Research Of Public Building Power Prediction Based On Combinatorial Model

With the increasing proportion of building energy consumption in the total energy consumption of the society,especially public buildings,their energy consumption per unit area is the largest and there are irrational uses of energy.Realizing rapid and accurate electricity consumption prediction for public buildings can help managers formulate economic and reasonable energy distribution plans,and ultimately achieve the purpose of building energy conservation.However,the changes in electricity consumption data have volatility and uncertainty.Traditional engineering methods and single prediction model can no longer meet the accuracy requirements of electricity consumption prediction.In contrast,the combined prediction model based on optimization algorithms,neural networks and data processing has greatly improved the performance of electricity consumption prediction.The thesis constructs a combined model based on the idea of decomposition-reconstruction-optimization-prediction,and selects the energy consumption of electricity in public buildings as the research object.The main research contents are as follows:(1)This paper uses the decomposition-reconstruction method to process power consumption data.Building power consumption data belongs to a random non-stationary sequence,and smoothing the data can effectively reduce the difficulty of model prediction.This paper uses fast complementary ensemble empirical mode decomposition,decomposes the energy consumption data into a set of intrinsic modal function components.Due to the excessive components generated by decomposition,they will affect the timeliness and accuracy of model prediction.The approximate entropy theory is used to reconstruct the intrinsic modal function components,which are divided into three components of high frequency,intermediate frequency and low frequency,and then obtain the relative a stable and small number of input sequences.(2)The improved whale algorithm optimizes the Elman neural network.To improve the defects of the whale optimization algorithm in the three aspects of initial population generation,algorithm convergence quality and algorithm internal parameter evolution mode.This paper uses 10 benchmark test functions to conduct an experimental analysis of the improved whale optimization algorithm and comparison algorithm,and verifies the effectiveness of the algorithm improvement strategy.Using the improved whale optimization algorithm,the weights and thresholds of the Elman neural network are optimized to solve the defects of the Elman neural network in the optimization process,such as the large randomness of the optimization effect and the poor quality of search accuracy.Finally,by selecting a large-scale public building in Xi'an,the power consumption data of the last 30 days as the experimental test set,the results show that the proposed combined model can reduce the prediction error to less than 6%.