Security Impact Assessment of Active Distributed Network (ADN) on the Future Grid Operations
Fossil fuel-based power generation contributes significantly to global warming. To decarbonize the electric power generation, a significant amount of renewable energy sources (RES) based-distributed generation (DG) units will be connected to the grid. High penetration of RES-DG units connected at different grid levels will provide technical merits if properly planned and operated. The concern, however, is how to guarantee the security of the grid with high penetration of RES-DG, considering the characteristics of the power grid and the behaviours of the connected RES-DG units under both steady-state and transient conditions. Penetration of DG units creates an active distribution system (ADN) with several impacts within the ADN and the transmission-distribution network boundary nodes. Variability of the grid load levels and power generation from RES-DG units will also impact the grid security. Voltage fluctuations, reverse power flow, power intermittency, and harmonics increase are some of the effects of high penetration of RES-DG units on the distribution network. The absence of mechanical inertia support from the RES-DG units during grid disturbance reinforces the need to be concerned about the penetration level. It is therefore important to understand the specific impact of high penetration of RES-DG units on the security of the different levels of the grid. The knowledge obtained will help operators plan toward increasing the penetration level of RES-DG units and apply appropriate measures and suitable control schemes to mitigate arising stability issues due to the high penetration.
This PhD research aims to assess the security impact of high penetration of RES-DG units on modern grid operations. The security assessment and proposed models border three key integrated grid sections: the transmission network, the distribution network, and the transmission-distribution network boundary. The results of investigating the impacts of RES-DG unit operations on the identified grid sections are presented in separate chapters. Two chapters are dedicated to the offline and online assessment and prediction of the security of the transmission grid.
Chapter 1 provides the background of this thesis. Also, it contains the research gaps, objectives, and specific contributions of the thesis. Chapter 2 provides a comprehensive literature review of the security impact of active distribution networks on modern grid operation. Chapter 2 also reviews the technical impacts of RES-DG penetration into the grid and discusses the role of renewable energy source generation in the grid transition. It also presents a comprehensive review of proposed techniques to limit the security impacts on the three identified grid sections. Lastly, it identifies several research gaps and drawbacks that eventually formed part of the research question in this thesis.
The proposed methods start with the optimal placement of RES-DG units in a distribution network in Chapter 3. The method in Chapter 3 involves developing a decision tree classification technique to optimize the network's security indices under the nodal hosting capacity and PVDG capacity constraints. The security indices considered in this paper are the branches' risk index (RI) and power loss (PL). The proposed method is a robust machine learning approach with deterministic and probabilistic security indices to determine the optimal placement of PVDG units. Chapter 3 also analyses the impacts of the different PVDG units' voltage control modes on the performance of the distribution network.
Chapter 4 presents a method to assess, classify and predict the security of the integrated grid using varying grid parameters with an offline machine learning approach. Firstly, an adaptive neuro-fuzzy inference system (ANFIS) suitable for real-time applications is developed to predict the critical clearing time considering varying load levels and grid inertia. Then, a continuous transient stability assessment technique is developed to generate the training dataset. A density-based clustering via classification approach is proposed to label the dataset. Finally, a predictive model was obtained from the labelled dataset using a Naïve-Bayes (NB) probabilistic classifier. The advantage of the proposed technique in this chapter lies in considering the impact of the penetration of RES-DG units on the grid’s security and the wide range of contingencies considered. Also, prior knowledge of the grid’s security state is not required with the proposed clustering approach.
After establishing the offline security assessment and prediction method in Chapter 4, Chapter 5 presents a framework for online security prediction and control for modern power grids. System operators must assess the grid's security for specific levels of inertia, load, RES-DG penetration, and fault location to plan, control, and operate the network securely. Unlike offline methods, an online security prediction framework can establish potential grid security in a reasonable time ahead. Thus, this chapter proposes an incremental machine learning training technique and intelligent security control system to ensure the grid's security. An incremental Naïve Bayes algorithm is applied to the training dataset developed from the responses of the grid to transient stability simulations. The intelligent security control system consists of a Gaussian process regression-based load shed value estimator. The proposed intelligent security control system ensures the grid's security for predicted insecure scenarios by estimating the load shed value to keep the operation of the grid in a security state.
The coordinated operation between the transmission and distribution networks is necessary to ensure the security of the modern grid. Interaction must exist between transmission and distribution systems operators acting independently to maintain grid voltage within the individual networks. The operation of the modern grid involves the utilization of the flexibilities of the active distribution network to achieve voltage supports during grid disturbance. It is therefore vital to ensure security at the transmission-distribution network boundary during flexibility operations. Consequently, Chapter 6 develops a bi-level model (economic and technical objective levels) to achieve optimal flexibility operation. The economic objective is achieved through a robust distribution network reconfiguration technique to minimize network losses. An improved decision tree classification technique is proposed for level two to determine the amount of flexibility in MW for optimal voltage support. The optimization problem in the first level is solved using a fmincon solver algorithm considering deterministic and probabilistic constraints. The proposed algorithm minimizes the power loss and optimizes the voltage support for specific grid operations. The thesis closes with Chapter 7, which presents the conclusion from the PhD research drawn from Chapters 2 to 6 and the future work and prospects to broaden the subjects addressed in this study.