The high uncertainty of renewables introduces more dynamics to power systems. The conventional way of monitoring and controlling power systems is no longer reliable. New strategies are needed to ensure the stability and reliability of power systems. This work aims to assess the use of machine learning methods in analyzing data from renewable integrated power systems to aid the decisionmaking of electricity market participants. Specifically, the work studies the cases of electricity price forecast, solar panel detection, and how to constrain the machine learning methods to obey domain knowledge.Chapter 2 proposes to diversify the data source to ensure a more accurate electricity price forecast. Specifically, the proposed two-stage method, namely the rerouted method, learns two types of mapping rules: the mapping between the historical wind power and the historical price and the forecasting rule for wind generation. Based on the two rules, we forecast the price via the forecasted generation and the learned mapping between power and price. The massive numerical comparison gives guidance for choosing proper machine learning methods and proves the effectiveness of the proposed method. Chapter 3 proposes to integrate advanced data compression techniques into machine learning algorithms to either improve the predicting accuracy or accelerate the computation speed. New semi-supervised learning and one-class classification methods are proposed based on autoencoders to compress the data while refining the nonlinear data representation of human behavior and solar behavior. The numerical results show robust detection accuracy, laying down the foundation for managing distributed energy resources in distribution grids. Guidance is also provided to determine the proper machine learning methods for the solar detection problem. Chapter 4 proposes to integrate different types of domain knowledge-based constraints into basic neural networks to guide the model selection and enhance interpretability. A hybrid model is proposed to penalize derivatives and alter the structure to improve the performance of a neural network. We verify the performance improvement of introducing prior knowledge-based constraints on both synthetic and real data sets.

Power systems are transforming into more complex and stressed systems each day. These stressed conditions could lead to a slow decline in the power grid's voltage profile and sometimes lead to a partial or total blackout. This phenomenon can be identified by either solving a power flow problem or using measurement-based real-time monitoring algorithms. The first part of this thesis focuses on proposing a robust power flow algorithm for ill-conditioned systems. While preserving the stable nature of the fixed point (FP) method, a novel distributed FP equation is proposed to calculate the voltage at each bus. The proposed algorithm's performance is compared with existing methods, showing that the proposed method can correctly find the solutions when other methods cannot work due to high condition number matrices. It is also empirically shown that the FP algorithm is more robust to bad initialization points. The second part of this thesis focuses on identifying the voltage instability phenomenon using real-time monitoring algorithms. This work proposes a novel distributed measurement-based monitoring technique called voltage stability index (VSI). With the help of PMUs and communication of voltage phasors between neighboring buses, the processors embedded at each bus in the smart grid perform simultaneous online computations of VSI. VSI enables real-time identification of the system's critical bus with minimal communication infrastructure. Its benefits include interpretability, fast computation, and low sensitivity to noisy measurements. Furthermore, this work proposes the ``local static-voltage stability index" (LS-VSI) that removes the minimal communication requirement in VSI by requiring only one PMU at the bus of interest. LS-VSI also solves the issue of Thevenin equivalent parameter estimation in the presence of noisy measurements. Unlike VSI, LS-VSI incorporates the ZIP load models and load tap changers (LTCs) and successfully identifies the bifurcation point considering ZIP loads' impact on voltage stability. Both VSI and LS-VSI are useful to monitor the voltage stability margins in real-time using the PMU measurements from the field. However, they cannot indicate the onset of voltage emergency situations. The proposed LD-VSI uses the dynamic measurements of the power system to identify the onset of a voltage emergency situation with an alarm. Compared to existing methods, it is shown that it is more robust to PMU measurement noise and can also identify the voltage collapse point while the existing methods have issues with the same.

Due to the large scale of power systems, latency uncertainty in communication can cause severe problems in wide-area measurement systems. To resolve the issue, a significant amount of past work focuses on using emerging technologywhich is machine learning methods such as Q-learning to address latency issues in modern controls. Although such a method can deal with the stochastic characteristics of communication latency in the long run, the Q-learning methods tend to overestimate Q-values, leading to high bias. To solve the overestimation bias issue, the learning structure is redesigned with a twin-delayed deep deterministic policy gradient algorithm to handle the damping control issue under unknown latency in the power network. Meanwhile, a new reward function is proposed, taking into account the machine speed deviation, the episode termination prevention, and the feedback from action space. In this way, the system optimally damps down frequency oscillations while maintaining the system’s stability and reliable operation within defined limits. The simulation results verify the proposed algorithm in various perspectives including the latency sensitivity analysis under high renewable energy penetration and the comparison with other machine learning algorithms. For example, if the proposed twin-delayed deep deterministic policy gradient algorithm is applied, the low-frequency oscillation significantly improved compared to existing algorithms. Furthermore, under the mentorship of Dr. Yang Weng, the development of a big data analysis software project has been collaborating with the Salt River Project (SRP), a major power utility in Arizona. After a thorough examination of data for the project, it is examined that SRP is suffering from a lot of smart meters data issues. An important goal of the project is to design big data software to monitor SRP smart meter data and to present indicators of abnormalities and special events. Currently, the big data software interface has been developed for SRP, which has already been successfully adopted by other utilities, research institutes, and laboratories as well.

The power-flow problem has been solved using the Newton-Raphson and Gauss-Seidel methods. Recently the holomorphic embedding method (HEM), a recursive (non-iterative) method applied to solving nonlinear algebraic systems, was applied to the power-flow problem. HEM has been claimed to have superior properties when compared to the Newton-Raphson and other iterative methods in the sense that if the power-flow solution exists, it is guaranteed that a properly configured HEM can find the high voltage solution and, if no solution exists, HEM will signal that unequivocally. Provided a solution exists, convergence of HEM in the extremal domain is claimed to be theoretically guaranteed by Stahl’s convergence-in-capacity theorem, another advantage over other iterative nonlinear solver.In this work it is shown that the poles and zeros of the rational function from fitting the local PMU measurements can be used theoretically to predict the voltage-collapse point. Different numerical methods were applied to improve prediction accuracy when measurement noise is present. It is also shown in this work that the dc optimal power flow (DCOPF) problem can be formulated as a properly embedded set of algebraic equations. Consequently, HEM may also be used to advantage on the DCOPF problem. For the systems examined, the HEM-based interior-point approach can be used to solve the DCOPF problem. While the ultimate goal of this line of research is to solve the ac OPF; tackled in this work, is a precursor and well-known problem with Padé approximants: spurious poles that are generated when calculating the Padé approximant may, at times, prevent convergence within the functions domain. A new method for calculating the Padé approximant, called the Padé Matrix Pencil Method was developed to solve the spurious pole problem. The Padé Matrix Pencil Method can achieve accuracy equal to that of the so-called direct method for calculating Padé approximants of the voltage-functions tested while both using a reduced order approximant and eliminating any spurious poles within the portion of the function’s domain of interest: the real axis of the complex plane up to the saddle-node bifurcation point.

Nowadays, the widespread use of distributed generators (DGs) raises significant challenges for the design, planning, and operation of power systems. To avoid the harm caused by excessive DGs, evaluating the reliability and sustainability of the system with high penetration of DGs is essential. The concept of hosting capacity (HC) is used to achieve this purpose. It is to assess the capability of a distribution grid to accommodate DGs without causing damage or updating facilities. To obtain the HC value, traditional HC analysis methods face many problems, including the computational difficulties caused by the large-scale simulations and calculations, lacking the considering temporal correlation from data to data, and the inefficient on real-time analysis. This paper proposes a machine learning-based method, the Spatial-Temporal Long Short-Term Memory (ST-LSTM), to overcome these drawbacks using the traditional HC analysis method. This method will significantly reduce the requirement of calculations and simulations, and obtain HC results in real-time. Using the time-series load profiles and the longest path method, ST-LSTMs can capture the temporal information and spatial information respectively. Moreover, compared with the basic Long Short-Term Memory (LSTM) model, this modified model will improve the performance in the HC analysis by some specific designs, which are the sensitivity gate to consider voltage sensitivity information, the dual forget gates to build spatial and temporal correlation.

As the field of machine learning increasingly provides real value to power system operations, the availability of rich measurement datasets has become crucial for the development of new applications and technologies. This dissertation focuses on the use of time-series load data for the design of novel data-driven algorithms. Loads are one of the main factors driving the behavior of a power system and they depend on external phenomena which are not captured by traditional simulation tools. Thus, accurate models that capture the fundamental characteristics of time-series load dataare necessary. In the first part of this dissertation, an example of successful application of machine learning algorithms that leverage load data is presented. Prior work has shown that power systems energy management systems are vulnerable to false data injection attacks against state estimation. Here, a data-driven approach for the detection and localization of such attacks is proposed. The detector uses historical data to learn the normal behavior of the loads in a system and subsequently identify if any of the real-time observed measurements are being manipulated by an attacker. The second part of this work focuses on the design of generative models for time-series load data. Two separate techniques are used to learn load behaviors from real datasets and exploiting them to generate realistic synthetic data. The first approach is based on principal component analysis (PCA), which is used to extract common temporal patterns from real data. The second method leverages conditional generative adversarial networks (cGANs) and it overcomes the limitations of the PCA-based model while providing greater and more nuanced control on the generation of specific types of load profiles. Finally, these two classes of models are combined in a multi-resolution generative scheme which is capable of producing any amount of time-series load data at any sampling resolution, for lengths ranging from a few seconds to years.

Modular multilevel converters (MMCs) have become an attractive technology for high power applications. One of the main challenges associated with control and operation of the MMC-based systems is to smoothly precharge submodule (SM) capacitors to the nominal voltage during the startup process. The existing closed-loop methods require additional effort to analyze the small-signal model of MMC and tune control parameters. The existing open-loop methods require auxiliary voltage sources to charge SM capacitors, which add to the system complexity and cost. A generalized precharging strategy is proposed in this thesis.

For large-scale MMC-embedded power systems, it is required to investigate dynamic performance, fault characteristics, and stability. Modeling of the MMC is one of the challenges associated with the study of large-scale MMC-based power systems. The existing models of MMC did not consider the various configurations of SMs and different operating conditions. An improved equivalent circuit model is proposed in this thesis.

The solid state transformer (SST) has been investigated for the distribution systems to reduce the volume and weight of power transformer. Recently, the MMC is employed into the SST due to its salient features. For design and control of the MMC-based SST, its operational principles are comprehensively analyzed. Based on the analysis, its mathematical model is developed for evaluating steady-state performances. For optimal design of the MMC-based SST, the mathematical model is modified by considering circuit parameters.

One of the challenges of the MMC-based SST is the balancing of capacitor voltages. The performances of various voltage balancing algorithms and different modulation methods have not been comprehensively evaluated. In this thesis, the performances of different voltage-balancing algorithms and modulation methods are analyzed and evaluated. Based on the analysis, two improved voltage-balancing algorithms are proposed in this thesis.

For design of the MMC-based SST, existing references only focus on optimal design of medium-frequency transformer (MFT). In this thesis, an optimal design procedure is developed for the MMC under medium-frequency operation based on the mathematical model of the MMC-based SST. The design performance of MMC is comprehensively evaluated based on free system parameters.

Transmission line capacity is an obstacle for the utilities because there is a load increment annually, and new power plants are being connected, which requires an update. Energy router (ER) is a device that provides an additional degree of freedom to the utilities by controlling the reactive power. The ER reactive power injection is demonstrated by changing the line's reactance value to increase its capacity and give the utility a deferral time for the project upgrade date. Changing the reactance manually and attaching Smart Wire's device to the branches have effectively solved the overload in three locations of a local utility in Arizona (LUA) system.

Furthermore, electric vehicle charging stations (EVCSs) have been increasing to meet EV needs, which calls for an optimal planning model to maximize the profits. The model must consider both the transportation and power systems to avoid damages and costly operation. Instead of coupling the transportation and power systems, EVCS records have been analyzed to fill the gap of EV demand. For example, by accessing charging station records, the moment knowledge of EV demand, especially in the lower order, can be found. Theoretically, the obtained low-order moment knowledge of EV demand is equivalent to a second-order cone constraint, which is proved. Based on such characteristics, a chance-constrained (CC) stochastic integer program for the planning problem is formulated. For planning EV charging stations with ER, this method develops a simple ER model to investigate the interaction between the mobile placement of power flow controller and the daily pattern of EV power demand.

With the rapid advancement in the technologies related to renewable energies such

as solar, wind, fuel cell, and many more, there is a definite need for new power con

verting methods involving data-driven methodology. Having adequate information is

crucial for any innovative ideas to fructify; accordingly, moving away from traditional

methodologies is the most practical way of giving birth to new ideas. While working

on a DC-DC buck converter, the input voltages considered for running the simulations

are varied for research purposes. The critical aspect of the new data-driven method

ology is to propose a machine learning algorithm. In this design, solving for inductor

value and power switching losses, the parameters can be achieved while keeping the

input and output ratio close to the value as necessary. Thus, implementing machine

learning algorithms with the traditional design of a non-isolated buck converter deter

mines the optimal outcome for the inductor value and power loss, which is achieved

by assimilating a DC-DC converter and data-driven methodology.

The present thesis investigates the different outcomes from machine learning al

gorithms in comparison with the dynamic equations. Specifically, the DC-DC buck

converter will be focused on the thesis. In order to determine the most effective way

of keeping the system in a steady-state, different circuit buck converter with different

parameters have been performed.

At present, artificial intelligence plays a vital role in power system control and

theory. Consequently, in this thesis, the approximation error estimation has been

analyzed in a DC-DC buck converter model, with specific consideration of machine

learning algorithms tools that can help detect and calculate the difference in terms

of error. These tools, called models, are used to analyze the collected data. In the

present thesis, a focus on such models as K-nearest neighbors (K-NN), specifically

the Weighted-nearest neighbor (WKNN), is utilized for machine learning algorithm

purposes. The machine learning concept introduced in the present thesis lays down

the foundation for future research in this area so that to enable further research on

efficient ways to improve power electronic devices with reduced power switching losses

and optimal inductor values.

Due to the rapid penetration of solar power systems in residential areas, there has

been a dramatic increase in bidirectional power flow. Such a phenomenon of bidirectional

power flow creates a need to know where Photovoltaic (PV) systems are

located, what their quantity is, and how much they generate. However, significant

challenges exist for accurate solar panel detection, capacity quantification,

and generation estimation by employing existing methods, because of the limited

labeled ground truth and relatively poor performance for direct supervised learning.

To mitigate these issue, this thesis revolutionizes key learning concepts to (1)

largely increase the volume of training data set and expand the labelled data set by

creating highly realistic solar panel images, (2) boost detection and quantification

learning through physical knowledge and (3) greatly enhance the generation estimation

capability by utilizing effective features and neighboring generation patterns.

These techniques not only reshape the machine learning methods in the GIS

domain but also provides a highly accurate solution to gain a better understanding

of distribution networks with high PV penetration. The numerical

validation and performance evaluation establishes the high accuracy and scalability

of the proposed methodologies on the existing solar power systems in the

Southwest region of the United States of America. The distribution and transmission

networks both have primitive control methodologies, but now is the high time

to work out intelligent control schemes based on reinforcement learning and show

that they can not only perform well but also have the ability to adapt to the changing

environments. This thesis proposes a sequence task-based learning method to

create an agent that can learn to come up with the best action set that can overcome

the issues of transient over-voltage.