No abstract

The area-wide installation of smart meters in low voltage (LV) grids prospectively provides information about the relevant operational system parameters, e.g. complex node voltages and line loads. Under the condition of neglected house to grid connection lines, a positive local measurement redundancy at every network node is obtainable. In general, this enables the implementation of special three-phase LV state estimation (SE) systems with the ability of bad data detection. In the future, such SE systems might be the basis for closed-loop network control systems without any operator interventions. This paper proposes a concept for a LV state estimation system based on smart meter data. In contrast to other approaches, a linear SE algorithm is used, so that the SE system is not prone to convergence issues. Input variables are voltage and current magnitudes as well as active and reactive currents. The bad data detection process is generally based on the well-known method of normalized residuals. To ensure correctly applied network topologies also in meshed networks, a novel algorithm for detecting topology faults is used. The presented results gathered from simulations and a field test are promising, showing appropriate accuracies and bad data detection probabilities especially for voltage magnitude and active current bad data.

No abstract

The area wide usage of smart meters in low voltage grids enables the identification of the three-phase system state with linear state estimation (SE) systems. In order to localize large measurement errors also bad data detection algorithms have to be applied. But as the measurement redundancy is typically small, the probability of bad data detection is usually small, too. This paper proposes a special three-phase SE approach which enables the reliable detection of bad data on the basis of the well-known normalized residuals method. In contrast to other algorithms active and reactive currents as well as absolute current values are used as input data for a linear SE system. Despite the simplicity of the process the results gathered from simulations and a field test are promising, showing appropriate bad data detection probabilities especially for voltage and active current bad data.

No abstract