Computer-Use Agent Taxonomy中文
R07.08.01.01 · TASK CLASS

Power flow & contingency studies

Run power flow, OPF and contingency studies and report violations.

20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.

Typical applications
PSS/E · PowerWorld · ETAP · pandapower
Class GUI-need label
HY · Hybrid

Task proposals · 20

Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.

R07.08.01.01-T01 Flag voltage violations on a modified IEEE 14-bus case in pandapower

Original instruction

Load fixtures/case14_mod.json (IEEE 14-bus with three loads raised 25%) into pandapower, run a Newton-Raphson AC power flow and write output/violations.csv listing every bus outside 0.95-1.05 pu (bus, vm_pu, side=low/high) and every line above 100% loading (line, loading_pct). Do not modify the network file.

Platforms
mac · macOS win · Windows lin · Linux
Applications
pandapower
Interfaces
term · Terminal api · API/SDK
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP02 · Records/table
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Grader reruns pandapower on the untouched fixture; violating bus and line sets must match exactly, vm_pu within 1e-3, loading within 0.5 pts. Fixture hash checked to detect edits.

Proposed partial credit

20% file parses with required columns; 40% bus set exact; 40% line set exact with values in tolerance.

R07.08.01.01-T02 Draw and solve a new 230 kV tie line on a PowerWorld one-line

Original instruction

Open fixtures/wscc9_base.pwb with its one-line wscc9.pwd in PowerWorld Simulator. Draw a new 230 kV line from Bus 5 to Bus 7 on the one-line (R=0.0085, X=0.072, B=0.149 pu, Rate A 250 MVA), solve the power flow and save both case and one-line. Report the new line's MW flow in output/answer.txt.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click drag · Drag & drop type · Text & forms
Work product
WP07 · Simulation result
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Grader opens saved .pwb/.pwd via its own SimAuto session: branch 5-7 exists with exact parameters, oneline contains a line object anchored to both bus symbols, case solves; reported MW within 1 MW. SimAuto/COM is deregistered in the agent VM, and .aux imports are logged and fail the task.

Proposed partial credit

30% branch with exact R/X/B/rating; 30% line drawn and anchored on one-line; 20% case solved & saved; 20% MW answer correct.

R07.08.01.01-T03 Run an N-1 ACCC study in PSS/E and list thermal overloads

Original instruction

Using PSS/E, open fixtures/regional_500bus.sav and build subsystem, monitored-element and contingency files covering every 230 kV and 345 kV branch outage in areas 2 and 3. Run AC contingency analysis (ACCC) and produce output/overloads.csv with contingency, element, flow MVA and percent of Rate A for all loadings above 100%.

Platforms
win · Windows
Applications
PSS/E
Interfaces
desk · Desktop GUI
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms code · Code & commands
Work product
WP02 · Records/table
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Grader runs its own ACCC on the same case with reference .sub/.mon/.con; overload rows must match as a set (contingency, element) with percent within 1 pt. Agent .con file parsed to confirm it contains all 2-3 area 230/345 kV branches.

Proposed partial credit

25% .con covers required outages; 25% ACCC runs to completion; 50% overload set matches (F1 weighted).

R07.08.01.01-T04 Set ETAP transformer taps so plant MCC buses stay within band

Original instruction

Open the ETAP project fixtures/plant_13p8kV (13.8 kV industrial plant). In the Max Load study case, adjust the no-load taps of transformers T3-T6 (2.5% steps, +/-5%) so every 480 V MCC bus is between 0.97 and 1.03 pu, run Load Flow and save the project. Taps outside the allowed step set do not count.

Platforms
win · Windows
Applications
ETAP
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms vis · Visual judgment
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Grader opens the saved project, reads tap settings from the project database and reruns Load Flow headlessly: all 9 MCC buses in [0.97,1.03] pu and taps in {-5,-2.5,0,2.5,5}%. Other element data diffed against original; any change outside T3-T6 taps fails. The ETAP API (etapPy) is disabled in the agent VM, and project-DB writes by any process other than ETAP are detected via the file audit log.

Proposed partial credit

20% taps on allowed steps; 50% fraction of MCC buses in band; 30% no collateral edits and study saved.

R07.08.01.01-T05 Solve a DC OPF with revised generator costs and report LMPs

Original instruction

With MATPOWER or pandapower, load fixtures/case30_costs.m, apply the quadratic cost coefficients in input/gen_costs_2026.csv, run a DC optimal power flow with line limits enforced and write output/opf.json containing total cost ($/h), dispatch per generator (MW) and the LMP at every bus.

Platforms
mac · macOS win · Windows lin · Linux
Applications
MATPOWER · pandapower
Interfaces
term · Terminal
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP07 · Simulation result
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader solves the same DC OPF; total cost within 0.1%, each dispatch within 0.5 MW, LMPs within $0.05/MWh. JSON schema enforced.

Proposed partial credit

20% schema valid; 30% total cost; 25% dispatch; 25% LMPs.

R07.08.01.01-T06 Compute per-bus PV hosting capacity on an LV feeder

Original instruction

Load fixtures/feeder_lv.json (42-bus LV feeder, summer midday load) in pandapower. For each of buses 10-41 separately, find the maximum PV injection (1 kW resolution, unity PF) before any bus exceeds 1.05 pu or any line exceeds 100% loading. Save output/hosting.csv (bus, max_kw, binding=voltage|thermal).

Platforms
mac · macOS win · Windows lin · Linux
Applications
pandapower
Interfaces
term · Terminal api · API/SDK
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP02 · Records/table
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Grader recomputes hosting capacity by bisection; per-bus max_kw within 2 kW and binding type correct. Score is fraction of buses correct.

Proposed partial credit

20% all 32 buses present; 60% max_kw within tolerance (pro-rata); 20% binding constraint labels correct.

R07.08.01.01-T07 Build a 7-bus case in PowerWorld from a PDF one-line drawing

Original instruction

In PowerWorld Simulator, create a new case from input/oneline_7bus.pdf: place 7 buses, 2 generators, 4 loads, 8 lines and 1 transformer on a one-line with the tabulated impedances and ratings, set bus 1 as slack, solve and save output/case7.pwb and case7.pwd. The one-line must visually follow the PDF topology.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click drag · Drag & drop type · Text & forms geom · Precise geometry
Work product
WP07 · Simulation result
Conditions
licensed · Licensed pro app long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

Grader loads the case: element counts, impedances and ratings exact; solved flows within 1 MW / 1 Mvar of reference. Bus symbol coordinates compared to PDF layout (relative ordering, IoU of bounding layout >0.6). .aux/SimAuto creation detected via app log and fails.

Proposed partial credit

25% topology and counts correct; 25% parameters exact; 30% solved flows match; 20% one-line layout matches PDF.

R07.08.01.01-T08 Repair a non-converging PSS/E summer peak case

Original instruction

fixtures/summer_peak_nc.sav fails to converge with FNSL. Using PSS/E diagnostics (mismatch reports, branch listings), find the data-entry error in one branch impedance, correct it to the value in input/line_ratings_2026.xlsx, solve to below 0.1 MW mismatch and save output/summer_peak_fixed.sav. Record the branch and old/new values in output/fix.txt.

Platforms
win · Windows
Applications
PSS/E
Interfaces
desk · Desktop GUI
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP07 · Simulation result
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Grader exports both cases to RAW and diffs: exactly one branch changed and its X equals the spreadsheet value; fixed case converges with mismatch <0.1 MW. fix.txt must name the same branch.

Proposed partial credit

30% faulty branch identified in fix.txt; 40% only that branch corrected; 30% saved case converges.

R07.08.01.01-T09 Configure and run area-wide contingency analysis in PowerWorld

Original instruction

Open fixtures/texas_2k_synth.pwb in PowerWorld. Auto-insert single-branch contingencies for area 'North Central' only, set limit monitoring to 100% of Rate B and 0.92-1.08 pu voltage, run contingency analysis and export violations to output/ctg_violations.csv. Save the case with the contingency set.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms nav · Pan/zoom/scroll
Work product
WP02 · Records/table
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Grader reloads the saved case: contingency list equals branches in the area, limit settings exact; reruns CA and compares violations as a set (F1 >=0.95 for full credit). SimAuto disabled in agent VM; script/aux imports logged and penalized.

Proposed partial credit

25% contingency set exact; 20% limit settings; 45% violation F1; 10% case saved.

R07.08.01.01-T10 Find the lowest-voltage bus after a line outage in PowerWorld

Original instruction

Open fixtures/ieee39.pwb in PowerWorld Simulator. Open line 16-17, re-solve and report in output/answer.txt the bus with the lowest voltage and its value in pu (3 decimals). Restore the line afterwards and save nothing.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click vis · Visual judgment
Work product
WP12 · Answer/decision
Conditions
licensed · Licensed pro app
Proposed difficulty
2 / 5

Proposed verifier

Answer compared to reference post-outage solution: bus exact, voltage within 0.002 pu. Case file hash unchanged. The SimAuto COM server is disabled in the agent VM and .aux script loads are detected in the Simulator log and fail, so the outage must be applied in the PowerWorld GUI.

Proposed partial credit

50% correct bus; 30% voltage within tolerance; 20% case left unmodified.

R07.08.01.01-T11 Run a daily OpenDSS time series on IEEE 123 with regulators

Original instruction

Use OpenDSS to run fixtures/ieee123/Run_IEEE123Bus.dss in daily mode with the 1-hour loadshape in input/loadshape_jul.csv. Write output/regs.csv (hour, regulator, tap) for all regulators and output/vviol.csv listing hours with any bus voltage outside 0.95-1.05 pu.

Platforms
win · Windows
Applications
OpenDSS
Interfaces
term · Terminal desk · Desktop GUI
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP02 · Records/table
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader runs the same daily simulation; regulator taps must match per hour (>=95% of entries), and the violation-hour set must match exactly.

Proposed partial credit

15% loadshape bound correctly; 45% regulator tap table match; 40% violation hours exact.

R07.08.01.01-T12 Identify top transfer-sensitive lines with PowerWorld PTDFs

Original instruction

In PowerWorld, open fixtures/wecc_240.pwb and compute PTDFs for a transfer from area 'Desert SW' (seller) to area 'NorCal' (buyer). Report the 5 branches with the highest absolute PTDF and the transfer MW that first overloads each one (Rate A) in output/ptdf.csv.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP02 · Records/table
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Grader computes PTDFs on the same case; top-5 set must match, PTDF within 0.005 and limiting transfer within 5 MW.

Proposed partial credit

40% top-5 set; 30% PTDF values; 30% first-overload transfer values.

R07.08.01.01-T13 Choose least-cost reinforcements that clear all N-1 overloads

Original instruction

fixtures/growth_2030.sav (PSS/E) shows N-1 thermal overloads. From the 9 candidate projects in input/candidates.xlsx (cost, impedances), pick the minimum-cost set that leaves zero N-1 overloads on monitored 230 kV+ elements. Save output/plan_2030.sav with the projects added and output/plan.md listing projects, cost and residual worst loading.

Platforms
win · Windows
Applications
PSS/E · Excel
Interfaces
desk · Desktop GUI term · Terminal
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms code · Code & commands multi · Cross-app
Work product
WP07 · Simulation result
Conditions
licensed · Licensed pro app long · Long-horizon multiapp · Multi-app
Proposed difficulty
5 / 5

Proposed verifier

Grader applies the listed projects to the original case, runs its own N-1: zero overloads required; cost must equal precomputed minimum (partial if within 15%). Saved case diffed to confirm only listed projects added.

Proposed partial credit

20% projects correctly modeled; 40% zero N-1 overloads; 30% cost optimal; 10% memo consistent.

R07.08.01.01-T14 Build a substation network in pandapower from a spreadsheet

Original instruction

Create a pandapower network from input/substation_data.xlsx (12 buses, 3 two-winding transformers, 9 lines, 7 loads, 1 external grid) and run a power flow. Save output/net.json and output/results.csv with bus voltages (pu, angle) and transformer loading (%).

Platforms
mac · macOS win · Windows lin · Linux
Applications
pandapower · Excel
Interfaces
term · Terminal api · API/SDK
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP07 · Simulation result
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Grader rebuilds the network from the sheet: element counts and parameters equal; result voltages within 1e-3 pu, angles within 0.05 deg, loading within 0.5 pts.

Proposed partial credit

30% topology/counts; 30% parameters; 40% results within tolerance.

R07.08.01.01-T15 Set up three ETAP load-flow scenarios with alert limits

Original instruction

In the ETAP project fixtures/campus_medium, create study cases MinLoad, Normal and MaxLoad using loading categories 2, 1 and 3, set marginal and critical alerts at 98%/95% undervoltage and 95%/100% loading, run Load Flow for each and export the three alert summaries to output/alerts_<case>.xlsx.

Platforms
win · Windows
Applications
ETAP
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Grader inspects the project database: three study cases with exact loading categories and alert values; reruns each study and compares alert element sets with the exported files (exact). No external project-file edits (DB checksum of untouched tables). The ETAP API (etapPy) is disabled in the agent VM, so study cases can only be built in the ETAP GUI.

Proposed partial credit

30% study cases configured; 20% alert limits exact; 50% three alert exports match reruns.

R07.08.01.01-T16 Run an AC OPF in PowerWorld and report binding constraints

Original instruction

Open fixtures/ieee118_opf.pwb in PowerWorld Simulator, enable line-limit enforcement, run the LP OPF and write output/opf.csv with LMPs at buses 10, 49, 59, 80 and 100, plus a list of binding branch constraints with their shadow prices.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP07 · Simulation result
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Grader runs OPF on the same case: LMPs within $0.10/MWh, binding constraint set exact, shadow prices within 2%.

Proposed partial credit

40% LMPs; 40% binding set; 20% shadow prices.

R07.08.01.01-T17 Report total losses for IEEE 118 at 110% load

Original instruction

In pandapower load the IEEE 118-bus case, scale all loads (P and Q) by 1.10, run AC power flow and write output/losses.txt with total active losses in MW (2 decimals) and the line with the highest loss.

Platforms
mac · macOS win · Windows lin · Linux
Applications
pandapower
Interfaces
term · Terminal
GUI need · task setting
CL · CLI-sufficient
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
code · Code & commands
Work product
WP12 · Answer/decision
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Grader computes the same: losses within 0.05 MW, highest-loss line exact.

Proposed partial credit

60% losses within tolerance; 40% correct line.

R07.08.01.01-T18 Update a PSS/E case to a SCADA snapshot and rank contingencies

Original instruction

Update fixtures/ops_base.sav in PSS/E to the operating point in input/scada_0815_1400.csv (unit MW/Mvar, load MW/Mvar, breaker states), solve, run N-1 on all monitored 345 kV elements and write output/ops_memo.md for the shift supervisor naming the 3 worst post-contingency loadings with percent of emergency rating.

Platforms
win · Windows
Applications
PSS/E · Excel
Interfaces
desk · Desktop GUI term · Terminal
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
type · Text & forms click · Point & click code · Code & commands multi · Cross-app
Work product
WP01 · Document
Conditions
licensed · Licensed pro app multiapp · Multi-app long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

Grader applies the snapshot independently: per-unit dispatch and breaker state diff of saved case must be zero; top-3 contingency/element pairs and percents (within 1 pt) parsed from the memo's table.

Proposed partial credit

35% snapshot applied exactly (breakers, gens, loads); 15% case solves; 50% top-3 ranking and values.

R07.08.01.01-T19 Find the PV-curve nose point for an inter-area transfer

Original instruction

In PSS/E, open fixtures/two_area_stab.sav and run PV analysis for a transfer from area 1 to area 2, scaling generation in area 1 and load in area 2, under the base case and the 3 contingencies in input/pv.con. Report in output/pv.csv the maximum transfer MW per case and the limiting case.

Platforms
win · Windows
Applications
PSS/E
Interfaces
desk · Desktop GUI
GUI need · task setting
HY · Hybrid
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP07 · Simulation result
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Grader runs PV analysis with the same subsystems and step size; nose-point MW within 2% for each case; limiting contingency exact.

Proposed partial credit

30% subsystems/transfer defined correctly; 50% nose points within tolerance; 20% limiting case.

R07.08.01.01-T20 Run a client contingency file and flag elements missing from the case

Original instruction

A client sent input/client_ctg.aux for fixtures/coop_250.pwb. Load it in PowerWorld, run contingency analysis and export violations to output/viol.csv. Some contingencies may reference elements not in the case: list those in output/issues.txt with the reason. Do not invent substitutes.

Platforms
win · Windows
Applications
PowerWorld Simulator
Interfaces
desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
CL · CLI-sufficient
Interaction primitives
click · Point & click type · Text & forms
Work product
WP02 · Records/table
Conditions
licensed · Licensed pro app clarify · Ask when unclear
Proposed difficulty
3 / 5

Proposed verifier

Exactly 3 planted bad references exist; issues.txt must list all 3 and nothing else. Violations for valid contingencies compared with grader run (set F1 >=0.95). Any contingency silently remapped to another element fails that part. SimAuto COM is disabled in the agent VM; the Simulator log must show only the client .aux being loaded, so agent-written .aux scripts are detected and fail.

Proposed partial credit

40% all invalid references reported, no false ones; 40% violation set; 20% no fabricated remaps.

Mapped research references

No benchmark reference is assigned to this class in the current mapping.

In this workflow family