Study On Mountain Flood Warning System Based On Growing-pruning Neural Network

Our country is a mountainous country with complex mountainous terrain. In recentyears, there have been frequent mountain flood of the provinces in China. Because ofthe mountain torrent disaster suddenly strong, great, artificial warning difficult, timelyand accurate decision-making and emergency response mechanism is the key to thecurrent flash flood defense.Aiming at technical problems existing in the mountain torrent disastrous defensesystem, the thesis according to the practical problems, analyses the mountain torrentdisaster outbreaks of the key factors, including rainfall, temperature and water flow. Asdesign of a low power consumption, stable performance and high precision of the dataacquisition unit, consider the environmental parameters. Data transmission and processare needed, after data collection complete. In the process of transmissioncommunication by915MHZ wireless sensor networks, a reliable transport mechanismand stable transmission device node is designed. After the data gathering completion,then through GPRS network sent to the monitoring center for processing.Monitoring center is responsible for the processing of data, at the same time issuedan order operation. Mountain torrent of data gathered through the flash flood forecastmodel for training and learning, and through the interface intuitive show thecorresponding information. Mountain flood forecasting model using improved growthand clip BP network, the network through flow distribution and likelihood functionanalysis of mountain flood water flow data, achieved advanced prediction effect.Through the practical application shows that the system runs stability and themonitoring center can remote view the information in the mountains. After a flash floodforecast model predictions, accurate to provide effective information for flood controldecision, and make prevention for early warning effect.