Sustainable Development Goals
Abstract/Objectives
A framework for distributed frequency control and intrusion detection in isolated microgrids (MGs) is presented in this work. First, a distributed secondary control for isolated MGs, consisting of the local droop control at the primary level and a distributed node-to-node update at the secondary level, is proposed for achieving a proportional power sharing. By casting it as a consensus optimization problem, we adopt the partial primal– dual algorithm for a totally distributed update requiring only neighboring information exchange. Attack models as well as countermeasures for malicious attacks on the communication network are also investigated. Two types of malicious attacks on the communication network, including link and node attacks, are studied. Model-based anomaly detection and localization strategies are developed by exploring the dual variable-related metrics. Numerical experiments have been performed to demonstrate the effectiveness of the proposed control and countermeasure metrics.
Results/Contributions

The full MG network dynamics, including power flow dynamics and droop controlled distributed interface converters (DICs), are considered. The problem of minimizing the voltage error and ensuring a proportional power sharing operation simultaneously is formulated as a consensus optimization problem. Assuming connected communication graph among DICs, we adopt the partial primal-dual (PPD) algorithm to solve the steady-state problem in closed-form. Main contributions of this work are two-fold:

Firstly, parts of the proposed update rules boil down to network power flow dynamics and thus are seamlessly implemented by the physical system itself. Hence, the proposed control design only requires the exchange of a few variables, while its stability follows directly from the PPD algorithm. Distinct from most of the previous work, the proposed control design can guarantee the MG stability through the selection of optimization step size. Secondly, the PPD-based design with localized dual variable information can be further utilized to improve the capability to attack detection. Earlier attack detection work for general distributed consensus methods typically requires system-wide information collection and accordingly has a very high computational burden. To overcome these limitations, we have developed the metrics for detection and identification by using local physical measurements and neighboring dual variable information. The centralized energy management system (EMS) collects all this information to decide the attack scenarios, as motivated by standards on cyber networks for integrating DER into power systems, and also by practical work studying cybersecurity framework in supervisory control and data acquisition systems (SCADA) of modern power systems. Different layers of protection schemes for managing cyber/physical security have been considered to evaluate intrusion probability for preventing possible cyber-attack. Accordingly, the proposed implementation is very scalable. Compared with previous work where the cyber-security of the MG has not been addressed, we have provided analytical understanding and mitigation strategies for cyber intrusions.

Keywords
microgriddistributed droop controlpartial primal-dual algorithmmalicious attackcyber-security
Contact Information
朱家齊
ccchu@ee.nthu.edu.tw