Research On Short-term Power Load Prediction Method Based On Combinatorial Of Signal Decomposition And Improved GRU

Short term power load forecasting plays an important role in power system.It can not only ensure the safe operation of power grid,but also affect the power utilization of power plants.The improvement of prediction accuracy can also make the power system safer and conducive to economic development.However,it is more and more difficult to build a high-precision load forecasting model at this stage.On the one hand,due to the continuous improvement of people’s recognition of smart grid,there has been a largescale distributed power system,which leads to the gradual complexity of the influencing factors of load forecasting;On the other hand,due to the popularization of new renewable energy,the structure of power grid is more complex,which leads to stronger volatility of power load data.Due to the interference of the above factors,it is quite challenging to carry out accurate power load forecasting.Aiming at the problem that the traditional single prediction model is less effective when fitting non-stationary and non-linear power load data,this paper proposes a prediction method based on signal decomposition and improved GRU by using the network model in deep learning,combined with heuristic optimization algorithm and signal decomposition method.Firstly,Aiming at the problem that the GRU parameters cannot be determined,the improved gray wolf algorithm is introduced to optimize the super parameters of the GRU model.Through an example,it is verified that the GWO algorithm(YGWO)with disturbance factor and reverse learning strategy converges faster than the GWO algorithm,And the accuracy is better.Then,due to the fluctuating and nonlinear characteristics of power load itself,which will affect the prediction accuracy,an improved ensemble empirical mode decomposition MEEMD is introduced to decompose the original load sequence and obtain several relatively stable signals.This method indirectly improves the quality of load sequence.Through experiments,MEEMD and the other three decomposition methods are compared,it is proved that MEEMD method is more suitable for processing load data.In order to verify the feasibility and effectiveness of the prediction model proposed in this paper,simulation experiments are carried out using the load data of New South Wales every 30 minutes.The seven models of GRU,GWOGRU,YGWO-GRU,EMD-YGWO-GRU,EEMD-YGWO-GRU,CEEMDYGWO-GRU and MEEMD-YGWO-GRU and the existing two methods are compared and analyzed.The results show that the MEEMD-YGWO-GRU model in this paper has higher convergence speed and better prediction accuracy,it is an effective short-term power load forecasting model.