IDENTIFICATION AND ELIMINATION OF HARMONICS IN POWER SYSTEMS

Abstract

This thesis focuses on reporting the research study on a problem area relating to the identification and elimination of harmonics in power systems. With the everincreasing use of the power electronic devices for more efficient operation of the electrical equipments, there has been a steady increase of harmonics in the power system. These power electronic controllers operate on the sinusoidal supply voltage but in return inject harmonics into the supply system. The injected harmonics are responsible for the distortion of the voltage and the current wave shapes. Primarily, these harmonics are of characteristics type, and a very substantial amount of these come from the high rating power rectifiers. Other devices viz. static var compensators, transformers, welding set, arc furnaces, cycloconverters etc. are some of the major devices responsible for the injection of harmonic currents at the point of common coupling in the power system. The loads responsible for the injection of harmonics in the system are termed as nonlinear loads. These nonlinear loads along with the distortion of the voltage and current wave shapes, are responsible for the increase in the overall system losses, malfunctioning of the protective devices, destruction of the reactive power compensating capacitors, creating dangerous operating conditions by means of parallel resonance between the supply system reactance and the power factor improvement capacitors. Therefore the research in the area of the harmonics propagation, harmonic sources identification, monitoring of the harmonic sources, placement of meters for the estimation of the harmonics and the elimination of harmonics has assumed a significant importance in recent times. In the present thesis methods for the identification of multiple unknown harmonic sources, placement of limited harmonic meters for static and dynamic harmonic state estimation of the power system and system wide harmonics elimination with optimally designed single tuned passive filters have been developed. Once the exact locations of the harmonic sources are known, harmonics level and resonance conditions if any in the system can be easily ascertained by means of harmonic load flow studies. Thereafter depending upon the desired requirement to bring down the level of harmonics, harmonic reduction techniques can be applied. The work accomplished in this thesis is as given below. Harmonic source identification is essential for designing effective harmonic mitigation schemes and for determining who is exactly responsible for the injection of harmonics in the system. The majority of the harmonic source identification techniques available in the literature assume a priori information about the existence of the harmonic sources in the power system. Consequently, the important question about the feasibility of identification of harmonic sources in the absence of any a priori knowledge is yet to be addressed. Hence, in the present thesis an attempt has been made to address the harmonic sources identification problem in the absence of a priori information about the existence of these sources. Towards this goal, a two-stage system wide state estimation technique is developed to identify the exact locations of multiple harmonic sources in a power system. Sometimes there exits a priori information about the loads in the power system. This information is useful for estimating the harmonic sources with limited optimally placed harmonic meters. Subsequently, the static or dynamic estimate of the harmonic sources can be carried out with limited meters. This type of state estimation of harmonic sources with limited harmonic meters is useful owing to reduced cost on the harmonic meters. The problem of placement of limited harmonic meters is formulated as an optimization problem that minimizes the expected value of the sum of squares of differences between estimated and actual parameter variables. Since the number of measurements is small the harmonic sources estimation problem is essentially an underdetermined one. The solution of such an underdetermined problem will be accurate only when the measurements used in the estimation are taken from the optimal locations within the system. To solve such an underdetermined problem some techniques based upon the minimization of the trace value of the covariance matrix of the unmeasured bus harmonic currents have been reported in the literature. The existing techniques in the literature for the placement of the limited harmonic meters are either exhaustive or are able of determining a near optimal solution. The complete enumeration technique (exhaustive) has the advantage of finding the optimal meter locations over the minimum variance and sensitivity meters placement techniques at the cost of computational burden. Therefore, there is a need to develop a new method for the placement of harmonic meters 11 giving the same solution as obtained by complete enumeration technique with much less computational burden. To fulfill the requirement of the placement of harmonic meters a Genetic Algorithm (GA) based optimal meter placement technique has been developed in the present thesis. The meter locations obtained by the proposed GA technique are the same as has been obtained by mean of complete enumeration. The GA technique is compared with other existing techniques available in the literature and it is found that GA outperforms all other techniques either in terms of computational burden or in term of the estimation accuracy. Once the harmonic meters are placed optimally within the power system the level of the unknown harmonic injection currents can be determined either of the following techniques i.e. minimum variance technique, sensitivity analysis or by means of Kalman filtering. Since the cost of the harmonic meters is quite high compared to their counterparts used for the measurements of fundamental quantities, monitoring the harmonic sources with limited meters is a very challenging task. Estimation of harmonic sources with already published methods viz. minimum variance technique, sensitivity analysis or Kalman filtering require two types of measuring devices i.e. devices able to measure harmonic voltages and currents. In case either of these two harmonic meters is not available then the known methods for the monitoring the known harmonic sources may prove to be inadequate. Use of artificial neural networks (ANNs) for harmonic source monitoring with limited harmonic meters, without any criteria for the meters placement in the power system has been suggested in the literature. Therefore, there is a need to develop a method to monitor the known harmonic sources with optimally placed available harmonic meters and to detect the presence of addition of a new harmonic source in the power system. In order to fulfill this requirement a method has been developed to monitor the known harmonic sources, detecting and finally identify the addition of a new harmonic source in the system with optimally placed harmonic meters. Basic problem is to place the limited harmonic meters optimally such that these meters can monitor the known harmonic sources with least possible measurement errors. For the purpose an optimal meter placement methodology based on the harmonic line currents sensitivity to the known injected harmonic currents is developed. The harmonic meters are placed on the most sensitive lines such that the meters are able to monitor the in harmonic sources with least possible measurement errors. In other words the estimated injected harmonic source currents are close to their true values. The knowledge of the injected harmonic source currents help to ascertain the level of the harmonics in the power system as well as to design suitable compensators. Since the available field measurements are limited, monitoring of the harmonic sources is an underdetermined problem. Therefore, application of trained artificial neural networks (ANNs) has been proposed to monitor the known harmonic sources. The data measured by the optimally placed harmonic meters act as the input to the initially trained feed forward neural networks. The ANNs are then able to monitor the known harmonic sources with least possible estimation errors. The advantage of this scheme is that only single type of harmonic meters is sufficient to accomplish the monitoring task. This kind of study is very useful when the accurate level of the harmonics is essential for the design of harmonic compensators for the known harmonic sources and to determine whether any resonant conditions do exist within the system. Occasionally the customer may add new harmonic source to the supply system without the knowledge of the concerned utilities. The present utilities are facing difficulties to detect the presence of the harmonic source added without their prior knowledge. Knowledge of the addition of such harmonic source is essential in order to take the preventive actions against the problems caused by such load on the supply system. To accomplish this important task of detecting the presence of newly added harmonic source, 'harmonic line current error index' based method has been developed in the present thesis. In case the harmonic current error index exceeds a pre-defined limit, the addition of an unknown harmonic source in the power system is indicated. The exact identification of the detected unknown harmonic source is essential for the appropriate preventive steps as well as for the design of harmonic filters. For this purpose, a two-stage methodology has been developed in this thesis to identify the exact location of the newly added unknown harmonic source. In the first stage of the developed method, tentative harmonic source locations at which the harmonic source could be present are determined by means of minimum variance technique. The data requirement for this technique is the harmonic line currents given by the permanently installed harmonic meters and the harmonic current injections of the already known harmonic IV sources. In the second stage the final verification of the exact location of the unknown harmonic source is done by means of static state estimation technique. Since the power system is a dynamic system, determination of the level of harmonics in real time is essential. The literature on fundamental frequency power system state estimation is quite rich but dynamic harmonic state estimation is still wanting. A precise determination of system wide harmonics magnitude and phase angles in real time is required in many practical situations such as for tuning passive filter or for controlling active filter. Therefore, an attempt has been made to estimate the system wide harmonic states dynamically with limited harmonic meters along with pseudomeasurements. Towards this goal, a Robust Extended Kalman filter has been proposed to estimate the harmonics in the power system in real time. The dynamic state estimation of harmonics is robust and the results obtained in the presence of different operating conditions (viz. normal operating condition, sudden load change and bad data conditions) are very close to their true simulated values. Harmonics if within the IEEE-519 limits need limited attention, but when the level of the harmonics present in the power system exceeds the prescribed limits proper filters installation at the optimal locations need attention. The filters locations and sizing problem is to be such that the optimal locations and sizes of these are determined in single stage itself. For this purpose in the present thesis, multi-objective optimal filters locations and sizing problem of single tuned filters has been solved using a nondominated sorting genetic algorithm (NSGA-II). The proposed NSGA-II algorithm determines the optimal values of the passive tuned filters in single stage. Another major advantage of the proposed methodology is that the effect of the harmonic sources phasors is also included in the design of the single tuned filters. The proposed method is able to give optimal designed values of the filters of a large distribution as well as an industrial power system. To conclude, methods have been developed for: • Finding the locations of the completely unknown harmonic sources in the power system using a two stage harmonic state estimation technique • Genetic algorithm based meter placement for static estimation of harmonic source Optimal placement of harmonic meters for estimation of harmonics in order to achieve the following (a) Monitoring the known harmonic sources with limited harmonic meters (b) Detecting the presence of the newly added harmonic source(s) (c) Identifying the exact locations of the unknown newly added harmonic source(s) Robust dynamic state estimation of power system harmonics Optimum placement and sizing of passive filters using NSGA-II for harmonics elimination