam pressure upon a part of a gaseous seam, for which neural network model of porosity has been built, is made. As verifying sets of entrance and target signals real data on gas collectors of the Bashkir circle in Astrakhan gas-condensate field have been used. The received results are evidence of efficiency of the given method of training.