Venturi, Alberto and Dolan, Michael James and Clarkson, Paul and Wright, Paul and Forbes, Alistair and Yang, Xin-She and Roscoe, Andrew and Ault, Graham and Burt, Graeme (2013) Evaluating the robustness of an active network management function in an operational environment. In: EU EURAMET EMRP Metrology for Smart Grids Workshop, 25-26 June 2013, 2013-07-25-2013-07-26., This version is available at https://strathprints.strath.ac.uk/44717/ Strathprints is designed to allow users to access the research output of the University of Strathclyde. Unless otherwise explicitly stated on the manuscript, Copyright and Moral Rights for the papers on this site are retained by the individual authors and/or other copyright owners. Please check the manuscript for details of any other licences that may have been applied. You may not engage in further distribution of the material for any profitmaking activities or any commercial gain. You may freely distribute both the url (https://strathprints.strath.ac.uk/) and the content of this paper for research or private study, educational, or not-for-profit purposes without prior permission or charge. Any correspondence concerning this service should be sent to the Strathprints administrator: strathprints@strath.ac.uk The Strathprints institutional repository (https://strathprints.strath.ac.uk) is a digital archive of University of Strathclyde research outputs. It has been developed to disseminate open access research outputs, expose data about those outputs, and enable the management and persistent access to Strathclyde's intellectual output.
Evaluating the robustness of an active network management function in an operational environment Alberto Venturi, Michael J. Dolan, Paul Clarkson, Paul Wright, Alistair Forbes, Xin-She Yang, Andrew J. Roscoe, Graham W.Ault, Graeme M. Burt Noordwijk, 26 th of June 2013
Introduction Incentives to renewable sources of energy are causing an increase of the number of generators connected to distribution networks. The cost of the network reinforcement are very high and utilities are interested in active network management solutions in order to manage the generator connection to the net, reducing to the minimum the network reinforcement. So, automatic control systems, based on software tools, are becoming more desirable in distribution power systems. Primarily, such schemes are expected to manage system voltage fluctuations, network power flows and fault levels. Functionalities include also power balancing, system frequency control and management of demand side resources for the primary system constraints.
Introduction A critical concern is the robustness of online and automatic active network management (ANM) algorithms/schemes. The ANM scheme s functionality depends on convergence to a solution when faced with uncertainty and its reliability can be reduced by data skew and errors. The work presented evaluates power flow management (PFM) functionality based on the Constraint Satisfaction Problem (CSP) in an operational environment. The objective is to assess performances when subjected to different levels of data uncertainty and verify the introduction of a state estimator (SE) in the ANM architecture to mitigate the data uncertainty effects on the control action.
Test Environment 15kW 4-wire back-back inverter jx jx 40kW, 50kVA Controllable loadbank R R LPC-controlled microgrid #1 LPC-controlled microgrid #2 6 x 3kW singlephase inverters Windy Boys 2kVA Synchronous generator 10kW, 12.5kVA Controllable loadbank 10kVA Inverter 10kW, 12.5kVA Controllable loadbank 4 regenerative MG sets
11kV supply R1.34J Test Environment Remote trip From control room, and control panel red lights from red phase Link box lines 3,4 11/0.4kV 500kVA MCCB 600A LV board 1 N/C 200A E-Stop Contactors 150A in Parallel 80A variac E-Stop Contactors 225A Link box lines 1,2 E-Stop Contactors 150A in Parallel 116kVA load max (167A, 241A DOL start) 146kVA gen max (210A) Tx line / primary substation impedance simulator. 0, or 2Ω to 43Ω in 9 discrete steps. May need to be shorted for some MGI experiments Where all loads and no generation, or vice versa, are engaged, due to current capacity. Rated 500A for 5 seconds. Continuous rating??x System neutral is earthed through this path, even when all fuseswitches and contactors are open Neutral is common throughout the entire system at present (no fuses or contactors/breakers in the path). DG3 F F 250A 250A DC Bridge 1/1 isolating Sync Contactor (key) 80kVA output Neutral exists but does not connect to transformer DG3 E E DG3 Switchboard GSP V ~ DC MOTOR 80kVA M-G SET DG1 to DG3 Optional Line Impedance type C DC GEN IM DG1 Switchbox type C RCD D20A type C 200A D20A 13A sockets from red phase DG1 to DG2 16A 3Ø Sockets Switch 200A LV board 2 D 32A 3Ø Sockets Small set power C D Siemens Motor Drives - Induction load/regen bus B10A 20kW 1kV DC supply Optional Line Impedance 0.34 Ω, 100W Resistors in series with 0.6mH Inductor (24A Max) D40A 09 th March 2012 Windy Boys Good ground via LV board 1 and earth strap D20A Spare DG2 to DG1 C16A type C Spare DG2 Switchbox Other:- DG2 to EastWall type C RCD type C Generators:- 80kVA (Big Gen) = 115A 3kVA (Small Gen) = 5A 23kVA (PV inverters) = 35A 10kVA (Inverter) = 15A 3+7+10+10=30kVA (Induction MG sets in gen mode) = 44A All gen = 80 + 3 + 23 + 10 +30 = 146kVA = 210A All gen apart from 80kVA = 66kVA = 96A Loads:- 50kVA Loadbank = 72A 2.2kW (3kVA) Induction motor = 6A (16A DOL start) 5.5kW (7kVA) induction motor = 12A (32A DOL start) 2 * 7.5kW (10kVA) Induction motor = 16A (35A DOL start) All induction motors = 23kW => 30kVA, 50A, (118A DOL start) 2 * 12.5kVA Loadbanks = 12.5kVA, 18A Single phase loads = 10.5kW => 10.5kVA, 15A. All loads = 50+30+12.5+12.5+10.5 = 115.5kVA, 166A (241A DOL start) All loads apart from 50kVA loadbank = 66kVA = 95A (169A DOL start) ring connection on East wall (outside control room) Load bank 40kW 27kVAR 50kVA D C B A DG3 D DG3 C DG3 B DG3 A 0.5 Ω, 300W Resistors in series with 0.4mH Inductor (24A Max) B16A 66kVA load max (95A) (169A DOL start) 36kVA gen max (52A) 66kVA gen max (96A) if infuction MG sets used also ring connection on North wall DG1 E E Tx line / transformer impedance 0.76mH per phase, 0.24 Ω @ 50Hz 6% pu @ 40kVA (58A) C10A 0.34 Ω, 100W Resistors switchable in series with Tx line / transformer impedance 33kVA load max (48A) (81A DOL start) 26kVA gen max (38A) 36kVA gen max (52A) if induction MG sets used also DG1 switchboard DG1 V DG2 E E Tx line / transformer impedance 0.76mH per phase, 0.24 Ω @ 50Hz 6% pu @ 40kVA (58A) C10A 0.34 Ω, 100W Resistors switchable in series with Tx line / transformer impedance 32.5kVA load max (47A) (88A DOL start) 10kVA gen max (15A) 30kVA gen max (44A) if induction MG sets used also DG2 switchboard DG2 V ~ 80kVA (60kW) Synchronous generator 10kVA load max (15A) (48A DOL start) 32A A DG1 A 40A B DG1 B 6A C DG1 C 40A D DG1 D 31kVA load max (45A) 20kVA load max (32A) (70A DOL start) 40A A DG2 A 10A B DG2 B 20A C DG2 C 20A D DG2 D Key to instrumentation symbols GSP E GSP V GSP C 3Ø CTs - 100:1A 3Ø VTs - 400/110V 3Ø VTs and CTs 40A DG1 RCD Load bank 10kW 20A 7kVAR 10A 16A 16AR 16AR 16AY 16AY 16AB 16AB 12.5kVA 20A 20A IM IG Siemens drive 2.2kW 3kVA Dynamic load (induction) IM IG Siemens drive 5.5kW 7kVA Dynamic load (induction) PV PV PV PV PV PV DC DC DC DC DC DC 6x3kW PV inverters, 224-600VDC in, Single or Three-phase out? 18kW @ 80%efficiency =23kW (33A) No.6 Currently Fronius To PV Bus (40A) 15kW Triphase Inverter 32A 3Ø +N+E Plug 22.7kW @ 80% efficiency = 29kW load max (42A) 30kVA gen max (44A) ~ Small Set 2kVA synchronous (with controlled rectifier& DC motor drive from LV board 2) Break out to 13A single phase ring mains 3 kettles 3 microwaves 10.5kVA load max (15A) To Induction load/regen Bus (40A) IM IG Siemens drive 7.5kW (A) 10kVA Dynamic load (induction) IM IG Siemens drive 7.5kW (B) 10kVA Dynamic load (induction) Spare connection 10kVA Inverter 20kW 1kV DC supply (powered from LV board 2, 3Ø) DG2 Load bank 10kW 7kVAR 12.5kVA
Test Environment Resistance, mohm/m 1.75 1.5 1.25 1 0.75 0.5 0.25 0 0 50 100 150 200 250 300 350 Conductor diameter, mm^2 Voltage drop, V 40 35 30 25 20 15 10 5 0 0 48 70 90 Load power factor angle Busbar 2 Busbar 3 Busbar 4 Busbar 5 The impedance of the cables has been estimated using values available in literature. The effect of the addition of extra impedances has been evaluated and new impedances have been added to the branches of the grid. Reactance, mohm/m 0.155 0.15 0.145 0.14 0.135 0.13 0.125 0.12 0 50 100 150 200 250 300 350 Conductor diameter, mm^2
Test Environment Part of the microgrid available at Strathclyde University has been configured to allow the integration and testing process of an ANM function. Busbar 1 includes a variable load bank. Load banks are also connected to two other busbars (4 and 5). Induction machines, which can also act as generators, are connected to buses 4, 5, 6 and 7. These units have a maximum real power output of 2.2kW, 5.5kW, 7.5kW and 7.5 kw.
PFM using a CSP approach Modelling the PFM problem as a constraint satisfaction problem entails expressing the problem as a set of variables with finite discrete domains and a set of constraints. For PFM, the problem to be solved is concerned with deciding what control actions to take, on the Distributed Generation (DG) units, in order to maintain the network within the thermal limits (i.e. power constraints) and maximize DG access. The variables of the CSP are the controllable generating plant power output setpoints and the domains are the discrete values that the generators set-points can assume. V:={gen 1, gen 2,,gen n } D gen1 :={control 1,, control n }
PFM using a CSP approach These values are the maximum values that a generator can output. However, the intermittent nature of most renewable generating plants means that DG output is such that its output is continuous up to this discrete set-point value. In addition to variables and their domains we have to set the constraints on the solution: Power flow constraints: no thermal overloads Contractual constraints: generators access rights Preference constraints.
PFM using a CSP approach (V gens, D Control Signal, C) (1) Where: V gens = {Gen 1, Gen 2... Gen n } (2) D Control Signal is: D Gen1 ={1,,0}, D Gen2 ={1,,0}, D Genn ={v 1,,v n } (3) C is the constraint applied to the sets of variables: max C Power Flow = { S S } ij (4) C Contractual = {k, l, m} (5) C MaxDG max N n 1 P (6) Gi ij
PFM using a CSP approach Modelling PFM, in this way, relies upon a load flow engine to evaluate the power flows within the network to determine any control actions that are required. The generators have been ordered in a last-in, first off (LIFO) manner to replicate the current connection regime used in the UK. Constraints Contractual Load Flow Preference Variables & DG Domains Output Loads CSP Solver Network Model Load Flow Engine PFM Algorithm N Ranked Control Solution M. J. Dolan, E.M.Davidson., G. W. Ault, K.R.W. Bell, S. D. J. McArthur, Distribution Power Flow Management Utilizing an Online Constraint Programming Method.Smart Grids, IEEE Transaction on.
Integration of the ANM and the Microgrid The PFM algorithm, presented above, was chosen as the ANM control approach and installed initially on a COM600 (Windows XP embedded industrial PC).
Integration of the ANM and the Microgrid The PFM algorithm, presented above, was chosen as the ANM control approach and installed initially on a COM600 (Windows XP embedded industrial PC). The communication system among the PCs of the Microgrid created some problems and different solutions has been tested. The solution chosen at the end is based on the OPC server/client architecture, with the integration of the OpenOPC functionalities.
Integration of the ANM and the Microgrid The PFM algorithm, presented above, was chosen as the ANM control approach and installed initially on a COM600 (Windows XP embedded industrial PC). The communication system among the PCs of the Microgrid created some problems and different solutions has been tested. The solution chosen at the end is based on the OPC server/client architecture, with the integration of the OpenOPC functionalities. In order to send the data to the RTS analogue inputs, a Beckhoff CX5010 Embedded PC with Intel Atom processor has been used. The Beckhoff software presents OPC functionalities and the data can be sent using the OPC standard. The Beckhoff unit can be set and controlled also via a standard pc.
Integration of the ANM and the Microgrid The sensor measurements coming from the microgrid are collected via a real time station (RTS) computer developed by ADI. The RTS has analogue and digital input/output (I/O) interfaces, can execute programs written with Matlab and Simulink, process directly the data collected, manage the electrical machines of the grid and guarantee their safe operation.
Integration of the ANM and the Microgrid The sensor measurements coming from the microgrid are collected via a real time station (RTS) computer developed by ADI. The RTS has analogue and digital input/output (I/O) interfaces, can execute programs written with Matlab and Simulink, process directly the data collected, manage the electrical machines of the grid and guarantee their safe operation. After a first elaboration, the measurements are then made available, mapped on the OPC server variables and sent from the PC, connected to the RTS, to the OPC server.
Integration of the ANM and the Microgrid The sensor measurements coming from the microgrid are collected via a real time station (RTS) computer developed by ADI. The RTS has analogue and digital input/output (I/O) interfaces, can execute programs written with Matlab and Simulink, process directly the data collected, manage the electrical machines of the grid and guarantee their safe operation. After a first elaboration, the measurements are then made available, mapped on the OPC server variables and sent from the PC, connected to the RTS, to the OPC server. The control software reads them, through the OPC server, and sends the control signals back through the RTS.
Experimental work and results The sensors installed on the microgrid and the data acquisition system guarantee a precision of 2.14% in the measurement of voltage and current magnitude, and consequently a precision of 4.5% in measuring the power flow. This level of precision is considered in literature enough to simulate the real operating conditions of an energy management system on a low voltage network. In order to show the effect of the uncertainty of the data on the performance of the ANM software a series of tests were executed on the microgrid. The induction machines located at buses 4, 5, 6 and 7 were set to compensate the load requested on the buses 4 and 5. Thermal constraints were set in the branches 1 and 3 by reducing the limits within the PFM software and microgrid network model.
Experimental work and results The generator access priorities were assigned to represent a LIFO connection arrangement. Gen 1 was set to 1 Gen 3 was set to 2 Gen 2 was set to 3 Gen 4 was set to 4 (i.e. this unit would be the first to be curtailed if a thermal breach was detected). Then, progressively, the loads were reduced to zero starting with the load on busbar 5. This caused a rising power flow through the branches 1 and 3, and a consequent thermal constraint violation.
Experimental work and results The response of the ANM function, for this scenario, was evaluated against the following data sets: An initial clean set of input data without any uncertainty A set of data as collected from the grid (with an uncertainty of 4.5%) A set of data in which the uncertainty of the loads and the machines power flow was artificially increased to 6% A set of data as calculated by a state estimator that reduces the uncertainty to 2%
Experimental work and results The different control signals sent by the ANM to curtail the power output of the generator on busbar 7, Gen 4, in presence of different levels of uncertainty The differences between the control signals (relative to the base case with no uncertainty) sent in presence of uncertainty
Experimental work and results The analysis found that no divergence of the load flow engine was encountered when erroneous measurements, up to 6.5%, were presented to the ANM software. However, with data uncertainty it can be seen that the error, in some situations, is large enough to either move the curtailment to a deeper set point (next domain value for the variable) or not curtail sufficient levels of DG. The studies have also highlighted the importance of the reduction in uncertainty, for example through the use of a SE. The uncertainty of the input data is reflected in the uncertainty of the final power flow calculation, so the operators taking in account of the uncertainty reduction introduced by a SE can adopt less conservative thermal detection limit to compensate for expected errors.