Jakov Krstulovic

Reconstructing Missing Data in State Estimation With Autoencoders

This paper presents the proof of concept for a new solution to the problem of recomposing missing information at the SCADA of energy/distribution management systems (EMS/DMS), through the use of offline trained autoencoders. These are neural networks with a special architecture, which allows them to store knowledge about a system in a nonlinear manifold characterized by their weights. Suitable algorithms may then recompose missing inputs (measurements). The paper shows that, trained with adequate information, autoencoders perform well in recomposing missing voltage and power values, and focuses on the particularly important application of inferring the topology of the network when information about switch status is absent. Examples with the IEEE RTS 24-bus network are presented to illustrate the concept and technique.

Towards an Auto-Associative Topology State Estimator

This paper presents a model for breaker status identification and power system topology estimation based on a mosaic of local auto-associative neural networks. The approach extracts information from values of the analog electric variables and allows the recovery of missing sensor signals or the correction of erroneous data about breaker status. The results are confirmed by extensive tests conducted on an IEEE benchmark network.

State estimation in distribution smart grids using autoencoders

This work proposes an innovative method based on autoencoders to perform state estimation in distribution grids, which has as main advantage the fact of being independent of the network parameters and topology. The method was tested in a real low voltage grid (incorporating smart grid features), under different scenarios of smart meter deployment. Simulations were performed in order to understand the necessary requirements for an accurate distribution grid state estimator and to evaluate the performance of a state estimator based on autoencoders.

Wind power plant as ancillary service provider

Rapid growth of wind power sector presents great challenge for power system operators in aspect of generation scheduling, grid management, balancing and ancillary services. Traditionally ancillary services are obtained from conventional power plants. However nowadays, through different set of control possibilities, wind power plants are able to partly participate in provision of ancillary services. This paper discusses the possibilities of providing ancillary system services by wind power plants, primarily regarding reactive power control and partially frequency control. In addition, real example is used to demonstrate the benefits from reactive power support by wind power plant to maintain regular voltage conditions in distribution network.

Denoising auto-associative measurement screening and repairing

This paper offers an efficient and robust concept for a decentralized bad data processing, able to supply in real-time a power system state estimator with a repaired measurement set. Corrupted measurement vectors are funneled through a denoising auto-associative neural network in order to project the biased vector back to the data manifold learned during an offline training process. In order to improve accuracy, a maximum similarity with the solution manifold, measured with Correntropy, is searched for by a meta-heuristic. The extreme robustness and scalability of the process is demonstrated in multiple characteristic case studies.

Selection of measurements in topology estimation with mutual information

This paper discusses mechanisms for establishing an efficient decentralized methodology for the reconstruction of topology in power systems. The maximum mutual information criterion is proposed as a selection criterion for the inputs of a distributed topology estimator, based on mosaic of local auto-associative neural networks. The proposed concepts offer some strong theoretical support for an information theoretic perspective on power system state estimation. The results are confirmed by extensive tests conducted on the IEEE RTS 24-bus system.

Wind power plant capacity allocation in areas with limited transmission capacity

This paper proposes a simulation framework for assessing the possibility of wind power plant integration into the existing transmission network in an optimal manner that takes into account the wind potential at each location scheduled for WPP construction. Wind power plant capacity allocation is analyzed from the static aspect considering maximum line loading under normal system operating conditions with WPP production modeled as stochastic variable. Primal-dual interior point method with a preprocessing phase which eliminates redundant constraints is employed to solve the optimization problem.