Dejan J. Sobajic

High-speed fault detection and classification with neural nets

This paper introduces a new neural net (NN) approach for automated fault disturbance detection and classification. The NN design and implementation are aimed at high-speed processing which can provide selective real-time detection and classification of faults. The approach is extensively tested using the Electromagnetic Transients Program (EMTP) simulations of two quite complex transmission system configurations. The results indicate that the speed and selectivity of the approach are quite adequate for a number of different transmission and distribution monitoring, control, and protection applications.

Multi neural network based fault area estimation for high speed protective relaying

The aim of this paper is to present a new approach to fault area estimation for high-speed relaying using feedforward neural networks. The suggested framework makes use of neurocomputing technology and pattern-recognition concepts. In contrast to conventional algorithms, our neural fault area estimator (NFAE) determines the fault area directly. This approach leads to very short propagation times and reliable classification results. Important attributes of artificial neural networks (ANNs) are their ability to learn nonlinear functions and their large input error tolerance. The obtained results indicate that these characteristics still result in reliable behaviour even if nonideal (real-world) effects pertain. A comparison of classification quality with conventional algorithms by simulating certain faults on a parallel transmission line shows the approaches advanced capability for protective relaying.

Determination of generator groupings for an islanding scheme in the Manitoba Hydro system using the method of normal forms

This paper deals with the application of the method of normal forms in the analysis of a specific aspect of system dynamic behavior in the Manitoba Hydro system. Following a major loss of transmission capacity on the Manitoba Hydro HVDC system (Nelson River system), and the subsequent operation of protection systems, there is a major deficit of generation in the remaining system, comprising Manitoba and Saskatchewan. The method of normal forms is applied to determine the natural groupings which are formed by the machines in Manitoba Hydro due to nonlinear interaction. This grouping then provides a basis for developing a systematic procedure to island the remaining system.

Neuro-fuzzy controller of low head hydropower plants using adaptive-network based fuzzy inference system

This paper presents an attempt of nonlinear, multivariable control of low-head hydropower plants, by using adaptive-network based fuzzy inference system (ANFIS). The new design technique enhances fuzzy controllers with self-learning capability for achieving prescribed control objectives in a near optimal manner. The controller has flexibility for accepting more sensory information, with the main goal to improve the generator unit transients, by adjusting the exciter input, the wicket gate and runner blade positions. The developed ANFIS controller whose control signals are adjusted by using incomplete on-line measurements, can offer better damping effects to generator oscillations over a wide range of operating conditions, than conventional controllers. Digital simulations of hydropower plant equipped with low-head Kaplan turbine are performed and the comparisons of conventional excitation-governor control, state-feedback optimal control and ANFIS based output feedback control are presented. To demonstrate the effectiveness of the proposed control scheme and the robustness of the acquired neuro-fuzzy controller, the controller has been implemented on a complex high-order nonlinear hydrogenerator model.

Neural-net based coordinated stabilizing control for the exciter and governor loops of low head hydropower plants

This paper presents a design technique of a new adaptive optimal controller of the low head hydropower plant using artificial neural networks (ANN). The adaptive controller is to operate in real time to improve the generating unit transients through the exciter input, the guide vane position and the runner blade position. The new design procedure is based on self-organization and the predictive estimation capabilities of neural-nets implemented through the cluster-wise segmented associative memory scheme. The developed neural-net based controller (NNC) whose control signals are adjusted using the on-line measurements, can offer better damping effects for generator oscillations over a wide range of operating conditions than conventional controllers. Digital simulations of hydropower plant equipped with low head Kaplan turbines are performed and the comparisons of conventional excitation-governor state-space optimal control and neural-net based control are presented. Results obtained on the nonlinear mathematical model demonstrate that the effects of the NNC closely agree with those obtained using the state-space multivariable discrete-time optimal controllers.

Neural-net based real-time economic dispatch for thermal power plants

This paper proposes the application of artificial neural networks to real-time optimal generation dispatch of thermal power plant units. The approach can take into account operational requirements and power network losses. The proposed economic dispatch uses an artificial neural network (ANN) for the generation of penalty factors, depending on the input generator powers and identified system load change. Then, a few additional iterations are performed within an iterative computation procedure for the solution of coordination equations, by using reference-bus penalty-factors derived from Newton-Raphson load flow. A coordination technique for environmental and economic dispatch of pure thermal power systems, based on neural net theory for simplified solution algorithms and an improved man-machine interface is introduced. Numerical results on two test examples show that the proposed algorithm can efficiently and accurately develop optimal and feasible generator output trajectories by applying neural net forecasts of power system load patterns.

Coordinated stabilizing control for the exciter and governor loops using fuzzy set theory and neural nets

Abstract This paper presents a design technique for a new hydropower plant controller using fuzzy set theory and artificial neural networks. The controller is suitable for real time operation, with the aim of improving the generating unit transients by acting through the exciter input, the guide vane and the runner blade positions. The developed fuzzy logic based controller (FLC) whose control signals are adjusted using the on-line measurements, can offer better damping effects for generator oscillations over a wider range of operating conditions than conventional regulators. Digital simulations of a hydropower plant equipped with a low-head Kaplan turbine are performed and the comparisons of conventional excitation-governor control, optimal state-feedback control and FLC performances are presented. The FLC, based on a set of fuzzy logic operations that are performed on controller inputs, provides a means of converting linguistic control requirements based on expert knowledge into an efficient control strategy. A fuzzy associative matrix is generated by using unsupervised learning of artificial neural networks. Results obtained on the nonlinear hydrounit mathematical model simulation demonstrate that the performance of the FLC closely agrees with that obtained if the optimal state-feedback multivariable discrete-time controller is applied.

A neural-net based short term load forecasting using moving window procedure

Abstract An improved neural-net approach based on a combined unsupervised/supervised learning concept is proposed. A ‘moving window’ procedure is applied to the most recent load and weather information for creating training set data base. A forecasting lead time that varies from 16 hours to 88 hours is introduced to produce the short term electric load forecasting that meets requirements of real electric utility operating practice. The unsupervised learning (UL) is used to identify days with similar daily load patterns. A feed forward three-layer neural net is designed to predict 24-hour loads within the supervised learning (SL) phase. The effectiveness of proposed methods is demonstrated by comparison of forecasted hourly loads in every single day during 1991 with data realized in the same period in the Electric Power Utility of Serbia (EPS). A better choice of input features and more appropriate training set selection procedure allow significant improvement in forecasting results comparing with our previous UL/SL concept characterized by a fixed neural-net structure and absence of re-training procedure. The improvement is illustrated by reduction of average error in daily energy forecasting for 0.83% and reduction of 90th percentile of 2.04%.

Fuzzy linear programming based optimal fuel scheduling incorporating blending/transloading facilities

In this paper, power system blending/transloading facilities are modeled using an interactive fuzzy linear programming (FLP), in order to allow the decision-maker to solve the problem of uncertainty of input information within the fuel scheduling optimization. An interactive decision-making process is formulated in which decision-maker can learn to recognize good solutions by considering all possibilities of fuzziness. The application of the fuzzy formulation is accompanied by a careful examination of the definition of fuzziness, appropriateness of the membership function and interpretation of results. The proposed concept provides a decision support system with integration-oriented features, whereby the decision-maker can learn to recognize the relative importance of factors in the specific domain of optimal fuel scheduling (OFS) problem. The formulation of a fuzzy linear programming problem to obtain a reasonable nonfuzzy solution under consideration of the ambiguity of parameters, represented by fuzzy numbers, is introduced. An additional advantage of the FLP formulation is its ability to deal with multi-objective problems.

On external network model development

Although state estimation is a mature technique which is widely available in the EMS industry, experience with practical implementation and use of state estimation and network security analysis functions at electric power utilities over the last two decades indicates recurring difficulties and problems attributable to inadequate external network modeling detail and data. This paper addresses the development of a set of guidelines for external network modeling and data exchange, based on the results of a recent project sponsored by EPRI. A general methodology is developed based on the results of a survey of a representative set of utilities and EMS suppliers, supplemented by subsequent analysis and simulation studies. Distinction is made between guidelines pertaining to external network topology, analog measurements, data exchange, and implementation procedures. A philosophy and approach for constructing and testing external models is also presented.