Improving The Accuracy Of Short-term Forecasting Of Electrical Loads Taking Into Account Meteorological Factors Based On Support Vectors Method

At the moment,there are a number of forecast models and software systems of domestic and foreign developers,which allow to perform forecasting of power consumption with the required accuracy for users,but in some cases,in particular,in the control room in power systems,the results on the accuracy of short-term and operational forecasting do not always meet the increasing requirements.Therefore,the task of improving the accuracy of forecasting power consumption remains relevant.The scientific novelty of the work is as follows:(1)A model for short-term forecasting of power consumption in the areas of operational zones of regional dispatching offices,based on the method of reference vectors and particle swarm algorithm,characterized in that it takes into account the values of natural light as one of the influencing factors,which allows to improve the accuracy of modeling and forecasting.(2)For the first time,the particle swarm algorithm was implemented to optimize the parameters of the regression model of reference vectors,in which both air temperature and illumination are taken into account as influencing factors.(3)It is shown on the basis of research of two most effective and perspective forecast models(neuro-fuzzy network and the method of reference vectors)that the regression model of reference vectors has the best approximating properties in the space of variables: power system,air temperature and natural light.The theoretical significance of the results of the dissertation work is to developed models based on support vector machine using particle swarm optimization algorithm to optimize the parameters of the model.The method allows to increase accuracy of establishment of nonlinear dependences between consumption of the electric power,air temperature and natural illumination.The practical significance of the developed model can be used to predict the energy consumption in the regional dispatching offices of the branches,wholesale generation companies and territorial generating companies,regional grid companies,energy sales companies,as well as in dispatching offices of the individual companies that are members of wholesale or retail electricity markets and power.A short-term forecasting computer program was developed in MATLAB power consumption for regional dispatcher management based on the method of reference vectors.