Forecasting On Electrical Net-load Especially In Smart Community Which Including Distribution Generations

Short-term forecast for renewable power generation and load in smart grid provides important foundation for the safety and economical operation of power system.From distribution perspective,it is important to find a prediction way which can combined forecasting the wind PV and load,may helpful for accordingly adjustment the control method and operation mode.With the influence of wind energy and solar energy,the power of wind and PV are unstable and randomness.The unpredictable makes them have a serious noise.The smart grid can be treated as a whole part to forecasting,and with the consideration of make the thermal generators controlling and distribution more optimal,there is necessary to build a combined forecasting model for wind power,PV and load,which is able to realize the one-time forecasting on net load and avoid the accumulate errors from separately forecasting one by one.Thus,it may helpful to cut down the expense for setting the forecasting system in control center.The defined net load represents the energy power difference between the energy load consumption and the renewable generations in smart community.So,the net load can reflect the energy supply level between the electric main network and the distribution network intuitively.The value of net load can be used for arranging the output of different unit direct.The accuracy of the net load can provide a decision opinion for dispatching in power system and it is vital for running the power system economy,safety and significance.With the updating of training data,the traditional way cannot able to obtain the forecasting value of net load accuracy.Therefore,a combined forecasting approach with model self-adjustment has been proposed.This approach can adjust the parameters with the training data update,which is not only with the well characters of SVM but also make up for the disadvantages in traditional SVM such as unable to update the training data and the poorly real-time ability.Obviously,comparing with the traditional method,the mentioned approach can forecasting the wind power,PV and load together.Moreover,it can update database in real time and adjust parameter by itself.In this paper,the experiment data is the actual power data form one micro-grid,finally,a contrastive analysis of the three models including the separately forecasting model for wind power,solar energy and load,the traditional SVM approach and the combined forecasting model based on a real-time model with parameter self-adjustment is given.The results show that the proposed method can avoid forecast errors caused by single forecasting model and overcome the problems of the traditional support vector machine,such as online update database and real time performance.It can also be utilized to forecast the model accurately.Further,the relationship between the weather factors including temperature,humidity,wind speed and the power data such as wind power,PV,load power has been illustrated in the paper.The forecasting method has considered more aspect such as weather element rather than just the historical power data,through looking up to the historical weather data in the same region.A concept of weather factor has been proposed at the first time,which is calculated according to the proportion of the different kinds of power data in this micro-grid.After integrating the various weather factors which is influenced net load,a contrastive analysis is shown in this paper.The forecasting results has been demonstrated on 3 different circumstance,specifically,the result of whether considering the weather factor or not and considering the weather factor according to the proportion in the micro-grid or not.