Improvement of Mathematical Model of Calculation and Power Loss Forecasting on the Basis of Neural Networks
Keywords
Abstract
Power losses in power networks are the most important indicator of their efficiency, a clear indicator of the state of the electricity accounting system, and the efficiency of energy supplying organizations.
According to international experts, electricity losses during transmission and distribution in electricity networks of most countries can be considered satisfactory if they do not exceed 4-5%. Electricity losses at the level of 10% can be considered the maximum permissible from the point of view of the physics of electricity transmission over networks. The sharp aggravation of the problem of reducing electricity losses in electric networks requires an active search for new ways of its solution, new approaches to choosing appropriate measures and organizing work to reduce losses.
Today, the main formalized means of analyzing the functioning and control of the grid modes is mathematical modeling, the basis of which is a set of mathematical models that adequately reflect the processes being investigated. Increasing the complexity of the power grids, the tendency to a comprehensive consideration of the processes occurring in them, the strengthening of the requirements for the efficiency of calculations leads to objective difficulties in constructing and applying traditional multidimensional nonlinear mathematical models. Need to revise and improve the classical mathematical models of calculation and forecasting of electricity losses that are obsolete and do not meet current requirements. Their use is ineffective and partially impossible. In addition, they work poorly with partial lack of input information. This substantiates the need for the introduction of modern mathematical models (in particular, neural networks) to improve the calculation and forecasting of electricity losses in the power lines of power grids.
For the calculation and forecasting of electricity losses in domestic networks at present, deterministic and probabilistic statistical methods are considered to be the most prevalent.
