Energy market participants' optimal bidding strategies and profit maximization is built upon accurate locational marginal price (LMP) predictions. In wholesale electricity markets, LMPs are strongly spatio-temporal correlated. Without access to confidential information on system topology, model parameters, or market operation conditions, market participants can only accept data-driven methods to utilize publicly available market data to predict LMPs. Most previous data-driven studies on LMP forecasting only leveraged temporal correlations among historical LMPs, and very few of them learned the spatial correlations to improve forecasting accuracy. In this dissertation, unsupervised data-driven approaches are proposed to predict LMPs in real-world energy markets from market participants' perspective. To take advantage of the spatio-temporal correlations, a general data structure is introduced to organize system-wide heterogeneous market data streams into the format of market data 2-dimensional (2D) arrays and 3-dimensional (3D) tensors. The system-wide LMP prediction problem is formulated as a sequence prediction problem. A generative adversarial network (GAN) based prediction model is adopted to learn the spatio-temporal correlations among historical LMPs preserved in the market data 3D tensors, then predict future system-wide LMPs. Multi-loss functions are introduced to assist the adversarial training procedure. A convolutional long-short-term memory (CLSTM)-based GAN is developed to improve forecasting accuracy. All LMP price components are jointly determined by the interactions between the market clearing process and the generator bidding process. The market participants’ LMP forecasting problem can be formulated as a sequential decision-making model considering the interactive market clearing and generation bidding decision-making processes. The spatio-temporal decision transformer is proposed to learn the underlying sequential decision-making model from historical spatio-temporal market data and forecast LMPs as the future actions of these interactive decision-making processes. A two-stage approach is proposed to incorporate historical generation bids into energy price prediction from market participants' perspective. Historical generation bids are taken as the first stage's output and the second stage's input in the training process. The implicit correlation among locational bids, demands, and energy prices is learned to improve price forecasting accuracy. The proposed approaches are verified through case studies using both real-world and simulated data.

Modeling protection devices is essential for performing accurate stability studies. Modeling all the protection devices in a bulk power system is an intractable task due to the limitations of current stability software, and the difficulty in updating the setting data for thousands of protection devices. One of the critical protection schemes that is not adequately modeled in stability studies is distance relaying. Therefore, this dissertation proposes two different methods for identifying the critical distance relays for any contingency, which are required to be modeled in stability studies. The first method is an iterative analytical algorithm and the second method is an ML-based method. The performances of both the methods are evaluated on the Western Electricity Coordinating Council (WECC) system and the results show that to have an accurate assessment of system behavior, modeling the critical distance suffices, and modeling all the distance relays in not necessary. Furthermore, modeling various generator protective relays in stability studies is also crucial. However, no comprehensive framework has been developed that provides guidelines on proper representation of generator protective relays in stability studies and evaluate their impact on the dynamic response of a system. To fill this gap, this dissertation proposes a comprehensive systematic framework which enables proper representation of generator protective relays in stability studies, thereby increasing the accuracy of these studies. The framework is tested on a particular area of the WECC system and the behaviors of different generator protective relays is evaluated.Finally, this dissertation proposes a comprehensive machine-learning (ML)-based online dynamic security assessment (DSA) method that broaden the concept of online DSA by predicting loss of synchronism (LOS) in generators, and the operation of critical protective relays in a power system. The performance of the method is tested on the WECC system in the presence of different noise levels and missing phasor measurement unit (PMU) data. The results reveal that the method can provide precise and fast predictions and is robust to noise and missing PMU data. Therefore, the method can be reliably used in power systems to enhance situational awareness by providing early warnings of impending problems in the system.

Modular multilevel converter (MMC) has become the most attractive and promising topology for multi-terminal high-voltage direct current (MTDC) transmission system. Currently, the dq controller and droop controller are widely used in MTDC systems. However, dq control needs phase synchronization by the phase-locked loop (PLL) and ignores the MMC inner dynamics, which jeopardizes the power decoupling and system stability. On the other side, inappropriate droop parameters can cause instability due to the complicated dynamics of MTDC systems. Moreover, the estimation of control parameters stability region will be helpful to guarantee safe operation of the MMC-MTDC systems. In this thesis, a generalized model of the MMC-MTDC systems is developed, which is precise to reflect transient dynamics, and applicable for arbitrary dc network topology and transmission line model. Furthermore, a nonlinear phase-unsynchronized power decoupling control for MMC is proposed. It realizes power decoupling without PLL and MMC output power dynamics are designed as second-order inertial systems for convenient parameter determination. Additionally, a nonlinear droop controller with a reference self-correct algorithm is proposed for improving regulation speed, reducing dc voltage deviation, and maintaining stability. For convenient stability analysis, an inequality-constraint-based method is proposed to efficiently estimate parameter stability regions through constructing the inequality constraints of parameters' variation. To verify the proposed methods, 4-terminal and 14-terminal MMC-MTDC systems are employed. A comparison of dynamic responses between the calculations of nonlinear state-space models in MATLAB and the EMT simulations in PSCAD/EMTDC is conducted to demonstrate the accuracy of the developed model. Then, the proposed phase-unsynchronized power decoupling control is verified by four cases in EMT simulations and four cases in the experimental prototype. Meanwhile, comparisons with the dq control are conducted to demonstrate the benefits of the proposed method. Furthermore, the zero dynamic stability is investigated and the influences of system parameters on stability are analyzed. For the MTDC control, the performance of the proposed nonlinear droop control is validated in the EMT simulations. At last, the effectiveness of the proposed estimation method of parameter stability regions is demonstrated by several examinations including the supremum tests of droop slopes, the stability region sketches on the accuracy, and the unstable operations with predicted improper droop slopes.