Research And Application Of Smart Campus Water Saving System Based On SSA-ConvLSTM-LSTM Short-term Water Demand Prediction

As an important part of urban water consumption,the per capita water consumption of university park is significantly higher than that of urban residents.Therefore,scientific and reasonable supervision of water consumption in university park is of great significance to save water.At present,the water supply of university park mainly depends on artificial experience decision-making,but the error of artificial experience decision-making water supply is large,which often leads to water pipe burst,insufficient water supply at peak water consumption,etc.In this paper,the characteristics,distribution and influence factors of historical water supply data are analyzed,and a hybrid water demand prediction model based on deep learning is proposed.In order to further improve the accuracy of short-term water supply in campus,the mixed water demand prediction model is encapsulated into Falsk service interface and applied to water affairs platform.The main contents of this paper are as follows:(1)A short-term water demand forecasting method based on SSA-Conv LSTMLSTM hybrid model is proposed to solve the problem of low precision of single-step and multi-step forecasting in short-term water demand forecasting.Because of the existence of singular values in the historical data samples of the campus,it is necessary to preprocess the sample data before the model training.Campus water use trend is influenced by water use behavior of water users,and it shows multi-cycle and multipeak characteristics in time dimension.In order to extract the above features,it is found that Conv LSTM(Convolutional Long Short Term Memory,Conv LSTM)has the ability of extracting spatio-temporal features through theoretical derivation.Although Conv LSTM can extract certain spatio-temporal regularity,the prediction accuracy still has room for improvement.In order to further improve the prediction accuracy,LSTM(Long Short Term Memory,LSTM)network is added to Conv LSTM to form Conv LSTM-LSTM hybrid model.However,many parameters of the hybrid model lead to difficulty in adjusting parameters.For the parameter adjustment problem of the mixed model,the Sparrow Search Algorithm(SSA)is used to optimize the number of convolution kernels and the number of hidden neurons in the Conv LSTM-LSTM mixed model.The simulation results show that the method based on SSA-Conv LSTM-LSTM hybrid model can effectively mine the trend of campus water demand and accurately predict the short-term campus water demand.(2)In view of the large error of short-term campus water supply and imbalance between supply and demand of water resources,a smart campus water-saving system based on Flask framework is established.The system is designed and developed based on SSA-Conv LSTM-LSTM prediction model and park water management business.Based on the historical data,the system realizes the function of water demand prediction and water supply early warning.In addition,the system can forecast the water demand and the trend of water demand in the future by using the forecasting model,which can provide data basis for the short-term water supply planning of the campus.The system test shows that the smart campus water saving system is safe and stable,and can effectively improve the intelligent level of campus water management.