If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact us at email@example.com.
Bridging Machine Learning for Smart Grid Applications
AltmetricsView Usage Statistics
This dissertation proposes to develop, leverage, and apply machine learning algorithms on various smart grid applications including state estimation, false data injection attack detection, and reliability evaluation. The dissertation is divided into four parts as follows.. Part I: Power system state estimation (PSSE). The PSSE is commonly formulated as a weighted least-square (WLS) algorithm and solved using iterative methods such as Gauss-Newton methods. However, iterative methods have become more sensitive to system operating conditions than ever before due to the deployment of intermittent renewable energy sources, zero-emission technologies (e.g., electric vehicles), and demand response programs. Efficient approaches for PSSE are required to avoid pitfalls of the WLS-based PSSE computations for accurate prediction of operating conditions. The first part of this dissertation develops a data-driven real-time PSSE using a deep ensemble learning algorithm. In the proposed approach, the ensemble learning setup is formulated with dense residual neural networks as base-learners and a multivariate-linear regressor as a meta-learner. Historical measurements and states are utilized to train and test the model. The trained model can be used in real-time to estimate power system states (voltage magnitudes and phase angles) using real-time measurements. Most of current data-driven PSSE methods assume the availability of a complete set of measurements, which may not be the case in real power system data acquisition. This work adopts multivariate linear regression to forecast system states for instants of missing measurements to assist the proposed PSSE technique. Case studies are performed on various IEEE standard benchmark systems to validate the proposed approach. Part II: Cyber-attacks on Voltage Regulation. Several wired and wireless advanced communication technologies have been used for coordinated voltage regulation schemes in distribution systems. These technologies have been employed to both receive voltage measurements from field sensors and transmit control settings to voltage regulating devices (VRDs). Communication networks for voltage regulation can be susceptible to data falsification attacks, which can lead to voltage instability. In this context, an attacker can alter multiple field measurements in a coordinated manner to disturb voltage control algorithms. The second part of this dissertation develops a machine learning-based two-stage approach to detect, locate, and distinguish coordinated data falsification attacks on control systems of coordinated voltage regulation schemes in distribution systems with distributed generators. In the first stage (regression), historical voltage measurements along with current meteorological data (solar irradiance and ambient temperature) are provided to random forest regressor to forecast voltage magnitudes of a given current state. In the second stage, a logistic regression compares the forecasted voltage with the measured voltage (used to set VRDs) to detect, locate, and distinguish coordinated data falsification attacks in real-time. The proposed approach is validated through several case studies on a 240-node real distribution system (based in the USA) and the standard IEEE 123-node benchmark distribution system.Part III: Cyber-attacks on Distributed Generators. Part III of the dissertation proposes a deep learning-based multi-label classification approach to detect coordinated and simultaneously launched data falsification attacks on a large number of distributed generators (DGs). The proposed approach is developed to detect power output manipulation and falsification attacks on DGs including additive attacks, deductive attacks, and combination of additive and deductive attacks (attackers use the combination of additive and deductive attacks to camouflage their attacks). The proposed approach is demonstrated on several systems including the 240-node and IEEE 123-node distribution test system. Part IV: Composite System Reliability Evaluation. Traditional composite system reliability evaluation is computationally demanding and may become inapplicable to large integrated power grids due to the requirements of repetitively solving optimal power flow (OPF) for a large number of system states. Machine learning-based approaches have been used to avoid solving OPF in composite system reliability evaluation except in the training stage. However, current approaches have been utilized only to classify system states into success and failure states (i.e., up or down). In other words, they can be used to evaluate power system probability and frequency reliability indices, but they cannot be used to evaluate power and energy reliability indices unless OPF is solved for each failure state to determine minimum load curtailments. In the fourth part of this dissertation, a convolutional neural network (CNN)-based regression approach is proposed to determine the minimum amount of load curtailments of sampled states without solving OPF. Unavoidable load curtailments due to failures are then used to evaluate power and energy indices (e.g., expected demand not supplied) as well as to evaluate the probability and frequency indices. The proposed approach is applied on several systems including the IEEE Reliability Test System and Saskatchewan Power Corporation in Canada.