Modeling and Simulating Soil Evaporation Law in Northern Cold Region Based on Particle Swarm Optimization-BP Network

摘要

According to the disadvantage of hard observation parameters, complex calculation formula, large calculation, and low simulation accuracy for traditional simulation methods of soil evaporation law, taking the standardized values of soil evaporation in Harbin from November 2008 to May 2009 and its six impact factors, including ground temperature, air temperature, relative humidity, wind direction, atmospheric pressure and vapor pressure, as input data of BP network and adopting particle swarm optimization algorithm (PSO for short) to optimize network structure, initial weights and thresholds, the PSO-BP network model of soil evaporation law simulation in northern cold region can be built. The results show that PSO-BP network model can accurately simulate frozen soil evaporation law and can avoid human interference of single BP network in training process. Therefore, it is an effective method to simulate complex soil evaporation process with PSO-BP network model, which can depict the variation law of soil evaporation. The results can provide a new thinking and a new way for modeling and simulating soil evaporation.