An Improved Grey Model For Mid-long Term Electricity Demand Forecasting

As the basic production data,electricity plays an indispensable role in the national economy and people’s livelihood.Accurate load forecasting is an important prerequisite for ensuring the balance of load supply and demand.Only by ensuring the balance between supply and demand of electric power can we ensure the safe and stable operation of the power grid.In the long run,China’s load data has not only the stability of year-on-year climbs,but also the instability of random changes.It is a classic gray system and can be forecasted using a gray model.However,with the improvement of the education level of the power market,the power system has also become complex.The accuracy requirements of the mid-long term power load forecasting have exceeded the capabilities of the traditional gray forecasting model.The grey forecasting model needs to be improved and enriched.The main work of the thesis is as follows:Firstly,This paper introduces the theory of grey system,studies the traditional grey GM(1,1)model from the modeling mechanism in detail,and points out the disadvantages of the traditional grey forecasting model in power load forecasting,bringing out the improvement of the model below.Secondly,this paper analyzes the grey prediction model error sequence.The analysis results show that most of the error sequences are in fluctuations;if the training sample fluctuations are relatively large or there are abnormal values,the prediction results are not accurate.For the first problem,this paper proposes a Fourier-Rolling Grey Model.The new model combines Fourier error-corrected gray-scale models with metabolic mechanisms,taking into account both the positive and negative aspects of errors,as well as the basic principles of data utilization.For the second problem,this paper proposes a Fourier-Rolling Dynamic Parameter Grey Model.This model introduces new parameters to dynamically adjust the gray parameters.The adjustment not only increases the weight of the current data,to a certain extent,it also reduces the error caused by excessive data fluctuations,Minimize the impact of numerical fluctuations on the forecast results and improve forecast accuracy.In order to verify the performance of the proposed model,we collected power load data for nearly a dozen years in various regions of Wuhan,and compiled them into annual forecasts and monthly forecasts based on the prediction time.The experimental results show that the two prediction models proposed in this paper have a greater improvement in the prediction accuracy than the improved prediction model.