Several Intelligent Computing Methods And Applications

Intelligenct computing methods include artificial neruel networks (ANN), fuzzy systemsand evolutionart algorithms. ANN processes the prekonwn knowledge in supervioused orunsupervioused way and has good learning ability. Fuzzy system handles the uncertainproblems in practice in the way of computer and it enables computers to solve the practicalproblems in real life and manufacturing. Evolutionary algorithm use simple representations todescribe all kinds of problems and guides the searching direction and learning based on therule of survival of the fittest. With the fast development of computer technology, theintelligent computing methods are more and more widely applied, especially in the fields ofsmart power grid and job shop scheduling problems.In this thesis, expert system, ANN, fuzzysystem and evolutionary algorithm are extensively studied. Experiments of smart power gridand job shop problems were performed to test the algorithms and got some conclusions. Themain contents of the dissertation include:1. The structure of expert system and uncertain factors were introduced and uncertainexpert system was mainly researched. Based on that, a neural network expert system (BPES)was developed combined with ANN. And it is applied to the process of waste water. BPESlearned and was trained based on the biochemical process of the complex waste water process.I.e. learning ability of crisp expert system is fined and the judgment accuracy of ANN wasimproved. This dissertation described the establishment and the learning mechanism of BPES.Compared with BP-neural network (BPNN), expert system (ES), BPES can produce moreaccurate results and has a better learning ability. In addition, it can provide complete diagnosisadvice and proved to be an effective tool.2. MBPES was proposed based on the flaws of BPES and adopted in smart power griddiagnosis. It established cooperative expert diagnosis system based on expert knowledgedatabase, serial, parallel and hybrid network. The inputs of it are the real-time data andprofession data. MBPES can sort the knowledge and data, improving the diagnosis speed andeffectiveness. The experimental results indicated that MBPES could provide a higher thanBPN and BPES. A complete electric grid faults diagnosis and evaluation knowledge systemwas established based on the foregoing method.3. In this dissertation, an intelligent diagnosis system, the input of which is the gasdissolved in the oil, was established. The system has diagnosis system based on the fuzzytheory based diagnosis method. The parameters of the system include input section, outputsection and the function module comprising the system. The experimental data indicated theeffectiveness of this method. Experimental results showed that fuzzy group system couldprovide a higher accuracy of estimation than BNN and fuzzy system. Also, a complete transformer diagnosis and evaluation system was established.4. This dissertation solved the problem of transformer aging evaluation with the fuzzytheory combined with bidirectional associative memory network (BAM)。 BAM is adouble-layer feedback neural network, which can handle the situation where the instability ofthe features of the isolated oil gas and the electric features of the isolated oil influence thetransformer. Fuzzy theory can fuzzify the present evaluation standard. The combination ofBAM and fuzzy theory can solve the evaluation of the aging of transformer. The simulationexperiment was performed to test the proposed method. The results indicated a good matchwith the practical situation. The effectiveness is tested.5. The application of intelligenct computing methods in scheduling problems wasdiscussed. This dissertation obtained the comprehensive fuzzy weights evaluation in problemswith makespan, average flow time, total tardiness and the total load as objectives. Then themethod filters the machines and operations based on the results of the evaluation to obtain thedevices and operations suitable to the objectives. Then, the makespan and the expecteddelivery time were used as the fuzzy membership function and the maximum averagesatisfaction rate was improved with immune GA. It can be tested that the proposed methodcould get better results through the experiments using traditional GA and immune GA.6. The fuzzy numbers are produced with experience, which relies heavily on experienceand subjective factors. In this dissertation the production data, i.e. the processing time in thisproblem, was used to generate the fuzzy number based on normal distribution. This decreasesthe influence of the subjective factors on the judgment of the production effect. In addition,fuzzy satisfaction rate was adopted as the fitness function of the artificial bee colonyalgorithm. The experimental results indicated a good effectiveness.In conclusion, several novehybrid intelligenct computingmethods were proposed basedon the present ones. These methods could obtain good results in different areas such as smartpower grids and scheduling problems. Experimental results show that the proposed methodscould get good effects.