G. V. Proskurina

The article proposes mathematical regression models which allow to raise the accuracy of the gradient method that underlies the supernumerary situation detection module. A place of the investigated method is specified in the decision-making support of a MG dispatcher in case of rupture occurrence, the necessity of machine experiment is proved. Author describes the management object and lists alternatives of the experiences initial parameters in a kind of the experiment plan. The model of the experiment response is constructed, concepts of the averaged on time characteristics of the method accuracy are entered. The dispersive analysis is applied for exposing of method accuracy dependences from the factors varied during an experiment. Check of investigated average characteristics is made for the proof of a correctness of application of the given statistical method on distribution normality. The dispersive analysis has shown absence of direct communications between the method accuracy and factors. Therefore it is necessary to enter the additional parameters, depending not only from experiences of values of experiment factors set. Variety of such parameters is resulted, the correlation analysis for the purpose of significant parameters revealing is carried out. On the basis of the correlation analysis two mathematical regression models are created, the basic estimations of their adequacy and accuracy are calculated. The models constructed under the initial data of one site of MG, also work on the data from other site, consequently, they are applicable for the method-s estimation and increases the accuracy of rupture place detection. Perfection of the method underlying the module of gas leaks detection on the main gas pipeline, will allow to minimize losses of gas at the expense of reduction of decision-making time by the dispatcher of a gas-transport enterprise.