Advances in Electric Power and Energy. Группа авторов

be less than 10 MVA. He does not distinguish between internal and external systems.

       1.4.4.1 Metrics to Evaluate SE Solution Quality

      More than one metric is used to evaluate the accuracy of the results of the state estimator solution:

       Cost index is also referred to as “performance index” or “quadratic cost.” In general, it measures the sum of the squares of the normalized estimate errors (residuals). Increasing cost index values could indicate deteriorating state estimator solution quality. This is the most commonly used indicator, whose values range between 45 and 58%.

       Chi‐squared criterion is the second most used, and its value ranges between 36 and 42%.

       Measurement error/bias analysis is used as a performance indicator.

       Average residual value is used as a performance indicator.

      The reliability entity should track the selected metric over time to establish the pattern and determine what indicates a problem with state estimator solution quality. Deviation from the “normal range” of these metrics should trigger state estimator maintenance and support. These metrics are important because they could affect the CA solution.

      Many factors affect SE solution‐quality metrics such as:

      1 Electrical device modeling, connectivity, and telemetry data mapping. If the topology is incorrect, the state estimator may not converge or may yield grossly incorrect results. A topology error may be caused by either inaccurate status of breakers and switching devices or errors in the network model.

      2 Availability and quality of telemetry data. Telemetry data are essential components of the state estimation process.

      3 Inadequate observability. State estimation is extended to the unobservable parts of the network through the addition of pseudo‐measurements that are computed based on load prediction using load distribution factors, or they can represent non‐telemetered generation assumed to operate at a base‐case output level. The quality of pseudo‐measurements may be bad if they are not updated regularly to reflect current conditions.

      4 Measurement redundancy of the network is defined as the ratio of the number of measurements to the number of state variables in the observable area of the network.

       1.4.4.2 Methods for Evaluating SE Solution Quality (Accuracy)

      The following methods are used to evaluate the accuracy of the state estimator results:

       Continually monitor and minimize the amount of bad data detected by correcting model, telemetry, and bad status.

       Compare critical telemetry with the state estimator solution (ties, major lines, large units, etc.).

       Use measurement error/bias analysis to detect and resolve telemetry and model problems.

       Periodically review all stations to correct high residuals and minimize all residuals as much as reasonably possible.

       Compare CA results to actual system.

       Compare power flow results with actual system.

       Compare state estimator actual violations to see if they closely match actual SCADA violations.

       Compare state estimator total company load/generation/interchange integrated over time to see if it closely matches billing metering.

      1.4.5 Using SE to Monitor External Facilities

      The state estimator solution quality when it is used to monitor external facilities depends on the accuracy of their models. The external network models could affect the quality of state estimator solutions by:

      1 Propagation of errors into the internal model solution from the external model solution. This applies to one‐pass state estimators if the external network model solution is mainly based on forecasted and/or pseudo‐measurements rather than telemetered data. The external network model equivalencing methods could also cause errors to propagate. For two‐pass state estimator, there could be boundary problems (between the internal/observable solution and the external/unobservable solution) that could cause the total network solution to not converge.

      2 Measurement density in the external system. Many buses in external models are measurement unobservable. The low values for the external‐status‐point‐to‐external‐bus ratios for many respondents (i.e. less than one status point per bus) indicate that many external buses do not have telemetered breaker/switch information, which implies a bus‐branch‐type external model (i.e. a planning model) for many buses.

      3 Convergence issues related to external models and/or telemetry data for external model. Measurements for the external network model usually originate from data links. As a result, data availability depends on data link availability.

      4 The impact of interchange transactions, especially for the external portion of the model, could influence the state estimator solution.

      5 Adding detail or expanding the external network model could affect the throughput (execution time) of the state estimator application.

      External network model improvements are expected to enhance the accuracy of the results.

       Adding breaker/switch detail to the external and internal models

       Adding extensive telemetry to the external and internal models

       Adding lower‐voltage detail to the external and internal models

       Adding one or more control areas to the external model

       Creating a new external model

      1.4.6 SE Maintenance/Troubleshooting and Support Practices

      Many users have state estimator support personnel available continuously. Most users monitor state estimator status on a continuous basis (24 × 7 × 365) and maintain their state estimators with in‐house staff, and some use vendor staff in addition to in‐house staff for support.

      Most users notify operators and control room staff of a state estimator failure. State estimator status is presented primarily via alarm tools and physical displays. Some users page and send email notifying of a state estimator failure.

      Operators attempt to resolve state estimator problems prior to notifying support personnel.

      Many users have a process to investigate and debug unsolved/non‐converged and bad/inaccurate state estimator solutions.

      The operator receives an alarm notification of state estimator problems and then calls for support personnel as needed to solve the problem. Alternatively support personnel are on call and connect remotely after business hours to fix reported problems. Support personnel may be paged automatically by the application(s) to troubleshoot the problem.

      For example, the Electric Reliability Council of Texas (ERCOT) requires the following state estimator performance measures [24]:

      1 State Estimator to converge 97% of runs during a one‐month period.

      2 On transmission elements identified as causing 80% of congestion cost in the latest year for which data is available, the residual difference between State Estimator results and Power Flow results for critically monitored transmission element MW flows are required to be less than 3% of the associated element emergency rating on at least 95% of samples measured in a one‐month period.

      3 On transmission elements identified as causing 80% of congestion cost in the latest year for which data is available, the difference between the MW telemetry value and the MW State Estimator value shall be less than 3% of the associated element emergency rating on at least 95% of samples measured in a one‐month period.

      4 On