METHOD TO DETERMINE CONTRIBUTION OF THE CUSTOMER AND THE POWER SYSTEM TO THE HARMONIC DISTURBANCE

C I R E D 17 th International Conference on Electricity Distribution Barcelona, 1-15 May 3 METHOD TO DETERMINE CONTRIBUTION OF THE CUSTOMER AND THE POWER SYSTEM TO THE HARMONIC DISTURBANCE Olivier GONBEAU Luc BERTHET Jean-Louis JAVERZAC Denis BOUDOU Electricité de France France Réseau de Transport Electrique - France Electricité de France - France olivier.gonbeau@edf.fr luc.berthet@edf.fr jean-louis.javerzac@rte-france.fr denis.boudou@edf.fr 1- INTRODUCTION Harmonic levels are becoming an essential element among Power Quality parameters. Indeed, the number of non-linear loads does not stop increasing as well the injection of harmonic currents into the distribution and transmission system. Harmonic voltage levels should be maintained below the compatibility levels so that all the equipment supplied with the public networks can operate under satisfactory conditions. In addition to limits by equipment, there is already a tendency towards the definition of limits by installations. It becomes necessary to be concerned with control methods associated with these limits expressed in term with harmonic current. However, the measured harmonic current at the customer point of common coupling is the vectorial sum of a current emitted by this installation towards the network and a current coming from the network. Within a contractual framework, only the current emitted by a customer s installation towards the network should be taken into consideration. Thus, it is necessary to make use of a measurement method making it possible to extract this current from the current measured at the point of common coupling. To date, there s no method recognized at the international level with this intention. Theoretical work was undertaken on this subject since the beginning of the 9 s [1]. Now, it remains to validate a method (even several) by defining their algorithms, their application field and their limits and then to obtain an international consensus. To be able to determine the exact customer contribution in all cases, EDF studied new estimation methods. In addition, new algorithms for application of these measurement methods, as well as existing methods, were developed. This article s aim is to present these estimation methods and the first results. In this article we present successively: the regulatory context, harmonics, the estimation methods, algorithms for application of all these methods, first results obtained in the network, prospects. - REGULATORY CONTEXT International standards and national decrees are available proposing guidelines to fix harmonic current emission limits in order to respect harmonic compatibility levels. The technical report IEC 61-3-6 [] gives general steps to elaborate connection rules for non-linear loads. Thus, the electrical public networks could run under satisfactory conditions for its users. Historically in France, the Émeraude contract was launched in 1995 after a negotiation between the public utility and the representatives of large companies. This contract, approved by the public authorities, was extended to all HV and MV customers whose subscribed power exceeds 5 kva. As far as harmonics are concerned, this contract only gives informative connection rules for the customers. Moreover, the Émeraude contract forecasted that the informative connection rules would become obligatory from 1998. More recently, the February law (opening the French electric market) allows connection and access to the electrical power supply networks of the various users (specifically generators). Moreover, this law provides that technical methods of connection must be specified by decrees (already done for generators). Mandatory connection rules will therefore see the light of day in a few months. These rules will describe commitments of the both parties in terms of power quality and particularly harmonics. In conclusion, the evolution of the French national regulatory context will require measurement methods to check the network operator s and users reciprocal commitments. 3- HARMONICS 3.1- Harmonic flow into power supply network Harmonics are a by-product of modern electronics. When this equipment is connected to a 5 Hz sinusoidal voltage, these loads don t absorb a 5 Hz sinusoidal current. Non-linear load connection to the network results in the apparition of a harmonic voltage at the network impedance. Generally, load impedances are definitely larger than the line ones. Harmonic currents tend to circulate towards upstream network. In the industrial installations, capacitor banks are connected in parallel to compensate the reactive power of inductive loads. These capacitor banks can form resonant circuits with line inductances, which can amplify the disturbances and modify the natural harmonic flow of the current. EDF_Gonbeau_A1 Session Paper No 3-1 -

C I R E D 17 th International Conference on Electricity Distribution Barcelona, 1-15 May 3 3.- Problematic with the measurement of the currents emitted by installations To check if an installation s harmonic emissions are in conformity with the regulations or standards, it is necessary to differentiate with precision harmonic current emitted downstream from the current measured. Downstream part : C Installation or customer side Ih=Ihs+Ihc Ihs Vh Ihc Upstream part : S Source side Figure 1: Power supply system model at the PCC the "downstream" part stands for the installation. With: Js : Harmonic current from the source side : Harmonic equivalent impedance on source side Jc : Harmonic current from the installation Zc : Harmonic equivalent impedance on the installation side Harmonic measurement are performed following the IEC 61-4-7 recommendations [3]. The harmonic voltage and current measurement are the defined by: ( Js Ih(i) ) ( Jc Ih(i) ) Vh(i) =. (1) Vh(i) = Zc. + () According to this model, three methods will be presented. The first one (described in [1]) uses the sign of R eal( Vh Ih). The two others (non-linear regression and iterative methods) were imagined by EDF recently. Unfortunately, the current measured (Ih) at the point of common coupling is the sum of the current emitted by the customer (Ihc) towards the network and the network towards the customer (Ihs). In general, the current emitted by the customer is very dominating but not always; in certain cases, the current from upstream can be notable because of a resonance. 4- ESTIMATION METHODS 4.1- Harmonic single-phase equivalent circuit at the point of common coupling From the network represented by Figure 1, a measurement at the point of common coupling gives: Vh : Complex harmonic voltage, Ih : Complex harmonic current. At this point of measurement, the model of Figure can characterize the network: Ih 4.- Method using the sign of R eal( Vh Ih) The sign of eal( Vh Ih) R determines the side where a parameter variation didn t occur and by this way establishes the impedance value that did not vary [1]. Determination Domain 18 Vh Real sign Ih - + 9-9 Determination Domain Zc Figure 3: Impedance determination domains As shown in Figure 3, the ratio Vh Ih provides: The impedance if Vh The impedance Zc if eal( Vh Ih) 4.3- Non linear regression method R < R > Js Vh Zc Jc The problem resolution is to obtain, Zc, Js, Jc through the minimization of the errors defined by these equations: Source Installation Figure : Single-phase equivalent circuit for harmonic analysis. The source stands for the "upstream" part of the measurement point, which often represents on the same voltage level the parallel of the network with other harmonic injections, and e 1 (i) = Vh(i) Js + Ih(i) (3) (i) e We can minimize for example: = Vh(i) ZcJc ZcIh(i) (4) EDF_Gonbeau_A1 Session Paper No 3 - -

C I R E D 17 th International Conference on Electricity Distribution Barcelona, 1-15 May 3 1 = (5) i ( ) ( Vh(i) Js + Ih(i) ) e (i) + e (i) i + ( Vh(i) ZcJc ZcIh(i) ) 4.4- Iterative method Equation (1) provides Vh. ( Js Ih) =. This corresponds to an equation with two unknown factors. The principle of the iteration method is to apply the harmonic current and voltage measurements in equation (1) with one unknown factor fixed to a value to calculate the other. Thus, the estimation method allows harmonic impedance network estimation ( ) at one iteration and the harmonic current from source side estimation ( Js ) at the next iteration using the impedance previously computed. 5- ALGORITHMS 5.1- Application fields Harmonic currents and voltages measurements must rely on standard recommendations IEC 61-4-7 [3]. The three methods described previously correspond to different application fields but are theoretically complementary. The goal is to obtain for each one an algorithm that controls the results accuracy. Thus, the method using the sign of eal( Vh Ih) R gives good results in case of a significant variation of the harmonic currents and voltages. The results are provided just after the variation. Thus, three consecutive measurement windows are sufficient to obtain an estimate. Concerning the non-linear regression method, the error minimization passes through least squares resolution. This requires a great number of measurements samples without a significant parameters variation. The iterative method follows the evolutions of the estimated parameters with time with or without harmonic variations. 5.- Method using the sign of R eal( Vh Ih) If the system undergoes an abrupt variation from a permanent harmonic state to a new permanent harmonic state without source (or installation) parameters variation, we can define two recording windows ( a and c ) and a data processing corresponding to the permanent harmonic states. The window b corresponds to the transient between the two permanent states. From (1) and () we have: - Stage a permanent harmonic state before the variation Vh(a). Js Ih(a) Vh(a) = Zc. Jc + Ih(a) = ( ) and ( ) - Stage c permanent harmonic state after the variation Vh(c). Js Ih(c) Vh(c) = Zc. Jc + Ih(c) = ( ) and ( ) In the hypothesis of a variation on the downstream side Vh Vh(a) Vh(c) = = (since Js and do not vary). Ih Ih(a) Ih(c) In the hypothesis of a variation on the upstream side Vh Vh(a) Vh(c) = = Zc (since Jc and Zc do not vary). Ih Ih(a) Ih(c) Once the impedance is known, the harmonic current is determined by (1) and (). 5.3- Non linear regression method This method principle is to be applied during steady state operations. The error minimization passes through a least squares resolution. Equation (5) cannot be solved because of Js and Zc Jc terms that would lead to a nonlinear regression problem. The system can be transformed into a linear regression problem by defining two variables Vs and Vc with Vs = Js and Vc = ZcJc. From the minimization, the matrix resolution cannot be carried out without a minimum number of samples. The algorithm requires several minutes of consecutive samples before solving the system. Once the real and imaginary parts are calculated, the parameters, Zc, Js, Jc are reconstituted. 5.4- Iterative method If we consider there is a small variation at the measurement point on the network side, we can use the iterative method. There are two different stages (N-1 and N) with for each one a parameter calculation using the other parameter previously computed. At the stage N-1, the current Js estimation is carried out using the measurement ( Vh(N 1), Ih(N 1) ). With the impedance previously calculated N 1 = N, the current estimation is: Vh(N 1) Js N 1 = + Ih(N 1) (6) N At the stage N, the impedance estimation is carried out using the measurement ( Vh (N), Ih(N) ). With the current previously calculated Js N = Js N 1, the impedance estimation is: N = Js Vh(N) N 1 Ih(N) (7) EDF_Gonbeau_A1 Session Paper No 3-3 -

C I R E D 17 th International Conference on Electricity Distribution Barcelona, 1-15 May 3 In order to make the system converging, the initial value can be set by = Vh Ih which can be deducted by the method using the sign of eal( Vh Ih) 6- FIRST RESULTS R. 3 1-1 - -3 8 6 4 - -4-6 -8 Voltage measurement (V) Current measurement (A) 6.1- Networks studied In order to apply the different estimation methods, measurement campaigns have been carried out on installations connected to the transmission and distribution networks. Measurements on the transmission system took place in a 15/4 kv substation where significant harmonic levels were regarded (THD 15 %). This situation was due to the presence of capacitor banks in the customer installation that moved the resonance frequency to 5 Hz. Several tests were realized to get information about the application fields of the various methods (capacitor cut-in/cut-off operations, load variations, steady state operation...). Measurement campaigns on the distribution network were representative of industrial customers connected to MV network. 6.- Measurements at the point of common coupling of an installation connected to the transmission network The measurement device was placed at the point of common coupling of a customer connected to 4 kv. The customer load was a 1 MW arc furnace with 9 Mvar of reactive compensation. Ih Vh Scc 15 kv 4 kv Figure 4: Measurement point model The application of the method using the sign of R eal( Vh Ih) gave fairly accurate results. Figure 5 gives an example of the application algorithm when a harmonic variation occurred. Figure 5: Algorithm with three recording windows For the 5 th and 7 th harmonics the estimation was Vh5 Ih5 =,35 14,9i and Vh7 Ih7 =,16 8,9i. Figure 6 compares the estimated impedances on the installation side with those obtained by simulation. Impedance (Ohm) 7 6 5 4 3 1 Harmonic impedance on the installation side 5 15 5 35 45 55 65 75 85 95 Frequency (Hz) Simulation of the harmonic impedance (Ohm) Estimation of harmonic impedance (Ohm) Figure 6: Comparison of impedance obtained by estimation and simulation The shape of the impedance curve is due to the fact that the 9 Mvar capacitor banks were directly connected to the 4 kv. We can notice that the estimated harmonic impedances are very close to those obtained by simulation. 7 6 5 4 3 1 195 19 185 18 175 17 165 16 155 15 :, :, :, :, :4, 1:, 1:, 1:4, :, :4, 1:, 1:, 1:4, :, :, :4, 3:, 3:, 3:4, 4:, :, :4, 3:, 3:, 3:4, 4:, Figure 7: Results of the non-linear regression method Using measurement on Figure 7, the non-linear regression method provided for the 5 th harmonic 5 = 13,6Ω and Zc5 = 13,4Ω. The two values show that there is a parallel EDF_Gonbeau_A1 Session Paper No 3-4 -

C I R E D 17 th International Conference on Electricity Distribution Barcelona, 1-15 May 3 resonance between inductance of the source and the capacitance of the installation. Harmonic voltage measurements and harmonic impedance simulations confirmed this resonance. Impedance (Ohm) 4 35 3 5 15 1 5 14,5 14,4 14,3 14, 14,1 14 13,9 :, :, :4, 1:, 1:, 1:4, :, :, :4, 3:, 3:, 3:4, 4:, 4:, 4:4, 5:, :4,4 :44,4 3:4,4 3:4,4 3:44,4 4:4,4 4:4,4 4:44,4 Harmonic equivalent impedance on source side 5 Current (A) 3 5 15 1 5 6 5 4 3 1 :, :, :4, 1:, 1:, 1:4, :, :, :4, 3:, 3:, 3:4, 4:, 4:, 4:4, 5:, :4,4 :44,4 3:4,4 3:4,4 3:44,4 4:4,4 4:4,4 4:44,4 Harmonic current from source side Js5 Figure 8: Application of the iterative method after a sudden harmonic variation at T=:4,4 The determination of and Js makes it possible to carry out a follow-up of these values by the iterative method. From an initial variation, the graphics on Figure 8 show that it is possible to follow equivalent circuit parameters on the source side ( Js, ) over the time. Once the parameters on the source side are fixed, the iterative method makes it possible to follow their respective value even if a harmonic variation occurs on the installation side. In other terms, the iterative method computes = Vh Ih. 6.3- Measurements at the point of common coupling of an installation connected to the distribution network Two installations connected to the kv distribution network have been instrumented. The first installation (Figure 9) corresponds to a rectifier with an irregular consumption. The second installation (Figure 11) corresponds to an industrial site with a very constant consumption. Js (equations 8 and 9). ( Js + Jc) ( Js Jc) Vh =. Zc. + Zc + (8) Jc = Vh Js (9) Impedance (Ohm) Curent (A) Current (A) 14 11 8 5 35 3 5 4 3 1 5th harmonic impedance on source side 5 5th harmonic current from source side :, :8, :16, :4, :3, :4, :48, :56, 1:4, 1:1, 1:, :, 1:8, :8, 1:36, :16, 1:44, :4, 1:5, :3, :, :4, :8, :48, :16, :56, :4, 1:4, :3, 1:1, 1:, :, 1:8, :8, 1:36, :16, 1:44, :4, 1:5, :3, :, :4, :8, :48, :16, :56, :4, 1:4, :3, 1:1, 1:, 1:8, 1:36, 1:44, 1:5, :, :8, :16, :4, :3, Js5 5th harmonic current from installation side Jc5 (mm:ss:ms) Figure 1: 5th harmonic currents Js Jc and network impedance evolutions This application shows the interest of the iterative method when taking for initial value the estimation of = Vh Ih the method makes it possible to follow, Js and Jc (Figure 1). On the second installation, the harmonic fluctuations allowed an application of the method using the sign of R eal Vh Ih for the equivalent parameters. ( ) Sc c 63 kv Scc=1549 MVA Scc Measurement point MVA kv Measurement point Scc=144 MVA kv kv Industrial loads 59 V PdN=66 kw IdN=88 A U dn=75 V ~ = Domestic loads Figure 11: installation model Figure 9: Installation model On the first installation, the harmonic current and load variations allowed a reliable application of the method using the sign of R eal( Vh Ih) for and Js estimation. In this particular case with the assumption of Zc >>, it is possible to calculate Jc according to the values of and During this measurement described on the graphs of Figure 1, the 5 th harmonic variations allowed to find 5 = 16,6Ω. Moreover, other 5 th harmonic variations leaded to Zc5 = 3,6Ω. EDF_Gonbeau_A1 Session Paper No 3-5 -

C I R E D 17 th International Conference on Electricity Distribution Barcelona, 1-15 May 3 1 1 8 6 4 8 7 6 5 4 3 1 installation connected to the distribution grid revealed this limit). Thus, this application method on installations with fluctuating harmonic levels is to be proscribed. But, it is interesting to notice that when source and installation side impedances are close to each other the method is reliable. Thus this method will have to be correlated with other results (statistical or with results of the preceding methods). :, :1, :4, :36, :48, 1:, 1:1, 1:4, 1:36, 1:48, :, :1, :4, :36, :48, 3:, 3:1, Vh5 Ih5 3:4, 3:36, 3:48, 4:, 4:1, 4:4, 4:36, 4:48, 5:, Figure 1: 5th harmonic distortion and power measurements Using measurement on Figure 1, the non-linear regression method gives: Zc5 = 4Ω and Jc5 = 6,54A 5 = Ω and Js5 =,11A The results of both estimation methods are in line with those obtained by simulation. Moreover we can note that the current flowing from the installation to the network is larger than the current originating on the source side. 7- PROSPECTS The general objective of this study is to obtain a "standardized" method, which could be used within a contractual framework with customers. To be able to determine the exact customer contribution in all cases, EDF studied new estimation methods of the customer contribution (non-linear regression and iterative methods). In addition, new algorithms for application of these measurement methods, as well as existing methods (method using the sign of R eal( Vh Ih) ), were developed. The definition of algorithms and two dedicated measurement campaigns made it possible to cross a stage. The results of the described methods are encouraging. 1) The method using the sign of R eal( Vh Ih) allowed coherent harmonic impedances estimations when significant harmonic variations occur. Without significant harmonic voltage and current variation the method does not have any utility. We have defined thresholds in terms of harmonic current ( Ih ) and voltage ( Vh ) variations which lead to reliable estimations. These thresholds are around Ih >% Ih in the transmission network and Vh >6%Vh in the distribution network. ) The iterative method enabled interest to follow on line equivalent parameters (impedance and current injection). The results obtained were very conclusive but this method requires a very precise initial value. In this direction the iterative method appears to be a very good complement with the method using the sign of R eal( Vh Ih). 3) The non-linear method did not work if harmonic variations occur (the harmonic fluctuation on the first Lastly, the estimate methods are very sensitive to harmonic phase angle and a procedure should also be specified to reduce measurement errors. 7- CONCLUSION Harmonic current flowing at an installation s metering point results from harmonics originating in this installation and also in other installation connected to the power supply system. This paper presents assessment techniques and algorithms to realize on line harmonic emission evaluation. The basic idea is to define an approach scrutinizing sudden harmonic variations at the point of common coupling in order to apply the method using the sign of R eal( Vh Ih). Once the impedance and current estimation is done the iterative method makes it possible to follow-up these values over the time. To complete the assessment for periods with no significant harmonic variation, a non-linear regression method is described. These methods were tested with encouraging results. We can reasonably think that a method will see the light of day in the medium term allowing to determine the contribution of the installation and the power system to harmonic disturbance with sufficient precision within a contractual framework application. We should be able to give inform about our progress before the next CIRED. REFERENCES [1] H Yang, P. Pirotte, A. Robert, 1994, Assessing the harmonic emission level from one particular customer, Proceedings of PQA 94. [] 1996, Electromagnetic Compatibility, Assessment of emission limits for disturbing load in MV and HV power systems IEC 61-3-6. [3], Electromagnetic Compatibility, General guide on harmonic and interharmonics measurements and instrumentation, for power supply systems and equipment connected thereto IEC 61-4-7 Ed.. EDF_Gonbeau_A1 Session Paper No 3-6 -