Yoshiakira Akimoto

Genetic algorithms approach to voltage optimization

The authors consider the use of genetic algorithms as a measure of voltage optimization of electric power system. Genetic algorithms are optimization and learning techniques based on natural selection and natural population genetics. A formation of a power system is encoded to a string of characters called an artificial chromosome the initial population of strings are generated at random, and then they are evolved by a genetic algorithm. The experiments with the prototype implementation are presented. These results verified the feasibility of genetic algorithms approach to power engineering.<<ETX>>

Fault diagnosis system for GIS using an artificial neural network

The authors present an artificial neural network (ANN) approach to a diagnostic system for a gas insulated switchgear (GIS). Firstly they survey the status of operational experience of failures in GISs and its diagnostic techniques. Secondly, they present how to acquire signal samples from the GIS and how to process them so as to be provided for an input layer of ANN. Finally they propose a decision-tree like network referred to as module neural network (MNN), and compare it with the well-known three-layered network, the straight forward neural network (SFNN).<<ETX>>

Distributed Power System Simulation on a Hypercube Computer

Abstract Distributed system is expected to realize characteristics of flexibility, high modularity, easy development, high speed processing and cost performance in the system of several fields. The use of distributed systems on parallel or distributed computers are studied for getting these characteristics. We developed a distributed simulation method of power systems to get high modularity such as easy attaching and detaching new components of power systems in the simulator for extension or maintenance. In this paper, we describe the detail of the distribution methods of power system simulation. First, we propose the distributed algorithm of solving network equations of power system simulation. Secondly, we describe the way to mapping tasks to be separated for distributed simulation to a hypercube computer. Thirdly, we show the flow of simulation including communication method with time stamp technique. Our approach is installed and evaluated in the hypercube computer, NCube/10. In the hypercube computer, each bus and each control circuit of generator in a power system are assigned to each processor. Finally, we show the result of installing and evaluating them.

A distributed simulator for power system analysis using a hypercube computer

Abstract The use of distributed systems is expected to realize characteristics of flexibility, high modularity, ease of development, high-speed processing and cost performance in several fields. Distributed systems on parallel or distributed computers are studied in order to obtain these characteristics. We have developed a distributed simulation method for the extension or maintenance of power systems in the simulator. In this paper, we describe the distribution methods of power system simulation. First, we propose a distributed algorithm for solving the network equations of power system simulation. Second, we describe a way of mapping tasks, which are to be separated for distributed simulation, to a hypercube computer, NCube/10. In the hypercube computer, each bus and control circuits of a generator in a power system are assigned to individual processors. Finally, we present the results of installing and evaluating the proposed algorithm and distributed simulation techniques on a hypercube computer.

Autonomous Distributed Network Architecture for Control System

Abstract With the advance of information technology, control system is going to be a large and complex one. We believe that key technology to solve it is distribution and autonomy. In this paper autonomous network architecture and its software design method for distributed control systems, with such autonomous functions as fault tolerance, flexibility, evolvability, extensibility, and conversely, partitionability, etc., are discussed. First, a network architecture is described. Then, the design method of network software with autonomous functions for distributed control systems is discussed. Autonomy of the system is realized in the network software, which is separated from the application software. Finally, the proposed network architecture and its software is installed in the distributed computer system composed of INMOS Transputers. Its autonomous functions, that is fault tolerance, etc., and the performance of control are evaluated in the distributed system, when a sorting algorithm and simple model of the load frequency control are applied.

Abnormality diagnosis of GIS using adaptive resonance theory

The paper presents an artificial neural network (ANN) approach using ART2 (Adaptive Resonance Theory 2) to a diagnostic system for gas insulated switchgear (GIS). To begin with, the authors show the background of abnormality diagnosis of GISs from the view point of predictive maintenance of them. Then, they discuss the necessity of ART-type ANNs, as an unsupervised learning method, in which neuron(s) are self-organized and self-created when detecting unexpected signals even if untrained by ANNs through a sensor. Finally, they present brief simulation results and their evaluation.<<ETX>>

Fault Diagnosis Of Gas Insulated Switchgear Using Adaptive Neural Networks

In this paper, Artificial Neural Networks (ANNs) approach to diagnostic methods for abnormality detection for Gas Insulated Switchgear (GIS) will be presented. The necessity of predictive maintenance, and current technologies of sensing devices and signal processing are discussed as the introduction. Thereafter, we will show the necessity of adaptive learning to achieve predictive maintenance which plays an important role as signal processing technology. ICLNN(lncrementa1 Cluster Learning Neural Network), that exhibits adaptive learning capability is proposed and evaluated. ICLNN conduct similar functionality as the convcntional clustering algorithm that classifies sensor signal in sclf-organising manner. Brief simulation results of the ICLNN conducted using the data obtained in a factory shows the great possibilities and availabilities of the ICLNN. Keyword: Gas Insulated Switchgear, Artificial Neural Network, ICL, ICLNN, Predictive Maintenance, Abnormality Diagnosis

APPROXIMATION CAPABILITIES OF NEURAL NETWORKS USING SAMPLING FUNCTIONS

Learning an input and output relationship, given an example of data, can be regarded as an approximation of a mapping function from input to output. In this point of view, we propose a neural network architecture using sampling functions as hidden units which reconstructs the multi-dimensional function. As this architecture employs sampling theory as the background, the number of hidden units required can be determined by the highest frequency of training data and faster learning can be obtained. 3 Summary We have proposed the neural network architecture using sampling functions in the hidden layer and its learning algorithm. Based on the idea that learning an input and output relationship of training data can be regarded as synthesizing an approximation of a mapping function, we have transformed the mapping function into the representation composed of sampling functions working as basic functions, and then, devised a learning algorithm for irregularly obtained training data. No learning algorithm is needed for regularly obtained training data, that is, every weight in the network can be determined definitely. The simulations have proved that our architecture makes dramatically fast learning possible.