Research On Reservoir Operation System Based On Knowledge Engineering And Agent Theory

Reservoir operation is the important matter in water resource management. It's a traditional research problem. The level of management and decision-making of reservoir operation is advanced with the development of other theories and methodologies. Artificial intelligence (AI) is a new margin subject in computer science. Knowledge Engineering (KE) and multi-Agent system (MAS) are contacted with each other in AI. The thesis applies theories and methods of KE and MAS into management and decision-making of reservoir operation. To hope solve some problems in reservoir operation or solve these problems better.There are both structured problems and non-structured problems in reservoir operation. So precise models and algorithms alone are not enough to make decisions. The knowledge and experiences of experts and general knowledge are necessary. Based on theories and methods of KE, the thesis sets up knowledge representation system of reservoir operation, acquires knowledge for reservoir operation by data mining tools, sets up the reasoning system by combining cased-base reasoning and rule-based reasoning, namely, problem solving system. By expressing expert knowledge and experience in computer system and fuzzy reasoning mechanism, knowledge-based reservoir operation system is set up. The system can solve both structured and non-structured problems.Reservoir operation system is a distributed and storied system in nature. The concerned data and knowledge are distributed in every reservoir. The control work of reservoir operation is distributed in logic and physics. This character fits Agent exactly. The thesis discusses the character of reservoir operation and fitness of Agent to reservoir operation, sets up the reservoir operation system on MAS, applies contract net protocol to scheduling of generating units, constitutes Parallel Genetic Algorithm for reservoir joint operation based on MAS, and applies cooperative game and MAS to reservoir operation and distribute cooperative benefit by Shapley value.Chapter one introduces the background of the thesis, namely, the problems in reservoir operation and the fitness of KE and MAS to solve these problems. This chapter analyses and sums up the character of reservoir operation. The research general situation of operation model, operation system, multi-object problem disposal method, uncertainty problem disposal method, AI theory and method application, 3S technology application, uniform management of water quantity and water quality, conjunctive ground water and surface water management are summarized. Some research directions are indicated and development directions of reservoir operation are pointed out.Chapter two sets up the knowledge class system and knowledge representation system for reservoir. That it's to say, using fact base, model base, rule base and case base to represent knowledge for describing, knowledge for process, knowledge for control and operation case separately. The representation methods of rule and case are researched detailedly. In order to use mature theory and technology of database, improve the efficiency of knowledge representation and reasoning, knowledge representation methods are set up based on knowledge representation theory and reservoir operation practice, which can be operated in computer system. Rule base and case base are studied and designed detailedly.Chapter three analyses the demand of data mining in reservoir operation. Firstly, there aremany data in reservoir operation system. But these data are used basically, the potential rules are not found from data. Secondly, the knowledge-based reservoir operation system needs a lot of knowledge. The emphasis of this chapter is to apply association rule mining method mining choice criterion of reservoir operation models (operation model). With the analysis work, it's pointed that there are problems when classical Apriori algorithm is used in data mining in reservoir operation case base. The reason is pointed out. So, the classical Apriori algorithm is mended and the mended one is fit to similar problems, which include reason attribute and result attribute and the value of every attribute in a Transaction is exclusive. Furthermore, this chapter puts forward the process of metarule-guided mining of association rules by SQL language.Chapter four puts forward reasoning mechanism for knowledge-based reservoir operation, constitutes the reasoning mode of combining rule-based reasoning and case-based reasoning. The method and realization technique of case index (numerical type, enumeration type and fuzzy type), case acquisition (in the manner of case index and SQL language) and case choice are studied detailedly. As to rule-based reasoning, the uncertain reasoning mechanism, conflict resolve strategy and work mode of reasoning machine are studied and realized. Rule-based reasoning and case-based reasoning are coupled in the manner of primary one and accessorial one. Rule-based reasoning gains the basic parameters and condition for case-based reasoning, case-based reasoning acquires the similar history flood and reuses it or uses for reference.In chapter five, the fitness of applying MAS to reservoir operation is pointed out by analyzing reservoir operation system. This chapter designs the basic structure of reservoir operation MAS, the function of the system and communication mechanism. The reservoir operation MAS includes reservoir Agent, sub-valley Agent and valley Agent. These Agents are managed and controlled in classification. In this chapter the theory of multi-Agent system is applied in scheduling of generating units. Every unit is regarded as an agent and the all agents constitute a multi-agent system. According to the characteristic of scheduling of generating units, contract net protocol is used in the communications and cooperation among unit agents. The information transfer format among unit agents is designed. Compared with traditional method, scheduling method based on multi-agent system lets units exert their autonomies and harmonization among units. So scheduling system based on multi-agent system can improve the stability, security, flexibility of the system, and shorten the operation time. In multi-reservoirs jointly operation, a type of parallel genetic algorithm is designed based on reservoir Agent. The algorithm will shorten the calculation time.In chapter six, how to realize jointly operation by cooperative game and Shapley value method is studied when these reservoirs have waterpower relationship, but belongs to different benefit parts each other. It's differ from that when these reservoirs belong to the same benefit parts. Because the physics factors (traditional research part) can be corresponded only the benefit is corresponded firstly. MAS provides information base for cooperation. Cooperative game provides a cooperation mode for reservoirs jointly operation. And Shapley value method provides a benefit assign mode. Shapley value method satisfies individual rationality, collective rationality and validity and it assures that all reservoirs with waterpower relationship will operate united. The reservoirs jointly operation model and benefit assign mode are valid when it comes to multi-reservoirs jointly operation or hydroand thermal power system.Chapter seven is for case study. Based on Bujihe River in Shenzhen City, by requirement analysis and application of theories and methods of knowledge representation, data mining and knowledge reasoning in knowledge engineering, knowledge-based flood operation system in Bujihe River (KBFOS-BR) is developed. In KBFOS-BR, the relation of safety runoff of Sungang flood detention area with other factors from history flood data by statistical method, and it's translated into association rules. The comparison and analysis of the operation result by present operation rules and the rules in KBFOS-BR shows that the rules and reasoning system in KBFOS-BR are effective. The operation work by rules in KBFOS-BR can make full use of Sungang flood detention area and downdream river. Ultimately it will improve the flood control criterion of Sungang flood detention area.Chapter eight is the summary and prospect for the whole thesis, which sums up the whole thesis and points out several development directions of applying theories and methods of artificial intelligence to reservoir operation research.Developing intelligence operation system is one of the reservoir operation development direction. The research work in the diesis is an exploration of using theories and methods of artificial intelligence in reservoir operation. The main is to set up intelligent reservoir operation system based on knowledge by doing so. The system has the ability for knowledge representation, data mining, knowledge reasoning by knowledge base and reasoning system. Reservoir operation system based on Agent is the higher demand for reservoir operation. In this system, the reservoirs operation jointly by communication and cooperation among Agents. Ultimately the whole valley runs in optimum state.