The Load Forecasting Of Campus Based On Recurrent Neural Networks And Electricity Planning

In term of campus electricity,there often exist waste and inappropriate misuse.In order to save the electricity and enhance school's social responsibility,so electricity planning has become a research focus.Because the forecasting accuracy provides a good basis for the analysis of campus energy demand and the decision-making of electricity planning,the key question in the electricity planning is how to improve the forecasting accuracy.Our goal is to achieve and improve an accurate forecasting load model.We use the LSTM model because of its better performance in forecasting accuracy,but there is a problem: the model training time is longer and can not be predicted in real time.The research work of this paper mainly includes the following aspects: 1.Summarizes the development of energy prediction model;2.Study the energy consumption prediction model and optimization based on RNN and LSTM deep learning model;3.We improve the LSTM model,combined with CW-RNN model,to propose CW-LSTM model,it reduce the training time of the model,to make up for its shortcomings;4.Taking the energy saving of air conditioner as the goal,this paper proposes energy saving model for adjusting the air conditioning temperature,and proposes the air conditioning adjustment decision suggestion based on the per capita comfort based on the forecasting curve;5.Taking staggering peak power consumption as the goal,this paper proposes hydrogen energy storage adjustment model,put forward the reasonable decision suggestions of staggering peak power consumption and the energy saving.In this paper,the historical energy consumption data of a campus of SCUT are taken as experimental data,and carries on the sufficient experiment.The experimental results show that there are several energy consumption forecasting models,illustrating that the CW-LSTM model has better performance on energy consumption prediction.In the energy-saving model experiment,the proposed energy-saving recommendations can effectively control the air conditioning in the classroom,and achieve better air-conditioning energy-saving effect under the premise of lower damage to the comfort of teachers and students.In the staggering peak model experiment,the proposed staggering peak recommendations can effectively switch the hydrogen storage at the right time,reduce the overall difference between valley and peak,and maintain below the rated power.The overall work can effectively predict future energy consumption data and clarify their future energy demand and rational allocation of energy use.