PARMENIDES – ideal voltage sensory placement for battery storage operation optimization

To optimize the operation of a low-voltage distribution grid, battery storage systems can be efficiently managed at an optimal point by using a grid capacity management system. However, this requires more detailed information about the state of a grid than is usually available. This gap can be bridged by installing additional sensors, and the number of sensors can be minimized through innovative state estimation approaches. The presented approach allows the training of Machine Learning (ML) models capable of predicting the grid state when provided with operational data obtained through load flow calculations from the same grid. The models are trained to predict unknown voltage values based on assumed known data. The conducted experiments aimed at identifying the optimal locations for these voltage measurements to determine the ideal placement for sensory equipment in the grid and assess the estimation error. The results revealed that for both grids specific sensor placement can be identified where models perform very well with a low error and demonstrate a high level of prediction accuracy. Moreover, statements about the ideal placement of the sensors were derived from experiments.