Water Level Intelligent Prediction Technology And Application Based On Deep Learning

The water level of the inland waterway is an important factor to guide the reasonable stowage of ships and ensure the safe navigation of ships.Reasonably predicting the short-term trend of water level changes is essential for improving the passage capacity of the waterway,ensuring the safety of ship navigation and scientifically carrying out waterway maintenance.In order to improve the prediction accuracy of inland waterway water level,this paper based on deep learning method,using gated recurrent neural network(GRU)and convolutional neural network(CNN),studies the intelligent prediction model of inland water level in-depth.And this paper develops an intelligent water level prediction service system to meet the needs of comprehensive waterway information services,and implements the application of water level prediction models.The main work of this article includes:(1)The application of recurrent neural network in the prediction of river water level is studied.A single water level prediction model based on GRU is established,and compared with the model based on long-term and short-term memory(LSTM)to analyze the cyclic neural network more suitable for water level prediction Structure—GRU.(2)Through the analysis of the space-time relationship of the water level stations,the GRU-based multi-water level station linkage prediction model and the CNN+GRU-based multi-water level station linkage prediction model are further established.The experimental results on the 30-year 8:00 water level observation data set of multiple water level stations in the lower reaches of the Yangtze River show that the CNN+GRU-based multi-station linkage water level prediction model can reduce the impact of the randomness of single water level data and better integrate the correlation of water level values between upstream and downstream water level stations is used,so it has higher prediction accuracy.Moreover,through comparative analysis of evaluation indicators such as Nash-Sutcliffe efficiency coefficient(NSE),root mean square error(RMSE)and mean relative error(Mean Absolute Percentage Error,MAPE),based on CNN+GRU's multi-station linkage water level prediction model also has higher prediction accuracy than water level prediction models based on differential autoregressive moving average(ARIMA)and wavelet neural network(WANN).(3)TensorFlow Serving technology is used to deploy and apply the CNN+GRU-based multi-station linkage water level prediction model.A smart water level prediction service system is developed using Spring Boot and Vue.js technologies based on the front and back end separation framework.The various parts of the system are independently deployed and docked through the RESTful API interface,which has good loose coupling and flexibility,and the prediction results can be displayed in various forms such as Web pages,APPs,and WeChat public accounts.This system provides convenient intelligent water level prediction service for inland navigation ships and other port and navigation users.