Development of Hierarchical Control Strategies for DC Microgrid Clusters
Loading...
Files
Size: 6.89 MB, File format: Adobe PDF
Date
Authors
AL-Tameemi, Zaid
Supervisor
Lie, Tek Tjing
Zamora, Ramon
Blaabjerg, Frede
Item type
Degree name
Journal Title
Journal ISSN
Volume Title
Publisher
Auckland University of Technology
Abstract
DC Microgrids (MGs) have attracted considerable research interest and are expected to play an important role in the development of power networks in the coming years. They are characterised by their ability to facilitate the straightforward integration of renewable energy sources into the grid. In the event of network failures, DC MGs can operate independently or support the main grid, providing a reliable backup power source. However, these MGs might not be able to meet the local energy requirements during critical operating conditions. Therefore, they can be interconnected to form clusters, thereby greatly reducing the operating stress on each MG and enhancing the network's lifespan, reliability, and overall efficiency. These interconnected microgrids need to be properly coordinated within a cluster to efficiently handle disturbances that may occur in each microgrid. In the literature, various control strategies, including centralised, decentralised, distributed, and hierarchical schemes, have been utilised to coordinate MGs in a cluster. The main objectives of adopting these control schemes are to achieve efficient power sharing and voltage regulation. However, the increasing complexity of modern power networks and the unpredictability of load demand impose the development of advanced control strategies rather than centralised, decentralised, or distributed control schemes. In this regard, a hierarchical control scheme, which includes three control layers, primary, secondary, and tertiary, and is characterised by high dependability and efficient operation, has been widely adopted in the literature to coordinate DC microgrids in a cluster. In the tertiary control layer, coordination can be realised through time-triggered and event-triggered consensus algorithms. Time-triggered control approaches, which rely on periodic communication, are proposed to achieve the primary control objectives of DC microgrids, including accurate current sharing and optimal voltage regulation; however, they fail to reduce the communication burden, as unnecessary communication occurs among microgrid controllers during steady-state operation. To address such issues, event-triggered control schemes have been proposed to enhance communication efficiency by transmitting information only when required. However, the adopted ETC schemes still face several challenges that need to be addressed to maintain proper coordination among microgrids during steady-state conditions and disturbances.
It is found that the triggering functions used in prior studies are state-dependent and may be overly sensitive to changes in state values, thereby leading to redundant data exchange in microgrids and increasing the likelihood of Zeno behaviour. Furthermore, control coefficients of these functions are set to fixed values, which may limit their adaptability to sudden changes in a cluster. Although adopting fixed-time consensus schemes can improve convergence rate and voltage recovery, prior studies have not considered adaptive selection of their control coefficients based on operating conditions. Therefore, the inadequate tuning of such coefficients may result in slow responses and oscillations under unpredictable operating conditions. In secondary control schemes, PI controllers have been extensively used. Furthermore, most previous studies extensively use classical PI controllers without considering the importance of optimally selecting these parameters for improving the cluster's dynamic response during critical operating conditions. Also, complex mathematical derivations have been adopted in prior studies to design event-triggering functions with constant coefficients, thereby increasing the implementation complexity of ETC schemes in real-life scenarios and unnecessarily burdening communication links.
To address these issues, this thesis proposes a robust control scheme that includes significant advancements in the secondary and tertiary control layers. In this regard, a new distributed event-triggered consensus algorithm (DETC) is proposed to decrease triggering instances while ensuring a positive interevent time to overcome Zeno behaviour. Although this control scheme has been further developed by using a saturation function and a time-dependent term, the control parameters of the proposed triggering function used in this stage are fixed, which may limit adaptability to sudden state changes and cause it to respond to insignificant errors. To address this, an adaptive mixed time-state-dependent event-triggered consensus protocol (AMDETC) is adopted to reduce the sensitivity of the previous event-triggering functions to minor errors by allowing the control coefficients to adapt to the cluster’s operating conditions. This contributes to 100% accuracy in identifying triggering instances, rapid current convergence within 0.02 s, and fast voltage recovery across different operating conditions.
Furthermore, the secondary control layer is significantly developed in this thesis to address its identified weakness in prior studies by adaptively selecting control parameters in a fixed-time consensus algorithm based on Artificial Neural Networks, which is used to regulate the control output and prevent oscillations within the clusters. Additionally, the Grey Wolf Optimiser algorithm is used to adjust the PI controller in the secondary control layer to improve its robustness against critical operating conditions. These enhancements play a vital role in improving current convergence and voltage recovery during disturbances. Based on these featured results regarding current convergence, voltage recovery, and data exchange, the main targeted objectives of this PhD thesis have been successfully achieved; however, implementation complexity remains to be addressed at this stage. Therefore, in the final stage, an artificial neural network is introduced as an alternative mechanism to reduce dependence on the mathematical derivations of the previous triggering functions for coordinating data exchange among microgrids in a cluster, achieving 100% accuracy in event detection. The proposed control schemes are tested in a four-DC MG ring cluster using MATLAB and validated with the OPAL-RT tool to assess their applicability in real-time scenarios.