Neural network pressure observer for a turbomechanism electromecha- nical system powered by a wind generator

Abstract

Ecological and economic production of electrical energy through the use of alternative energy sources is an urgent direction due to the trend of increasing prices of energy carriers used in the electrical energy production and as a result of significant damage of the energy system of Ukraine in consequence of the war on the country territory. It is worth noting that in some areas it is possible to use only autonomous power generation systems, since the laying of electrical networks in these districts is impractical and unprofitable. Usually, the mentioned systems are based on a combination of a wind or hydro turbine - drive motor, and an electric generator. Such systems are characterized by high resource, reliability, low cost, and complexity of maintenance. Sometimes people's lives and the possibility of communication with the outside world depend on the operation of an autonomous electric power generation system, which is especially important in the conditions of martial law. At the same time, the lack of stabilization of the hydraulic network pressure of the water supply system can lead to the household conditions aggravation, the emergency situations occurrence, and the technological process disruption. In view of the mentioned factors, there is a need to measure the pressure of the hydraulic network, which is possible by using technological coordinates observers built on the basis of the artificial networks theory. In the paper a modern turbomechanism electromechanical control system powered by an alternative electrical energy source under the conditions of pressure stabilization of the hydraulic network when using a technological coordinates observer, namely a pressure estimator, is proposed. A mathematical description of the main elements of the investigated system is given. A hydraulic network pressure observer based on the artificial neural networks theory is built and studied. Features of design and training of technological coordinate estimators based on neural networks with feedback are described. The operation of the sensorless system during the pressure stabilization at a given level when the resistance of the hydraulic network changes within the typical daily cycle of water consumption is considered on a specific example. The results and analysis of the investigation of the developed observer in standard and sensorless control systems are shown.