Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Saved Studies

A Study is a saved case with its changes and results. Open Studies, give the current case a name, and select Save study. A planning goal is optional. Saving keeps the current result when it matches the case; an unsolved case can also be saved. Saving does not move the network or reset the camera.

Case, History, and Plan

  • Case shows the selected saved state. Demand rows show base demand, current demand, and the accumulated change in MW. Edits at different buses remain part of one cumulative set of changes.
  • History holds saved states and activity. Select a state to inspect it or branch from it to explore another option. Observations, edits, solves, planning trials, and decisions have distinct labels. An observation adds evidence without inventing an electrical state.
  • Plan is optional. Choose a goal, then explicitly choose the whole network, an area, or selected equipment. Searchable tables show bus names and IDs, line endpoints, weights, limits, and units. The displayed counts include every selected element; there are no hidden bus or line limits.

Reset to base case restores the original electrical inputs and retains previous states and activity. The base is the network data supplied to the Study, which can differ from its starting state when demand was already edited. Older imported Studies without a base input report that reset is unavailable.

The saved state currently being inspected, the recommended candidate, and the applied state remain distinct. Selecting a candidate changes the displayed network. An agent proposal requires an explicit Apply action. Live case returns to the case outside the Study.

Planning

The OPF objective describes operating cost inside the power-system calculation. A planning goal describes what to improve across candidate cases, such as a weighted LMP or voltage target. Custom expressions remain available through the structured API: weighted observables, sums, scaling, squared target deviations, and direct intervention penalties.

Line capacity uses MW for DC OPF and MVA for SOCWR. Demand uses MW. Limits and increments apply to the cumulative changes from the goal’s starting state. Demand placement allocates a stated additional total; redistribution preserves the total through paired transfers.

Find a proposal uses the objective gradient to choose candidate edits, solves each candidate, and retains verified improvements. Every attempted solve counts against Solve budget, including failed trials and any required starting-point solve. The result is the best verified candidate found within that budget. It does not establish a global optimum. Expand the evidence for prediction error, active-constraint changes, failed trials, and numerical settings.

The derivative uses a combined adjoint calculation, including direct penalties, without constructing a dense observable-by-decision matrix. Changing the goal creates a new revision and retains earlier results. Changing a proposal’s goal or starting state invalidates its approval.

Save, reopen, and continue

The browser stores completed operations atomically in IndexedDB. Export creates a portable bundle containing PowerIO generation-2 inputs and solutions, deduplicated SHA-256 artifacts, goals, states, and activity. Geographic layers, line paths, drawings, and the selected view travel with it. Import checks versions, hashes, identities, and references. It never restores approval tokens or executes imported text. Older journals remain historical evidence when their electrical states are unavailable.

A failed save reports the problem instead of discarding history. Free storage or export the saved Study before retrying. Cancellation retains completed trials and the best candidate after the running solve finishes.

Native and agent access

cargo build -p tellegen-cli --features conic
tellegen describe
tellegen study create study.json < create-request.json
tellegen study inspect study.json
tellegen study run study.json < operation-request.json
tellegen study export study.json > portable-study.json

tellegen describe lists commands and generated JSON schemas. Rust definitions also generate the TypeScript types. Browser controls, WebMCP, and the CLI use the same Study operations and revision checks.

PowerMCP’s Tellegen server (powermcp run tellegen) invokes the native CLI. Its solve, solve_module, and plan tools handle calculations; study_create, study_inspect, study_run, study_export, and study_import handle saved cases. study_create also accepts a grid file for PowerIO to read. The server provides read-only capability descriptions and leaves applying a proposal to an explicit user action: study_run refuses apply.

WebMCP list_cases and select_case change the visible case without proposal approval. Network queries identify the displayed case and saved state, so an LMP question refers to the same result shown on screen. Creating a Study is needed only when saving or planning is requested.

Supported calculations are DC OPF, AC power flow, and, with the conic build option, SOCWR. Unsupported objective and solver combinations are rejected before planning.

Model details

Texas7k uses an explicitly configured convex quadratic or linear fit to its piecewise generator costs. Model details records the method, original and prepared costs, affected generators, units, and measured breakpoint errors. Those errors describe an approximation; they are not a relaxation bound. The prepared costs persist through saved cases, exports, and reset.

The same preparation is available for any balanced case through tellegen prepare-model. The command returns prepared PowerIO IR and its approximation report. Other cases retain their declared costs by default.