Optimal deep learning based aggregation of TCLs in an inverter fed stand-alone microgrid for voltage unbalance mitigation

Abstract

The rapid growth of distributed generating (DG) units due to its advancement in modern day power systems has put a major effect in the operation of low inertial stand-alone Microgrids (MG). Moreover, the limited availability of ancillary services pose a severe challenge in power quality issues especially during unbalanced loading condition. To perpetuate its resiliency, this paper presents an inverter fed Space Vector based Droop controlled Positive-Negative sequence compensation along with deep learning-based aggregation of Thermostatic loads for incorporating Demand Side Management (DSM) to mitigate voltage unbalance (VU) adhering to IEC-61000-3-13 and without compromising the thermal comfort of customers. The major portion of Thermostatic Control Loads (TCLs) comprise of air conditioners that occupy a wide sector in residential and commercial areas; hence the aggregation and scheduling of such loads re-establishes dynamic reactive power of DG by alleviating VU during peak hours. The novel one-time registration of loads is also introduced in this paper by incorporating Artificial Neural Network (ANN) for easy access to customer's newly connected loads to participate in DSM. The test system comprises of eight bus system connected with Solar Photovoltaic, Battery unit, Slow responding Diesel Generator (DZ) along with single and three phase loads constituting of TCLs. The performance of the proposed VU mitigation control strategy is evaluated by considering various case studies and the results are validated through simulations performed in MATLAB / Simulink environment.