Smart Busbar Protection Based ANFIS Technique for Substations and Power Plants

Size: px

Start display at page:

Download "Smart Busbar Protection Based ANFIS Technique for Substations and Power Plants"

David Lee
5 years ago
Views:

1 Smart Busbar Protection Based ANFIS Technique for Substations and Power Plants 1 Mohamed A. Ali, 2 Sayed A. Ward, 3 Mohamed S. Elkhalafy 123 Faculty of Engineering Shoubra, Benha University 1 mohamed.mohamed02@feng.bu.edu.eg, 2 drsayedward@yahoo.com, 3 eng_ms4ever@yahoo.com Abstract In this paper, an Artificial Intelligent approach using ANFIS for busbar differential protection for reliable fault detection with enhanced properties. The properties are phase comparison with current directional, dynamic differential characteristic with low/high current mode, and saturation detection algorithm based on AI technique. A comprehensive ANFIS structure denoted by Smart Busbar Protection Algorithm- is created with the three properties described above to distinguish between actual internal fault in a sub-half cycle and the external faults to send a blocking signal to avoid mal-operation of the busbar protection system. The smart busbar protection algorithm is practically tested on a typical 220kV busbar system in Egypt with secondary injection tester on a busbar protection system as exists on the substation. Then the actual testing results are compared with the simulated smart busbar protection system. The results show the validity and effectiveness of the proposed smart algorithm with very fast action making the proposed algorithm suitable for real-time protection. Index Terms Adaptive Busbar Protection, ANFIS, Busbar Protection, Directional Detection, Saturation Detection, Fault Detector. I. INTRODUCTION Differential protection schemes are applied for high voltage busbars. Failure to- trip on an internal fault, as well as false tripping of a busbar during a load service or in case of external fault or heavy inrush currents, both have disastrous effect on the stability of power systems [11]. The challenge of bus differential protection is the issue of false differential current due to CT saturation and ratio mismatch [1-4, 8-11]. The bus bar protection can be classified as high impedance and low impedance types. High impedance relays are used to provide low cost bus protection, but have limitations due to complex arrangements and use of multi ratio current transformers. The low impedance measuring principle employs the zone selective differential current as the operating quantity and the sum of the current magnitudes as the stabilizing signal [8-10]. The measuring principle must ensure protection with CT saturation on external faults. A low impedance busbar protection operates during CT saturation by using a principle, which discriminates between saturated and unsaturated wave forms. Recently, many novel differential techniques have been proposed to overcome CT saturation that will be discussed below. For external faults, the differential current should be zero, but errors caused by CT saturation can result in a non-zero value. To prevent mal operation, the operating threshold is raised by increasing the bias setting. Raising the bias threshold has detrimental effect on the relay sensitivity as it prevents the detection of in-zone resistive faults. The impact of CT ratio bias characteristics reduces the sensitivity of the relay. New digital relaying technique for busbar protection using phase angle change in sequence current of incoming CTs and outgoing CTs which based on directional detection is produced [3]. The paper [2] presents a new approach for busbar protection with stable operation of current transformer during saturation, using neuro- fuzzy and symmetrical components theory. The authors' technique uses symmetrical components of current signals to learn the hidden relationship. Also, an analysis of the performance during CT saturation conditions is discussed. Reference [5] proposes a busbar protection scheme based on phase changes in positive sequence current of incoming and outgoing line current transformers (CTs). The angle differences of during fault and pre-fault current signals of incoming and outgoing CTs are the indicators of external or internal faults for busbar protection. The advantage of the method is that it does not use magnitude information of the current and thus overcomes the CT saturation issues. Whereas, in reference [6] it has been shown that by looking into the property of only three quantities it is possible to make a numerical busbar differential protection relay which can: Operate quickly for internal faults due to the fact that input 2331

current becomes much larger than output current during an internal fault (i.e. current flows into the faulty bus but it doesn t flow out from the faulty bus).

2 current becomes much larger than output current during an internal fault (i.e. current flows into the faulty bus but it doesn t flow out from the faulty bus). Remain stable for external faults followed by CT saturation due to a fact that for short period of time, immediately after the fault current zero crossing, i.e. input current will be equal to output current (i.e. before any CT goes into saturation). Detect an open CT secondary circuit in any one of the connected feeder CTs. fault sensitivity. I, I r, and I d are a minimum measuring capability of the embedded PMU into the relay which may be denoted the measuring cut-off value. Reference [7] discussed a novel busbar protection scheme based on arificial intelligent technique. The authors employe wavelet multi -resolution signal decomposition for fast busbar protection system. This research paper produces an Artificial Intelligent technique using ANFIS for busbar differential protection for reliable fault detection with enhanced properties. The properties are phase comparison with current directional, dynamic differential characteristic with low/high current mode, and saturation detection algorithm based on AI technique. A comprehensive ANFIS structure denoted by Smart Busbar Protection Algorithm- is created with the three properties described above to distinguish between actual internal fault in a sub-half cycle and the external faults to send a blocking signal to avoid mal-operation of the busbar protection system. II. PROPOSED ALGORITHM The proposed busbar protection algorithm depends on many sub-functions that detect current transformer saturation due to high external faults, fault location i.e. internal or external, distinguishing between normal heavy loading currents and high fault currents via the Intelligent Fault Detector (IFD). All the previous sub-functions are implemented into ANFIS structure for modern busbar protection system to achieve a sub-cycle tripping actions for more reliable busbar protection system. The following sections present the different sub-functions employed in this algorithm. III. INTELLIGENT FAULT DETECTOR The Intelligent Fault Detector (IFD) element is a sensitive current fault detector that detects any fault on the protected system. So that, the IFD is able to distinguish between normal heavy loading currents and high fault currents. The IFD is intended for use in conjunction with Phasor Measurement Unit (PMU), blocking of current based elements to prevent mal-operation as a result of the wrong settings. The proposed IFD responds to the changes in magnitude of the sequence currents and the rate of change of the differential and restraining currents as shown in Figrue.1. Thus, for more security, any current-based function like percentage differential should not be declared unless the IFD is activated that reflects a real fault condition. The values of K, K 1, and K 2 are a sensitivity factors that determines the required degree of Figure 1: Intelligent Fault Detector (IFD) algorithm IV. INTELLIGENT SATURATION DETECTION ALGORITHM External faults near substation typically result in very large time constants of DC components in the fault currents. Also, when energizing a step-up transformer, the inrush current being limited only by the machine impedance may be significant and may last for a very long time. In order to provide additional security against mal-operations during these events, the busbar protection should incorporate saturation detection algorithm. When saturation is detected the busbar protection system will make an additional check on the angle between the substation currents into the protected zone (Directional Detection). If this angle indicates an internal fault then tripping is permitted. Figure.2 depicts the differential and restraining current trajectory due to internal faults and external faults which lead to CT saturation. Firstly, in case of external faults: differential current stays very low during the initial period of linear CT operation while the rate of change of restraining current increase quickly as marked in red at Figure.2. This rapid increase in restraining current will lead to CT saturation which in turn leads to false differential current. Thus, this case will require an additional security function to guarantee reliable stable operation for the busbar protection system. Whereas, for internal faults: the rate of change of both differential and restraining current increase simultaneously as marked in blue at Figure.2. CT saturation condition is declared by the Saturation Detector sub-function when the magnitude of the restraining current becomes larger than the higher breakpoint and at the same time the differential current is below the first slope. 2332

3 Case No. Table 1: Illustration for the proposed CT saturation detector d d d Id Ir Description Ir Id Id dt dt dt > d dt Ir 1 +H +H 1 H H Internal Fault 2 +L +H 0 L H External Fault Figure 2: Trajectory of differential & restraining currents due to saturation and internal fault cases Figure.3 depicts the saturation detection algorithm during the operation of the busbar protection system. When Saturation (SAT) detector declares i.e. SAT=1, the sub-function checks if the differential current is below the first slope with low rate of change with respect to time, also a higher rate of change for restraining is monitored leads to differential-restraining trajectory out of the differential characteristics for certain period of time. Figure 3: Flowchart of saturation detection algorithm as per B90 IED 3.1 +L +H 0 L H CT saturation due to external 3.2 +L -H 1 H L fault 4 L L 1 L L Internal Fault Table.1 illustrates the proposed saturation detection algorithm using ANFIS technique. Case No.01: states that during internal faults, the rate of change of differential current and restraining current will be highly increasing (+H). Also, the rate of change of differential current is higher than the rate of change of restraining current. While, the differential and restraining currents are high. Case No.02: During external faults the rate of change of differential current will be increased Low (+L) with simultaneous increase in the rate of change of restraining current (+H). Also, the analogue comparator of the rate of change declares "0". Case No.03: Also, during external faults; same as Case No.02 but with extended event shown in case No. 3.2 as described via Figure.2 Case No.04: Internal fault with low current mode as described in row four of Table.1. As shown in Table.1, the ANFIS structure of the saturation detector is comprises of five inputs and one output as depicted in Figure.4. This ANFIS structure will be merged with the comprehensive ANFIS for total busbar protection system. The proposed ANFIS structure is employed for saturation prediction because it depends on measuring the rate of change quantities as shown in Table.1, whereas, the conventional algorithm in protection relays is for saturation detection. i.e. the proposed ANFIS structure is adequate for real-time busbar protection as it is very fast and based on prediction not detection. 2333

Figure 6: Illustration for low/high current mode differential operation Figure 4: ANFIS structure for saturation detector After the above analysis, the busbar protection system requires and

The additional security function must be directional-based to guarantee the operation during internal fault and blocking during external faults as shown in figure.5.

For high differential current mode, the directional principle is included only if demanded by the saturation detector (dynamic 1-outof- 2 / 2-out-of-2 mode).

The dynamic inclusion/exclusion of the directional principle is not applied for the low differential currents but is included permanently only because it is comparatively difficult to reliably detect

4 Figure 6: Illustration for low/high current mode differential operation Figure 4: ANFIS structure for saturation detector After the above analysis, the busbar protection system requires and additional security beside the saturation detector algorithm. The additional security function must be directional-based to guarantee the operation during internal fault and blocking during external faults as shown in figure.5. For low differential current mode, the biased differential element operates on the 2-out-of-2 basis utilizing both the differential and directional principles as shown in Figure.6. For high differential current mode, the directional principle is included only if demanded by the saturation detector (dynamic 1-outof- 2 / 2-out-of-2 mode). Typically, the directional principle is slower, and by avoiding using it when possible. The dynamic inclusion/exclusion of the directional principle is not applied for the low differential currents but is included permanently only because it is comparatively difficult to reliably detect CT saturation occurring when the currents are small, i.e. saturation due to extremely long time constant of the DC component or due to multiple auto-reclose actions. V. DIRECTIONAL COMPARISON ALGORITHM For more reliable and secure tripping actions due to internal faults and blocking due to external faults, the directional comparison algorithm is highly needed. Figure 7: Current Phasor of external faults External Fault Case: One current flow in an opposite direction compared with the sum of the remaining currents as shown in Figure.7. Internal Fault Case: All fault currents flow in one direction i.e. in-phase currents for all bays as shown in Figure.8. Figure 5: Saturation and directional detection algorithm 2334

5 Figure 8: Current Phasor of internal faults Practical Implementation Steps: Step 1: select fault contributors A contributor is a circuit carrying significant amount of current. A circuit is a contributor if its current is above higher break point and is also above a certain portion of the restraining current. Step 2: check angle between each contributor and the sum of all the other currents. Step 3: compare the maximum angle to the threshold Which is 90 degrees as depicted in Figure.6 and Figure.7. Below are a simple IF-THEN rules implemented in the AI technique to check the directionality and the faulted bay in the system. Where: INT_FLT : Internal Fault flag n: Total number of bays I C: Contributor bay current I S-I C: is the summation of all bay currents in the protected zone. For J=1:n; //n=number of bays IF I Cj > B1 or I Cj > 0.1 pu IF abs (ang(i Cj)-ang(I S-I C)) < 90 INT_FLT=1; Disp ("Fault at Bay No. ", %J) ELSE INT_FLT=0; // i.e. external fault END END END VI. PRACTICAL TESTING RESULTS Typical substation configuration in Egypt is chosen for testing the presented algorithm in this paper. The substation is 220/66/11kV, the substation is divided into two main sections each one contained double buses with one circuit breaker. Each bay is connected via two disconnectors to represent the substation bus image. Figure.10 depicts the substation configuration under testing to validate the proposed algorithm. The busbar protection system is GE MULTILIN B90 comprises of eight IEDs. IED 1, 2, and 3 are designated for protection of phase A, B, and C respectively of Bus 1 and Bus 2. IED 4, 5, and 6 are designated for protection of phase A, B, and C respectively of Bus 3 and Bus 4.as the system is considered phase segregated scheme. IED 7 is for collecting all isolator status for dynamic bus replica. IED 8 is for breakers failure for the whole substation. Finally, IED 9 is for end fault protection for faults at dead zones to trip the remote substation in case of outside fault. Figure.11 depicts the busbar protection system at the substation under test. Figure 10: Practical test system at typical 220/66/11kV substation in Egypt Figure 9: Proposed fault detection algorithm Figure.11: Busbar protection system at the substation under test 2335

mal-operation and tripping decision by the busbar protection system.

6 As shown in Table.2, the ANFIS structure consists of four inputs and one output. The tripping action for the busbar protection system depends on the declared flags by each sub-functions which are Dif -L, Dif-H, DIR, and SAT. mal-operation and tripping decision by the busbar protection system. Table 2: Input/output of the ANFIS structure Dif L DIR SAT Dif H Expected Action Actual Action Description BLOCK TRIP BLOCK BLOCK BLOCK TRIP BLOCK TRIP BLOCK Figure.12: ANFIS structure of the proposed busbar protection TRIP BLOCK BLOCK TRIP TRIP TRIP TRIP Figure 13: Actual external fault with one CT is saturated Figure.13 shows an actual external fault happened near to the substation under testing. Thus, the busbar protection relays recorded the waveform capture of the currents in the fault recorder. This wave form is then retrieved from the fault recorder via COMTRADE files format - according to IEEE to be used for practical secondary injection testing using "Fault Recurrence" option in the used test set which is Kingsine K1066i. The retrieved COMTRADE file is used for secondary injection for Bay 1 and Bay 2 shown in Figure.10. The current of Bay 1 is F1 which is marked in blue in Figure.13. The current of Bay 2 is F2 which is marked in red. The current F2 is heavily saturated due to very close external fault current. The saturated CT associated to F2 will make a false differential current which in turn will lead to Figure 14: Actual response of the busbar protection system due to practical testing of the proposed algorithm 2336

7 Dif L Table 3: Practical testing results due to secondary injection testing DIR SAT Dif H Description Actual Measured Tripping Time By K1066i By Conventional Protection Relay By Smart Protection Algorithm BLOCK N/A N/A TRIP 29 ms 9 ms BLOCK N/A N/A BLOCK N/A N/A BLOCK N/A N/A TRIP 23 ms 10 ms BLOCK N/A N/A TRIP 28 ms 11 ms BLOCK N/A N/A TRIP 22 ms 10 ms BLOCK N/A N/A BLOCK N/A N/A TRIP 25 ms 11 ms flag. i.e. the SAT flag was very fast enough to declare firstly to be used for blocking. Finally, the total system response is less than half-cycle which means it is suitable for real-time protection with reliable operation and substation protection. Table.3 depicts the busbar protection system response upon employing the proposed smart protection algorithm presented in this paper. The actual tripping time for each case is practically measured by using Kingsine model: K1066i test set. The tabulated results show that the response of the smart protection algorithm is a sub-cycle with average approximately half-cycle whereas, the conventional busbar protection relay response time was measured to be more than one cycle and approaches in some cases one-and-half cycle. VII. CONCLUSION In this paper, an Artificial Intelligent approach using ANFIS for busbar differential protection for reliable fault detection with enhanced properties. The properties are phase comparison with current directional, dynamic differential characteristic with low/high current mode, and saturation detection algorithm based on AI technique. A comprehensive ANFIS structure denoted by Smart Busbar Protection Algorithm- is created with the three properties described above to distinguish between actual internal fault in a sub-half cycle and the external faults to send a blocking signal to avoid mal-operation of the busbar protection system. The smart busbar protection algorithm is practically tested on a typical 220kV busbar system in Egypt with secondary injection tester type Kingsine K1066i on a busbar protection system type GE MULTILIN B90 IEDs. Then the actual testing results are compared with the simulated smart busbar protection system. The results show the validity and effectiveness of the proposed smart algorithm with very fast action making the proposed algorithm suitable for real-time protection TRIP 27 ms 9 ms TRIP 29 ms 8 ms TRIP 24 ms 12 ms Same COMTRADE file is used for simulated injection for simulated busbar protection algorithm using the proposed modern busbar protection algorithm. Simulation is done using MatLAB M-files and the AI tool box for the ANFIS to check the validity and effectiveness of the proposed algorithm. Figure.14 depicts the response of the smart busbar protection system due to heavy saturation for one during external fault. The smart busbar protection system is blocked from tripping due to saturation detection via declaring the SAT flag. Differential pickup is declared but after declaring the SAT VIII. REFERNCES 1. L.G.Hewitson, Markbrown, Ramesh and Balakrishman, Practical power system protection, M.R Aghaebrahimi, H.Khorashadizadeh (Department of power engineering,university of Birj, Fuzzy neuro approach to busbar protection design and implementation, International Journal Of Information and Communication Engineering, 2:5, Bodankasztenny, Gustawo Brunello and Lumomin Sevov, Digital low impedance bus differential protection with reduced requirements for C.T, GE Power Mangement CANADA Transmission and Distribution Conference and Exposition 2001, IEEE/PES Volume 2, Issue, 2001, Pages: VOL

8 4. J.G Andrichak and Jorge Cardenas, Bus differential protection, Western Protective Relay Conference Spoken, Washington, October 24, P.Jena and A.Kpradham, Busbar protection a solution to C.T saturation, Fifteenth National Power System Conference (NPSC,ITT) Bombay Dec Z. GAJIĆ, Design principles of high performance numerical busbar differential protection, CIGRE, Relay Protection and Substation Automation of Modern Power Systems (Cheboksary, September 9-13, 2007). 7. Shiyong Wang, Xinzhou Dong, and Shenxing Shi, A novel bus bar protection scheme based wavelet multi resolution signal decomposition. Dept Of Electrical Engineering Tsinghua University, Beihing China. 8. Y.C.kang, J.S.Yun, B.E.Lee, S.H.Lee, S.H.Kang, S.I.Jang and Y.G.Kim, Bus bar differential protection in conjunction with a current transformer compensating algorithm, IET Gener, Transm. Distrib.Vol.2, No, 1, Jan A.M.Dmitrienko and A.Y.U.Sinichkin, Fast differential bus bar protection based on REB670, Russian Electrical Engineering, Dec S.R.Samantaray, L.N.Tripathy, P.K.Dash, G.Panda, S-transform based directional bus bar protection, National Institute Of Technology Rourkela, India and Siliscon Instituted of Technology Bhubanes war, India. 11. Mohamed A. Ali, Ahmed F. Bendary, " Modern Philosophies of Inrush Current Detection Algorithm and Their Impact on Transformer Protection", International Electrical Engineering Journal (IEEJ), Vol 06 (2015), No.12, pp

Bus Protection Fundamentals

Bus Protection Fundamentals Terrence Smith GE Grid Solutions 2017 Texas A&M Protective Relay Conference Bus Protection Requirements High bus fault currents due to large number of circuits connected: CT