Contact
Leaderboard
Team
Stars
ASU
Mountain Bees
Notice

The GO Competition is currently in the Beta Testing Phase, also known as Phase 0.

During this phase, we provide you the opportunity to get familiar with the competition platform: the problem to be solved, formats for input and output files, registration, algorithm submission and scoring. We encourage you, as a potential competitor, to try out the platform and raise questions through the Forum or Contact Us directly. Your participation in this phase will help improve the competition platform and process.

Participation at this point carries no obligation for the future. Team membership is flexible at this point.

The formal competition is subject to appropriation of funding.

Datasets

Use of the term "scenario"

The term "scenario" in the context of power systems often means a load snapshot from a coupled sequence of time data. That is not how it is used here. Here a scenario is used to represent a snapshot instance with no correlated relationship to any other instance in the list of scenarios in the dataset. Not only can the load change between one of our scenarios and another, but so can any of the other parameters, such as line limits or generator limits. The basic network model, the number of buses, generatiors, lines, etc., remains fixed for all the scenarios within a network model dataset.

Initial Trial Dataset

Please use this dataset for your initial submission to verify that you are producing valid forms of the files solution1.txt and solution2.txt used for scoring. This dataset contains only one scenario while all other datasets contain multiple scenarios (10 for the 179 bus datasets, 100 for the IEEE 14 and RTS96 datasets, and 105 for the Original Dataset) so please make sure this one is being processed correctly before attempting the others.The results of this submission will not be visible on the leaderboard, only to the individual/team that made the submission. An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 70 kB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Filename Available Formats

Phase 0: IEEE 14-Bus (1 Scenario)

.zip (15.2 KB)
Download Datasets
Filename Available Formats

Phase 0 : RTS96springwday_1

(100 scenarios)

.zip (10.3 MB)

Phase 0 : Modified RTS96springwday_1

(100 scenarios)

.zip (9.75 MB)

Phase 0 : IEEE 14-Bus

(100 scenarios)

.zip (1.41 MB)

Phase 0 : Modified IEEE 14-Bus

(100 scenarios)

.zip (1.43 MB)
Phase 0 : Feasible 179-Bus (10 scenarios) .zip (12.0 MB)
Phase 0 : Infeasible179-Bus (10 scenarios) .zip (12.7 MB)

Phase 0 : Original Dataset

(105 scenarios; 2 network models)

.zip (4.22 MB)
Dataset Contents

At the root level of each zip file is the scorepara.csv file that contains the dataset name and scoring parameters. For more details on these parameters, pleas see the scoring section:

  • Nominal time value, tnom.,A: seconds to complete the longest scenario in a power systems network model with the reference GAMS algorithm, rounded up a whole single digit
  • Nominal objective value, cnom.,A: maximum objective across scenarios in a power systems network model with the reference GAMS algorithm, rounded up a whole single digit)
  • Time scale: when multiplied by tnom.,A, this gives the scenario dependent threshold,  tA,i
  • Constraint violation penalty scale: when multiplied by cnom.,A, this gives the score, yA,i, when constraints are violated or an algorithm times-out
  • Time violation penalty scale: when multiplied by tnom.,A, this gives the score, xA,i, when an algorithm returns a solution in time greater than  tA,i
  • Maximum infeasibility: constraint violations that exceed this value are not considered satisfied (1e-6 in most scenarios )
  • Number of scenarios associated with each power systems network model in the dataset

Each network model dataset contains 1 or more folders labeled scenario_1 to scenario_n. These scenarios are each independent scenarios instances with no coupling to any of the other scenarios.  Each scenario folder contains the following files:

  • powersystem.raw
  • generator.csv
  • contingency.csv
  • pscopf_data.gms
  • pscopf_data.mat
  • pscopf.m
  • solution1.txt (in Phase 0 datasets only)
  • solution2.txt (in Phase 0 datasets only)

powersystem.raw is a PSSE Raw file version 33.5, containing bus, load, fixed shunt, generator, branch, transformer, and other control variable data for a power system. Raw files may contain other data not relevant to the PSCOPF problem. This format is used because it is a common way to describe network configurations. It is not an endorsement of PSSE or imply that PSSE is necessary to read or in any way process the data. A powerflow solver is not necessary to provide the PSCOPF solution to these problems and a powerflow solver is not used in evaluating the correctness of the results.

generator.csv is a comma separated value (csv) file that defines the generator cost functions. The headers are Bus Number, Generator ID, Term ID, and Value. The generation cost is a quadratic function, the constant term has Term ID 0, the linear term with ID 1, and the quadratic term has ID 2. Term ID 9 is for participation factor of a generation. The cost data and participation factor are NOT per unit. Default value of the field "Value" is 0.0. 

The first 4 lines (out of 235) of generator.csv for scenario_1 of the Phase 0: RTS96sprindwday_1 dataset are:

			Bus Number,Generator ID,Term ID,Value 
			101,1,0,1375.9
			101,1,1,140.3842614
			101,1,2,2.082286932
			101,1,9,5

This means the cost function of generator 1 linked to bus 101 is $1375.9+140.3843 p + 2.082287 p^2$, where $p$ is active power in MW. The participation factor is 5.0.

contingency.csv is a comma separated value (csv) file that defines a list of contingencies. The headers are Contingency ID, Type, From, To and CID. The first field provides the index for each contingency. These indices will be used in the standard output file of solutions. Field "Type" identifiers are B for branch, T for transformer and G for generator. For Type B, fields "From" and "To" are the indices of the end buses of the failed branch and CID is the ID of the circuit linking the two buses. For Type T, fields "From" and "To" are the indices of the end buses of the failed transformer. For Type G, field "To" is the index of the failed generator and "From" is the index of the bus to which the failed generator is connected. With Type G, the CID field is not used and should be empty.

A contingency list includes all contingencies to be considered in a PSCOPF problem. Each line represents a single contingency.

All 11 lines (10 contingencies) of contingency.csv of the Phase 0: RTS96sprindwday_1 dataset are:

			Contingency ID,Type,From,To,CID
			1,B,101,102,1
			2,B,106,110,1
			3,B,111,114,1
			4,B,116,117,1
			5,B,121,122,1
			6,B,205,210,1
			7,B,216,219,1
			8,B,312,323,1
			9,B,317,322,1
			10,B,325,121,1

Line 2 of the file says the first contingency is of type Branch between bus 101 and bus 102 over circuit 1.

pscopf_data.gms contains the same data from the first three files, but in GAMS compatible format.

pscopf_data.mat contains the same data from the first three files, but in MATLAB binary format.

pscopf.m contains the same data from the first three files, but in MATLAB compatible format.

solution1.txt contains the required output for the base solution described under Evaluation.

solution2.txt contains the required output for the contingency solutions described under Evaluation.

Please note:  The solution1.txt and solution2.txt files are included in the dataset for demonstration purposes only for Beta Testing Phase (Phase 0).  They will not be included during competitive phases.

Dataset Descriptions

Phase 0: IEEE 14-Bus (1 scenario )

The IEEE 14-bus test case is a model of the American Electric Power System in 1962. IEEE14 has had many additions over the years: line limits, generator costs, etc. For the PSCOPF problem, a few security contingencies were defined. This is one of the feasible scenarios generated by random perturbation of some of the limits (e.g. on generator power output and branch flows). This problem is small enough that it should not pose any challenge to off-the-shelf optimization solvers, commercial or otherwise.

This should be the dataset used with your first submission. Unlike the other datasets, it consists of only a single scenario so any error, i.e., in creating solution files, is not repeated multiple times. The results are not tracked on the leaderboard. Only the person who made the submission, and their team, can see the results. Once you have recieved a score for this dataset equal to the objective value of your solution (the reference value is 2,912,857.26) you are ready to submit multiple scenario datasets.

An example of the information returned from the GO Competition Reference GAMD submission for this dataset is available here, a 70 kB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Phase 0: IEEE 14-Bus (100 scenarios)

This dataset uses the same power system network model for the IEEE 14-bus system as above but is composed of 100 scenarios, all known to be feasible, generated by random perturbation of some of the limits (e.g. on generator power output and branch flows). These problems are small enough that they should not pose any challenge to off-the-shelf optimization solvers, commercial or otherwise.

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 6.7 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Phase 0: Modified IEEE 14-Bus (100 scenarios)

This dataset uses a modified power system network model based the IEEE 14-bus system mentioned above. The modification consists of removing one line, linking buses 7 and 9. The 100 scenarios, all known to be feasible, were generated by random perturbation of some of the limits (e.g. on generator power output and branch flows). These problems are small enough that they should not pose any challenge to off-the-shelf optimization solvers, commercial or otherwise.

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 6.6 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Phase 0: RTS96 (100 scenarios)

This dataset is composed of 100 scenarios, all known to be feasible, based on the 1996 update (“The IEEE Reliability Test System - 1996”, IEEE Transactions on Power Systems, Vol. 14, No. 3, August 1999) of the IEEE Reliability Test System, originally published in 1979. The scenarios were generated by random perturbations of some of the limits (e.g. on generator power output and branch flows) taken from the spring windy day 1 data.

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 96 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Phase 0: Modified RTS96 (100 scenarios)

This dataset is composed of 100 scenarios, all known to be feasible, based on the 1996 update (“The IEEE Reliability Test System - 1996”, IEEE Transactions on Power Systems, Vol. 14, No. 3, August 1999) of the IEEE Reliability Test System, originally published in 1979, modified by removing one line, linking buses 121 and 325, in order to generate a distinct power system network model. The scenarioswere generated by random perturbations of some of the limits (e.g. on generator power output and branch flows) taken from the spring windy day 1 data.

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 86 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Phase 0: Feasible 179-Bus (10 scenarios)

This dataset is composed of 10 scenarios, all known to be feasible, based on the 179-bus model of the Western Systems Coordinating Council (WECC/WSCC) developed at the University of Wisconsin – Madison ("DC Multi-infeed Study," Electric Power Res. Inst. TR-104586, Dec. 1994.) For this dataset special emphasis was placed on generating a large set of security contingencies, with scenarios  generated by random perturbations of some of the limits (e.g. on generator power output and branch flows).

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 69 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation.

Phase 0: Infeasible 179-Bus (10 scenarios)

This dataset is composed of 10 scenariosbased on the 179 bus model of the Western Systems Coordinating Council (WECC/WSCC) developed at the University of Wisconsin – Madison ("DC Multi-infeed Study," Electric Power Res. Inst. TR-104586, Dec. 1994.) For this dataset special emphasis was placed on generating a large set of security contingencies. Multiple scenarioswere generated by random perturbation of limits, e.g. on generator power output and branch flows, and ten were selected has having no known feasible solution. Since the initial selection, scenarios 2, 3, and 4 have been shown to be feasible.

The reactive power flow balance constraint violations (pu) for each scenario observed by the GAMS reference solution are given in the table below. Note that the violation for scenario 7 is for a generator PV PQ switch violation and not for reactive power flow balance. The reactive power flow balance constraint violation was less than the threshold.

scenario Violation used in scoring Violation from GAMS Time in minutes
1 108.3277324 108.3277324 0.9
2 7.12E-7 2.06E-7 2.1
3 5.69E-7 5.78E-8 7.0
4 6.44E-7 1.22E-7 1.4
5 112.387487 112.387487 1.9
6 107.5294908 107.5294908 1.0
7* 0.013162441* 0.013162441* 12.9
8 11.6900435 11.6900432 43.8
9 99.48148377 99.48148377 1.0
10 119.8815278 119.8815278 1.1

* constraint violation is for a generator PV PQ switch violation in scenario 7. All other values are for reactive power flow balance constraint violations.

A solution is considered feasible if it is less than the maximum constraint violation set by the maximum infeasibility paramater in the scoreparam.csv file, in this scenario 1.e-6.

The purpose of this dataset is to challenge solvers. Solvers that find feasible solutions to the scenarios currently considered infeasible are considered noteworthy. Solvers that can recognize infeasible solutions and quit early are also noteworthy.

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 62 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation. The GO Competition Reference GAMS implementation independently calculates the constraint violations and writes them to solution3.txt for each scenario. Comparing the solution3.txt results to score.csv shows very good agreement except when the violations are less than 1.0e-6, when the single precision accuracy of the some of the input values influences accuracy. For scenario 2, GAMS finds a slightly larger violation of 1.55E-7 for a real power flow balance constraint violation.

Phase 0: Original Dataset (105 scenarios)

This dataset is an example of a multiple network dataset of the format that will be used in the Phase I and Phase II competition. It consists of 75 scenarios taken from the Phase 0 IEEE 14 Bus dataset and 30 scenarios from the Phase 0 RTS96 dataset for a total of 105 scenarios.

An example of the information returned from the GO Competition Reference GAMS submission for this dataset is available here, a 34 MB tar.gz file. Note: some of the information in the tar.gz file is specific to the Reference GAMS implementation. The score.csv file in the tar.gz file, as with all the single network model datasets, gives a score for each scenarios and at the end of the scenarios for the first network model gives a Network model score that is the geometric mean of the individual scenario scores for the network model. Since this dataset has two network models, the scenario scores for the second network model follow along with its Network model score. The last line of the file gives the Final dataset score that is the geometric mean of the two Network model scores. This is the score used to rank the performance of the submission.