Short-term Power Load Forecasting Model Which Considering Renewable Energy Generation Output

The load data has a high degree of nonlinearity and is influenced by external factors such as weather and electricity prices.The net load data,in particular,is closely tied to the solar power generation.As a result,the structure and selection of features for load data prediction have a direct impact on the model’s accuracy and prediction cost.A shortterm model for load forecasting is proposed through a data-driven approach,with the following main findings achieved:(1)To address the problem of multidimensional features affecting model prediction accuracy and training cost,In this paper,a short-term forecast model of electric load using Phase Space Reconstruction(PSR)and Stochastic Configuration Networks(SCN)is designed.Firstly,the proposed method uses Principal Component Analysis(PCA)to analyze and dimensionally reduce the dimension of relevant data,and then the reduced data and load data are composed into multivariate data according to the time axis.Then,we use the mutual information method and the spurious nearest neighbor method in phase space theory to obtain the required parameters.Finally,we construct a high-dimensional structure of multivariate series based on the requested parameters and model the electric load using a random configuration network.The experimental results on real data sets show that,compared with Long Short-Term Memory network(LSTM),Autoregressive Integrated Moving Average Model(ARIMA),etc.,the method which designed in this paper has the advantages of low manual involvement and fast computation rate,which shows its considerable practicality.(2)To address the problem of new energy generation output affecting the accuracy and noise resistance of power load forecasting,a short-term power load forecasting model based on Variational Mode Decomposition(VMD)and improved N-BEATS network is proposed in this paper.By calculating the Relative Entropy of different modal numbers of load data,the predetermined value of decomposition modal number is determined,and the load data is decomposed into different Intrinsic Model Function(IMF)components by VMD,then a prediction model is established separately for load data as well as photovoltaic power output data by using the improved N-BEATS network combined with Soft Thresholding to finally obtain the net load prediction model.The effectiveness of the proposed method is verified with the European grid public dataset,and the results show that the proposed method is comparable to the VMD-N-BEATS model,N-BEATS network,modified N-BESTS network,long and short-term memory network,Gated Recurrent Unit(GRU),Support Vector Regression(SVR)and Random Forest(RF),the method designed in this paper reduces the number of models while considering the influence of the load frequency domain,and can construct highly discriminative features with better noise resistance.(3)A Python-based Django framework was used to design a power load forecasting and analysis system.