Power system load prediction method using fuzzy decision-based neural network model

Abstract

The invention discloses a power system load prediction method using a fuzzy decision-based neural network model. The method includes the following steps that: the degree of membership mui of historical load samples is obtained by using the fuzzy decision-based neural network model, and a historical sample membership degree curve is drawn; historical sample data of which the degree of membership is smaller than mui are removed, processing is performed, and therefore, samples which contain missing data and data that vary sharply can be basically eliminated; and after data pre-processing, an appropriate neural network model is selected, and change rules are found out based on the historical data of power load, and a power load neural network prediction model is established. According to the power system load prediction method using the fuzzy decision-based neural network model of the invention, the advantages of fuzzy decision and neural networks are combined, so that the convergence precision and convergence speed of the method can be both improved, and dependence on an initial value can be decreased, and therefore, the accuracy of prediction can be improved.

Description

technical field [0001] The invention relates to a fuzzy decision-based neural network model power system load forecasting method. Background technique [0002] At present, the change of power load is mainly governed by people's production and life rules and presents regularity, and is affected by weather and other factors. The total load of an area is the sum of individual loads that are difficult to count, so there must be random components in the load. The periodicity and randomness of load changes are a pair of contradictions. The fluctuation between the two determines the predictability of the load and is an important factor affecting the accuracy of the load forecast. Improving the accuracy of load forecasting is the goal pursued by all researchers engaged in load forecasting, but the unpredictable forecasting error has always troubled many researchers. In fact, the historical load data used for modeling, the error of the model itself, There will be some internal lin...

AI Technical Summary

Problems solved by technology

At present, the key to achieving the goal of improving forecasting accuracy is how to consider the influence of meteorological factors on load more reasonably, because the proportion of meteorological sensitive load in the total load is increasing. For a long time, given that the meteorological department cannot provide real-time Most of the forecast models established by the power system are based on daily characteristic meteorological factors, such as daily maximum temperature, minimum temperature, etc. These forecast models do not consider the factors that affect the power load, and are only used as general time series Therefore, its accuracy is not satisfactory, especially for forecasting ultra-short-term and short-term power loads in my country.

Images