A FUZZY EXPERT SYSTEM FOR QUANTIFYING VOLTAGE QUALITY IN ELECTRICAL DISTRIBUTION SYSTEMS

Transcription

1 A FUZZY EXPERT SYSTEM FOR QUANTIFYING VOLTAGE QUALITY IN ELECTRICAL DISTRIBUTION SYSTEMS Fuat KÜÇÜK, Ömer GÜL Department of Electrical Engineering, Istanbul Technical University, Turkey ; SUMMARY Power quality is usually taken as a voltage quality. When the impacts on voltage quality are known, some information about power quality can be obtained. Various types of device connected to power distribution systems expect good voltage quality, but their expectation levels may differ. Device sensitivities to the voltage quality problems are usually given with the curves. Commonly used for the computer systems is the CBEMA curve. Using device sensitivity characteristics can be a reasonable way to represent voltage quality. CBEMA curve strictly separate acceptable - unacceptable regions, and should be fuzzied. Different methods can be tried for evaluation of voltage quality. Because of including vagueness and uncertainty voltage quality has chance to be studied with fuzzy mathematics. quality problems can be separately examined. If the effect of each individual problem on voltage quality is known, then general voltage quality can be considered as superposition of individual voltage qualities. In this work, a method was proposed for quantifying the voltage quality. To achieve this, software was written using Matlab and its Fuzzy Toolbox. 1. INTRODUCTION There are several definitions for power quality, depending on producing or consuming power. According to customers, good power quality has voltage, current and frequency that do not cause failure or miss operation of any equipment. Power quality is usually taken as a voltage quality. [1,2]. Devices connected to a power utility expect good voltage quality. Because their sensitivities to voltage quality problems are different, their quality expectation levels are different. sags, swells, outages and harmonics are the main parts of the voltage quality problems. They caused fault or malfunction depending on voltage variations and its duration their repetitions also have effect on voltage quality. This type of problems usually rises from nonlinear loads such as adjustable speed drives, electronic ballasts for fluorescent lamps and power supply for welding machines, events such as momentary supplying large power loads and short circuit in power utilities. Although various studies and standards (some of them given in reference [3-7]) have been written about voltage (or power) quality problems, there are few works related to voltage quality evaluation. Because of including vagueness and complexity, voltage quality evaluation is difficult with classical methods, but has a great chance to study with fuzzy mathematics. In one of the works about voltage quality evaluation with fuzzy logic, voltage quality is quantified just for one voltage sag [8]. The other works are complicated and do not say which type reference sensitivity characteristics were used [9, 10]. This paper commences with some useful information about voltage quality problems and fuzzy mathematics, and then continuous developed method for voltage quality evaluation with fuzzy logic. Finally it is shown with an example that developed method can give successive results. 2. FUZZY MATHEMATICS Fuzzy logic includes fuzzy set theory and rules. In conventional set theory, belonging and not belonging to any set is denoted with 1 and 0 respectively. There is no more choice between them. This is also called crisp set. It is not possible to express the vagueness and uncertainty with the crisp set. Fuzzy sets can manage this problem. Fuzzy sets are considered to be an extension of crisp sets [9]. The main part of the fuzzy set is the membership functions. Belonging degree of value x to the set A is determined by the membership function, where set A is defined on a universal set. This definition of fuzzy set and membership function given as follows: µ A : X {0,1} (1) µ A (x) is called membership value or the grade of membership of x œ X. The universal set (or variable) X may include one or more membership functions. Value x may belong to two sets at the same time. The inference of variables formed by membership functions is defined by if then rules. To perform the fuzzy logic, inference rules are needed such as: IF x is A and y is B THEN z is C (2) where A, B, C are fuzzy sets and x, y, z are variables. 3. USING DEVICE SENSITIVITY CHARACTERISTIC FOR VOLTAGE QUALITY Devices connected on a power utility may have different sensitivities to voltage quality problems. For example, voltage including sags and harmonics has not equal quality for computers and motors.

2 It is important to know which voltage quality problem is acceptable. In other words, which voltage quality problem has effects on device working with fault or malfunction. This is usually given with envelope or curve and called safety operating region. One of the most important curves is the voltage deviation versus duration curve. Commonly used for the computer systems is the CBEMA curve, which is shown in Figure 1. for the fuzzy logic part. Software can be divided into 3 parts. In the first part, it detects and separates voltage quality problems while inspecting the voltage. In the second part, each fuzzy section takes its quality problem and gives individual voltage quality values VQ 1, VQ 2, VQ 3... VQ n. In the third part, these quality values are multiplied by coefficients k 1, k 2, k 3,... k n. These coefficients are described as weighting factors that give importance degree of the voltage quality problem. Their values change from 0 to 1 and should be defined by customer. Finally, software gives a general voltage quality value. Power utility Sampling Figure 1. CBEMA curve When looked in the Figure 1, it can be seen that acceptable and unacceptable regions are strictly separated. That means, any inside point near to curves is in acceptable region but any outside point near to curves is not. Surely, there is a little difference between them. In practice, device may work normal at any outside point near to curve and may not work any inside point near to curve. For this reason, the curves should be fuzzied. That is, near to curves should be divided into regions. It can also be mentioned about curves for other voltage quality problems, such as harmonics. In the literature, it has not been encountered any curve about harmonic sensitivity of devices. Harmonic sensitivity is as important as voltage sensitivity. For this reason, harmonic tolerance that device can work without any failure should be known. Using device sensitivity characteristics may be a reasonable way to represent voltage quality. If device sensitivities to voltage quality problems are known, general voltage quality can also be evaluated. 4. PROPOSED METHOD FOR QUANTIFYING VOLTAGE QUALITY quality problems can be separated into its components such as voltage sags, swells, harmonics etc. If each variation, duration, repetition are known for each problem, individual voltage quality is easily evaluated by fuzzy logic. Then a general power quality can be found multiplying each voltage quality by coefficients. Basic flow chart of the proposed method is given in Figure 2. To evaluation of the voltage quality, software has been written using Matlab. Fuzzy Toolbox has been also used sags VQ 1 swells quality problem identification FL 1 FL 2 FL 3 FL n VQ 2 harmonics VQ 3 VQ n k 1 k 2 k n General Quality In this work, three types of voltage quality problems were considered (Figure 2). Input variables of each fuzzy system are as follows: voltage sags - duration, voltage swells duration, and voltage harmonics duration. The output variables are the individual voltage qualities named VQ 1, VQ 2 and VQ 3. Figure 3, Figure 4 and Figure 5 give membership functions of input and output variables used for fuzzy section of the voltage sag. k 3 Figure 2. Flowchart of proposed method

3 µ VS CL L N Magnitude Figure 3. Membership functions for magnitude of the voltage sag 1 CS VS S L TABLE 1 Meanings of the fuzzy sets for voltage sag Magnitude CL Completely Low Very Low L Low N Normal CS Completely Short VS Very Short S Short L Long Very Long Quality CP Completely Poor VP Very Poor P Poor NG Nearly Good G Good TABLE 2 Rule base for the voltage sag Figure 4. Membership functions for duration of the voltage sag µ VQ CP VP P NG G 1 Magnitude CL L N CS G G G G VS VP P NG G S CP VP P G L CP CP VP G CP CP CP G Quality Figure 5. Membership functions for quality of voltage sag As seen in the figures, four and five membership functions are used for input variables while five membership functions are used for the output variable. All the membership functions are chosen in trapezoidal type. Because of the number of membership function of the input variables, 20 rules are needed. During the design of membership functions and the rules, practical applications and the CBEMA curve were considered. The two input variables, voltage magnitude and duration are expressed in percentage from 0 to %100 and in second from 0 to 60 s respectively. Besides, output variable, voltage quality is expressed in percentage from 0 to % 100. Table 1 and Table 2 explain the meaning of each fuzzy set and the rules respectively. 5. CASE STUDY A power distribution system is studied in this case study. Its nominal values are 380 V, 50 Hz. The rms variation of the voltage and its harmonic are shown in Figure 6 and Figure 7 respectively. rms and its total harmonic distortion are sampled in about 70 s while sampling time is 0.01 s. Frequency value is assumed to be constant. rms (V) Time Figure 6. Rms variation of the voltage used in case study

4 The weighting coefficients are chosen the same value for this case study. That is, all voltage quality problems have the same effect on the voltage quality evaluation. harmonic TABLE 4 Results for voltage sag Average Rated Quality Time Figure 7. Variation of voltage harmonic Table 5 presents voltage quality according to the problems. It also gives general voltage quality while taking all weighting factors 1. When software runs, general voltage quality can be obtained. Not only software gives the general voltage quality, it can also give average values of each voltage sag, swell, harmonics, their durations and individual voltage qualities. Table 3 and Table 4 show each event, duration and individual voltage quality of voltage sag and swell. TABLE 5 - General results Problem Quality Weighting factors Sag 80 1 General Quality TABLE 3 Results for voltage swell Average Rated Quality Swell 53 1 Harmonic CONCLUSION A method including fuzzy mathematics was proposed for quantifying voltage quality. For this case, software was written using Matlab and Matlab Fuzzy Toolbox. Proposed method was applied to the voltage that was sampled from a 380 V, 50 Hz power distribution system. While using the CBEMA curve, acceptable and unacceptable regions were fuzzied. Three types of voltage quality problems have been considered: voltage sags, voltage swells and voltage harmonics. Each event of voltage quality problems was detected according to the average value and duration. The particular and general results were presented in case study part. General voltage qualities were expressed considering device sensitivity curves and weighting factors. The weighting factors were taken 1 for the case study, but it may change according to the customer s devices. If it is needed, general voltage quality levels may be constructed. Before any poor voltage caused a damage or malfunction, these quality levels can be used for protection or can allow taking steps. 67 In this work, voltage quality was found for about 70 s time interval. Besides, voltage quality values are only for the computer based devices, because CBEMA curve is used while constructing the membership functions. If it is intended to find voltage quality value according to the

5 various types of devices, membership functions should be reconstructed. Time interval should also be extended to the hours, days or years. Then, more accurate and more applicable voltage quality value will be obtained. REFRERENCES [1] R. C. Dugan, 2002, Electrical Power Systems Quality, McGraw Hill [2] B. W. Kennedy, 2000, Power Quality Primer, McGraw Hill [3] IEEE Recommended Practice for Monitoring Electric Power Quality, IEEE Standard [4] IEEE Recommended Practice for Evaluating Electric Power System Compatibility with Electronic Process Equipment, IEEE Standard [5] G. T. Heydt, R. Ayyanar, R. Thallam, 2001 Power Acceptability, IEEE Power Engineering Review, v.21 p [6] Y. Liaoa, J. B. Leeb, 2004, A Fuzzy Expert System for Classifying Power Quality Disturbances, Elsevier Electrical Power and Energy Systems, v. 26 p [7] P.K. Dash, I. L. W. Chun, 2003, Power Quality Data Mining Using Soft Computing and Wavelet Transform, TENCON Conference on Convergent Technologies, v. 3, p [8] B. D. Bonatto, T. Niimura, 1998, A Fuzzy Logic Application to Represent Load Sensitivity to Sags, IEEE 8th International Conference on Harmonics and Quality of Power, v. 1, p [9] K. Tanaka, 1997, An Introduction to Fuzzy Logic for Practical Applications, Springer