A new neural network paradigm called diagonal recurrent neural network (DRNN) structure is presented, and is used to design a neural network controller, which includes both a neuroidentifier (DRNI) and a neurocontroller (DRNC). An unknown plant is identified by a neuroidentifier, which provides the sensitivity information of the plant to a neurocontroller. A generalized dynamical backpropagation algorithm (DBP) is developed to train both DRNC and DRNI. An approach to use an adaptive learning rate scheme based on the Lyapunov function is developed. The use of adaptive learning rates not only accelerates the learning speed but also guarantees the convergence of the neural network.<>
{"title":"Diagonal recurrent neural network for controller designs","authors":"C. Ku, K.Y. Lee","doi":"10.1109/ANN.1993.264344","DOIUrl":"https://doi.org/10.1109/ANN.1993.264344","url":null,"abstract":"A new neural network paradigm called diagonal recurrent neural network (DRNN) structure is presented, and is used to design a neural network controller, which includes both a neuroidentifier (DRNI) and a neurocontroller (DRNC). An unknown plant is identified by a neuroidentifier, which provides the sensitivity information of the plant to a neurocontroller. A generalized dynamical backpropagation algorithm (DBP) is developed to train both DRNC and DRNI. An approach to use an adaptive learning rate scheme based on the Lyapunov function is developed. The use of adaptive learning rates not only accelerates the learning speed but also guarantees the convergence of the neural network.<<ETX>>","PeriodicalId":121897,"journal":{"name":"[1993] Proceedings of the Second International Forum on Applications of Neural Networks to Power Systems","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1993-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126581503","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
T. Kawagoshi, A. Kumamoto, T. Hikihara, Y. Hirane, K. Oku, O. Nakamura, S. Tada, K. Mizuki, Y. Inoue
The increase of power devices adopted in built-in power supplies is accompanied with an increase of many distributed harmonic generating loads and the harmonic problem in 6.6 kV power distribution lines is becoming prominent. To cope with harmonics, either conventional LC filters or active filters are used as additional compensators near the harmonic generating load. However such devices are not optimal when adopted as compensators in the power distribution system. The authors describe a neural network controlled active filter to realize both stable and adequate compensation for parameter variation due to impedance change or load variation and then discuss a computer simulation followed by the results obtained using a small scale laboratory model.<>
{"title":"Harmonic voltage suppression by active filter with neural network controller","authors":"T. Kawagoshi, A. Kumamoto, T. Hikihara, Y. Hirane, K. Oku, O. Nakamura, S. Tada, K. Mizuki, Y. Inoue","doi":"10.1109/ANN.1993.264343","DOIUrl":"https://doi.org/10.1109/ANN.1993.264343","url":null,"abstract":"The increase of power devices adopted in built-in power supplies is accompanied with an increase of many distributed harmonic generating loads and the harmonic problem in 6.6 kV power distribution lines is becoming prominent. To cope with harmonics, either conventional LC filters or active filters are used as additional compensators near the harmonic generating load. However such devices are not optimal when adopted as compensators in the power distribution system. The authors describe a neural network controlled active filter to realize both stable and adequate compensation for parameter variation due to impedance change or load variation and then discuss a computer simulation followed by the results obtained using a small scale laboratory model.<<ETX>>","PeriodicalId":121897,"journal":{"name":"[1993] Proceedings of the Second International Forum on Applications of Neural Networks to Power Systems","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1993-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125025279","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The authors discuss a knowledge base for power system stability in an AI/expert system environment. With the knowledge on parametric resonance, a power system is interpreted as a restoration characterized system, in which the 'jump phenomena' could occur due to the ill-combination of the system parameters. A stability index, called the restoration characteristics index (RCI), is derived by considering the specific combination of parameters to make the resonance curve triple-valued. And considerations on the monitoring and control strategy for AI/expert system approach are discussed.<>
{"title":"Monitoring and control strategy of power system stability based on the restoration characteristics index","authors":"Y. Tamura, Y. Huang, S. Tsukao","doi":"10.1109/ANN.1993.264339","DOIUrl":"https://doi.org/10.1109/ANN.1993.264339","url":null,"abstract":"The authors discuss a knowledge base for power system stability in an AI/expert system environment. With the knowledge on parametric resonance, a power system is interpreted as a restoration characterized system, in which the 'jump phenomena' could occur due to the ill-combination of the system parameters. A stability index, called the restoration characteristics index (RCI), is derived by considering the specific combination of parameters to make the resonance curve triple-valued. And considerations on the monitoring and control strategy for AI/expert system approach are discussed.<<ETX>>","PeriodicalId":121897,"journal":{"name":"[1993] Proceedings of the Second International Forum on Applications of Neural Networks to Power Systems","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1993-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115250218","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The BP (backpropagation) process is a popular learning algorithm for neural networks. Despite of many successful applications, the BP process has some known drawbacks. These drawbacks stem from that the BP process is a gradient based optimization procedure without a linear search. In this paper, a new learning algorithm is presented based on a solution method of nonlinear equations. Compared with the former optimization procedure, the proposed method often converges faster to desired results. Newton's method is basically applied to solve the nonlinear equations. However, the major difficulty with Newton's method is that its convergence depends on an initial point. In order to assure a global convergence, independent of an initial point, the Homotopy continuation method is employed.<>
{"title":"Learning algorithm for neural networks by solving nonlinear equations","authors":"K. Aoki, M. Kanezashi, C. Maeda","doi":"10.1109/ANN.1993.264305","DOIUrl":"https://doi.org/10.1109/ANN.1993.264305","url":null,"abstract":"The BP (backpropagation) process is a popular learning algorithm for neural networks. Despite of many successful applications, the BP process has some known drawbacks. These drawbacks stem from that the BP process is a gradient based optimization procedure without a linear search. In this paper, a new learning algorithm is presented based on a solution method of nonlinear equations. Compared with the former optimization procedure, the proposed method often converges faster to desired results. Newton's method is basically applied to solve the nonlinear equations. However, the major difficulty with Newton's method is that its convergence depends on an initial point. In order to assure a global convergence, independent of an initial point, the Homotopy continuation method is employed.<<ETX>>","PeriodicalId":121897,"journal":{"name":"[1993] Proceedings of the Second International Forum on Applications of Neural Networks to Power Systems","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1993-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129954007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The representation of load dynamic characteristics remains an area of great uncertainty and it becomes a limiting factor of power systems dynamic performance analysis. A major difficulty, both for component-based and measurement-based methods, is the lack of data for dynamic load modeling. A way of solving this problem for measurement-based methods is to interpolate and extrapolate the models identified from wide voltage variation data recorded during naturally-occurring disturbances or field experiments. This paper deals with data measured in Chinese power systems using two models: a multilayer feedforward neural network (ANN) with backpropagation learning, and difference equations (DE) with recursive extended least square identification. A comparison between the two approaches was done. The results show that the DE models interpolation and extrapolation are nearly linear, and they cannot describe the voltage-power nonlinear relationship of load dynamic characteristics. However, the ANN models can represent well this nonlinear relationship, they are promising dynamic load models.<>
{"title":"Comparison of dynamic load modeling using neural network and traditional method","authors":"He Ren-mu, A. Germond","doi":"10.1109/ANN.1993.264338","DOIUrl":"https://doi.org/10.1109/ANN.1993.264338","url":null,"abstract":"The representation of load dynamic characteristics remains an area of great uncertainty and it becomes a limiting factor of power systems dynamic performance analysis. A major difficulty, both for component-based and measurement-based methods, is the lack of data for dynamic load modeling. A way of solving this problem for measurement-based methods is to interpolate and extrapolate the models identified from wide voltage variation data recorded during naturally-occurring disturbances or field experiments. This paper deals with data measured in Chinese power systems using two models: a multilayer feedforward neural network (ANN) with backpropagation learning, and difference equations (DE) with recursive extended least square identification. A comparison between the two approaches was done. The results show that the DE models interpolation and extrapolation are nearly linear, and they cannot describe the voltage-power nonlinear relationship of load dynamic characteristics. However, the ANN models can represent well this nonlinear relationship, they are promising dynamic load models.<<ETX>>","PeriodicalId":121897,"journal":{"name":"[1993] Proceedings of the Second International Forum on Applications of Neural Networks to Power Systems","volume":"282 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131769312","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：