Network Observability and Pseudo Measurements:
A minimum amount of data is necessary for State Estimation (SE) to be effective. A more analytical way of determining whether a given data is enough for SE is called observability analysis. It forms an integral part of any real time state estimator.The ability to perform state estimation depends on whether sufficient measurements are well distributed throughout the system. When sufficient measurements are available SE can obtain the state vector of the whole system. In this case the network is observable. As explained earlier that this is true when the rank of measurement Jacobian matrix is equal to the number of unknown state variables. The rank of the measurement Jacobian matrix is dependent on the locations and types of available measurements as well as on the network topology. An auxiliary problem in state estimation is where to add additional data or pseudo measurements to a power system in order to improve the accuracy of the calculated state i.e. to improve observability.
The additional measurements represent a cost for the physical transducers, remote terminal or telemetry system, and software data processing in the central computer. Selection of pseudo measurements, filling of missing data, providing appropriate weightage are the functions of the observability analysis algorithm.
Observability can be checked during factorization. If any pivot becomes very small or zero during factorization, the gain matrix may be singular, and the system may not be observable.
To find the value of an injection without measuring it, we must know the power system beyond the measurements currently being made. For example, we normally know the generated MWs and MVARs at generators through telemetry channels (i.e. these measurements would generally be known to the state estimator). If these channels are out, we can perhaps communicate with the operators in the plant control room by telephone and ask for these values and enter them manually. Similarly, if we require a load-bus MW and MVAR for a pseudo measurement, we could use past records that show the relationship between an individual load and the total system load.
We can estimate the total system load quite accurately by finding the total power being generated and estimating the line losses. Further, if we have just had a telemetry failure, we could use the most recently estimated values from the estimator (assuming that it is run periodically) as pseudo measurements. Thus, if required, we can give the state estimator with a reasonable value to use as a pseudo measurement at any bus in the system.
Pseudo measurements increase the data redundancy of SE. If this approach is adapted, care must be taken in assigning weights to various types of measurements. Techniques that can be used to determine the meter or pseudo measurement locations for obtaining a complete observability of the system are available. A review of the principal observability analysis and meter placement algorithms.