The possibility of using an artificial neural network for the short-term and operational prediction of the load of an electric power system is considered. A structure of an artificial neural network, which predicts its "own" hour is proposed, and the optimum set of input data is determined. The advantages of using this method compared with other methods employed at the present time are proved using the example of the Samara power system.Keywords: artificial neural network, load prediction, outside air temperature, power system. Short-term and operational load prediction plays a key role in ensuring the economic and safe operation of a power system. The choice of the structure of the generating equipment, the planning of the dispatcher graph and the estimate of the power plant stability require solutions of this problem.The load graph of the majority of consumers possesses the property of quasisimilarity: the consumption during the last period, as a rule, is similar to the consumption in the previous periods. The main reasons for the difference in the load as a function of the measurement period (the hour, day, month and year) are as follows:-load dispersion -the change in the mode of operation of major consumers; -a change in the composition (the connection of new and the disconnection of old) or parameters of the consumers;-the effect of meteorological factors (the temperature or illumination), since a considerable portion of the load is used to maintain normal conditions in buildings (heating, ventilation and air conditioning).The solution of the problems of short-term and operational prediction of the load is based on the use of the following model [1]:where P (t ) is the power consumption per 1 h t, F (t ) are the trend values (the hour before, the days before, the months before etc.), which have the greatest influence on the predicted value, K (t ) is a correction function, which takes into account the dispersion, the change in the loads and the temperature of the outside air, and k and m are the values of the time shift, determined by the trend values that have the greatest effect on the predicted value. Regression methods are employed to solve problems of short-term and operational load prediction. However, they are based on linear models, while the load series which they model are nonlinear functions of exogenic variables. Moreover, regression methods do not enable the effect of meteorological factors on the value of the predicted load to be taken into account accurately in view of the nonlinearity of this effect. The change in the load structure, which is occurring at the present time (the considerable increase in the portion of communal-everyday and nonindustrial load), means that meteorological factors have a considerable influence on it [2].With the development of the theory of artificial intelligence, it has been suggested that the problem of predicting an electrical load can be solved using models based on artificial neural networks (ANN). The theory of artificial intelligence assumes the existence ...
Состояние вопроса. Научная проблема исследования заключается в необходимости прогнозирования электропотребления собственных нужд электростанций с минимальной ошибкой. Решением задач краткосрочного прогнозирования ранее занимались на уровне электроэнергетических систем и промышленных предприятий. Что касается прогнозирования электропотребления собственных нужд электростанций, то в качес тве прогнозных значений использовались ретроспективные данные по электропотреблению. Данная проблема сохраняет свою актуальность согласно Постановлению Правительства РФ от 27 декабря 2010 г. № 1172, в котором отм ечено, что электростанции берут на себя ответственность за потребление электроэнергии, объем которого вышел за рамки установленного. Отклонение в электропотреблении на 2 % и более от установленного значения приводит к дополнительным финансовым расходам. В связи с этим актуальным является выбор метод а прогнозирования электропотребления собственных нужд ТЭЦ с низкой погрешностью. Материалы и методы. Для решения задач краткосрочного прогнозирования выбран метод, основанный на искусственных нейронных сетях, и проведено обучение данных сетей с помощью методов численной оптимизации: алгоритма обучения Бройдена-Флетчера-Гольдфарба-Шанно; метода Сопряженных градиентов; метода градиентного спуска, которые практически использовались для решения различных задач в электроэнергетике. Для определения почасовых значений электрической нагрузки собственных нужд ТЭЦ использован программный пакет Statistica Neural Networks. Результаты. Выбран метод, основанный на искусственных нейронных сетях «многослойный персептрон» и определен алгоритм его обучения Бройдена-Флетчера-Гольдфарба-Шанно, с помощью которого на ТЭЦ появляется возможность прогнозировать электропотребление системой собственных нужд со средней абсолютной погрешностью 0,43 %. Выводы. Предложенная методика краткосрочного прогнозирования электропотребления СН ТЭЦ протестирована и утверждена в Филиале АО «СО ЕЭС» ОДУ Средней Волги для оценки прогнозных значений электропотребления электростанций в процессе планирования баланса электроэнергии. Ключевые слова: прогнозирование электропотребления, собственные нужды ТЭЦ, искусственные нейронные сети, алгоритмы обучения, ошибка прогноза
Direct operating current system is a set of electrical devices and appliances, including power sources required for the functioning of the main technological complexes of enterprises. The features of direct current systems are a significant spatial distribution, susceptibility to strong electromagnetic interference and noise. A frequent fault in DC operating current systems isolated from the ground is the damage of the insulation of one pole to ground. This fault is not accompanied by large currents, but can cause false operation of the electrical system. In this paper we consider the method of searching for the faulty section, and analyze the occurrence of interference in the search system. The improved search system based on the external reference voltage superimposition method has confirmed its performance in real objects.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2025 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.