Research On Elevator Group Control System Based On Fuzzy Neural Network

With the development of Socio-economic, there are more and more high buildings, elevator group is playing a more important role in high-rise buildings and intelligent buildings, so the elevator group control system (EGCS) is now the focus of the researchers at home and aboard. There are two main contents in this thesis, including the elevator traffic pattern recognition and the scheduling algorithm of the elevator. Intelligent control is introduced in the paper.Firstly, the thesis puts forward background, purpose and significance. Then, the history and status quo are reviewed.Secondly, the basic characteristics of the EGCS are studied, including uncertain characteristics, imperfect characteristics, nonlinear characteristics, multi-target characteristics. Then the basic concept of the traffic flow is given, and the forecast methods are analyzed. The performance evaluation of the EGCS is studied, including target of time evaluation, target of energy consuming evaluation. Then the composing of the EGCS is studied.Thirdly, the fuzzy neural network (FNN) based upon the mamdani model is studied, which applied in the EGCS. The FNN Combines the advantages of fuzzy logic and artificial neural network, the knowledge can be expressed easily and the study capacity is strong. The structure of the FNN is given in the paper, which is the basic theory for the traffic pattern recognition and the elevator assignment.Then, the FNN is applied to identify the traffic pattern. The six kinds of typical traffic patterns with the characters are studied. Three-stage hybrid learning algorithm is used in the FNN. Network I is used to identify the upper peak pattern, the down peak pattern, the off peak pattern and the inter-floor pattern. When the proportion of the inter-floor pattern is smaller, the Network II is not need. When the proportion of the inter-floor pattern is bigger, the Network II is used to identify the two way pattern, the four way pattern and the passenger flow balance pattern. The samples are used to train the FNN, and the actual traffic flows are used to test the network. Finally, the scheduling algorithm for the elevator group control is studied. Elevator scheduling is a typical multi-objective programming. In this paper, the FNN based upon the mamdani model is used to optimize the elevator group. The control targets are the average waiting time, the average riding time and the energy consumption. According to expert rules, the model of the FNN is ascertained. Then BP algorithm is used to train the FNN. Then we use the FNN for actually, and the effect of the algorithm is validated.