Multi-agent Model for Dam Management Based on Improved Reinforcement Learning Technology

摘要

In order to achieve efficient management of tbe dam,on the basis of analyzing the dynamic and uncertain characteristics of engineering and its relationship with the environment, the new algorithms such as reinforcement learning, Synergetic, Structural Risk Minimization and Particle Swarm Optimization are used to establish a Cooperative Wavelet Least Squares Support Vector Machine Model. For the purpose of improving the convergence rate and making full use of knowledge and advice of mechanics and hydraulics of the dam, two kinds of models are used cooperatively, which are WLS-SVRM and WLS-SVCM. Before the training online, mapping provides training samples for WLS-SVCM. During the course of training online, the numerical simulation and WLS-SVCM will provide knowledge and advices for WLS-SVRM. Case study shows that the model can provide timely information of gate opening and management information of the dam so as to provide decision support for engineering management.