In power systems, frequency control is essential to ensure stable grid operation. With the increasing share of renewable energy, frequency fluctuations have become more complex and unpredictable. To address this chall...
详细信息
In power systems, frequency control is essential to ensure stable grid operation. With the increasing share of renewable energy, frequency fluctuations have become more complex and unpredictable. To address this challenge, the paper proposes a cooperative control method for power grid frequency based on the Expert-Guided Deep Deterministic Policy Gradient (EGDDPG) algorithm. This method is data-driven and continuously interacts with the environment to adaptively optimize the control strategy. On the one hand, EGDDPG incorporates expert data during training to guide the agent, enabling faster learning of effective strategies and reducing training convergence time. On the other hand, random disturbances are introduced to simulate the uncertainties in the power system, encouraging the agent to learn control strategies across different environmental states. This approach avoids the overfitting issues associated with fixed disturbance training, enhancing the adaptability and robustness of the algorithm. Simulation results show that the proposed EGDDPG algorithm converges faster and demonstrates stronger control capability and adaptability across various scenarios. It effectively reduces frequency deviation fluctuations and overshoot.
The tight production objectives and dynamically evolving conditions within industries necessitate the diligent monitoring and evaluation of control assets' operational health. During the past decades, a lot of dat...
详细信息
ISBN:
(纸本)9798350373981;9798350373974
The tight production objectives and dynamically evolving conditions within industries necessitate the diligent monitoring and evaluation of control assets' operational health. During the past decades, a lot of data-driven methods, using fundamentals of signal processing and process control or machine learning, have been developed to detect performance issues in control loops. One of those, machine learning based methods, has also become popular in recent years. However, the complexity of algorithms used, incapability of predicting more than "good" or "bad" or need for process excitations limits the practical use of these methods in large industrial scale. In this paper, an easy-to-use classifier has been developed which is based on only routine industrial closed-loop data available and common for many controlsystems. The developed classifier is able to classify the control loops as acceptable, aggressive tuning, sluggish tuning, stiction and external disturbance, which account for almost all common and major problems experienced in closed-loop PID controllers. The features calculated and given to the classifier are immensely easy-to-obtain metrics based on histograms of control error, auto-correlation function, and impulse response. The developed classifier has an 88% training and 85% test accuracy. The classifier has also been tested with a set of industrial loops assessed extensively by process control engineers and able to predict the true class of 88% of the loops, with a 3% false negative rate.
A surrogate model that accurately predicts distribution system voltages is crucial for reliable smart grid planning and operation. This letter proposes a fixed-point data-driven surrogate modeling method that employs ...
详细信息
A surrogate model that accurately predicts distribution system voltages is crucial for reliable smart grid planning and operation. This letter proposes a fixed-point data-driven surrogate modeling method that employs a limited dataset to learn the power-voltage relationship of an unbalanced three-phase distribution system. The proposed surrogate model is designed using a fixed-point load-flow equation, and the stochastic gradient descent method with an automatic differentiation technique is employed to update the parameters of the surrogate model using complex power and voltage samples. Numerical examples in ieee 13-bus, 37-bus, and 123-bus systems demonstrate that the proposed surrogate model can outperform surrogate models based on the deep neural network and Gaussian process regarding prediction accuracy and sample efficiency.
Power system operation is gaining complexity due to the changes imposed by the energy transition. Especially, the increased share of intermittent and decentralized renewable generation units in the energy mix, an incr...
详细信息
Power system operation is gaining complexity due to the changes imposed by the energy transition. Especially, the increased share of intermittent and decentralized renewable generation units in the energy mix, an increased uncertainty regarding the supply of energy, and the predominantly market-driven cross-region and cross-border transport of electricity impose new challenges on the operation of power systems in Europe. In particular, power system operators must facilitate higher utilization of the grid capacity and coordinate more with neighboring transmission system operators (TSOs) and distribution system operators (DSOs). To deal with these new challenges, there is a pressing need to improve the observability and controllability of key system parameters to safeguard the reliability of power systems. Furthermore, the aforementioned developments and challenges go hand in hand with the need to improve the system resilience from the cybersecurity and system stability points of view. In the future, these challenges cannot be met without innovation towards intelligent decision support systems and assistant functions, which allow a look ahead combined with fast response and proactive actions. Here, the rather novel digital twin (DT) concept in combination with data-driven (i.e., machine learning) applications can be purposefully applied.
The multicut problem, also known as correlation clustering, is a classic combinatorial optimization problem that aims to optimize graph partitioning given only node (dis)similarities on edges. It serves as an elegant ...
详细信息
The multicut problem, also known as correlation clustering, is a classic combinatorial optimization problem that aims to optimize graph partitioning given only node (dis)similarities on edges. It serves as an elegant generalization for several graph partitioning problems and has found successful applications in various areas such as data mining and computer vision. However, the multicut problem with an exponentially large number of cycle constraints proves to be NP-hard, and existing solvers either suffer from exponential complexity or often give unsatisfactory solutions due to inflexible heuristics driven by hand-designed mechanisms. In this article, we propose a deep graph reinforcement learning method to solve the multicut problem within a combinatorial decision framework involving sequential edge contractions. The customized subgraph neural network adapts to the dynamically edge-contracted graph environment by extracting bilevel connected features from both contracted and original graphs. Our method can learn to infer feasible multicut solutions end-to-end toward optimization of the multicut objective in a data-driven manner. More specifically, by exploring the decision space adaptively, it implicitly gains heuristic knowledge from topological patterns of instances and thereby generates more targeted heuristics overcoming the short-sightedness inherent in the hand-designed ones. During testing, the learned heuristics iteratively contract graphs to construct high-quality solutions within polynomial time. Extensive experiments on synthetic and real-world multicut instances show the superiority of our method over existing combinatorial solvers, while also maintaining a certain level of out-of-distribution generalization ability.
In this paper, we solve the optimal output regulation problem for discrete-time systems without precise knowledge of the system model. Drawing inspiration from reinforcement learning and adaptive dynamic programming, ...
ISBN:
(纸本)9798350301243
In this paper, we solve the optimal output regulation problem for discrete-time systems without precise knowledge of the system model. Drawing inspiration from reinforcement learning and adaptive dynamic programming, a data-driven solution is developed that enables asymptotic tracking and disturbance rejection. Notably, it is discovered that the proposed approach for discrete-time output regulation differs from the continuous-time approach in terms of the persistent excitation condition required for policy iteration to be unique and convergent. To address this issue, a new persistent excitation condition is introduced to ensure both uniqueness and convergence of the data-driven policy iteration. The efficacy of the proposed methodology is validated by an inverted pendulum on a cart example.
This paper proposes a method for calibrating control parameters. Examples of such control parameters are gains of PID controllers, weights of a cost function for optimal control, filter coefficients, the sliding surfa...
详细信息
This paper proposes a method for calibrating control parameters. Examples of such control parameters are gains of PID controllers, weights of a cost function for optimal control, filter coefficients, the sliding surface of a sliding mode controller, or weights of a neural network. Hence, the proposed method can be applied to a wide range of controllers. The method uses a Kalman filter that estimates control parameters, using data of closed-loop system operation. The control parameter calibration is driven by a training objective, which encompasses specifications on the performance of the dynamical system. The performance-driven calibration method tunes the parameters online and robustly, is computationally efficient, has low data storage requirements, and is easy to implement making it appealing for many real-time applications. Simulation results show that the method is able to learn control parameters quickly, is able to tune the parameters to compensate for disturbances, and is robust to noise. A simulation study with the high-fidelity vehicle simulator CarSim shows that the method can calibrate controllers of a complex dynamical system online, which indicates its applicability to a real-world system. We also verify the real-time feasibility on an embedded platform with automotive-grade processors by implementing our method on a dSPACE MicroAutoBox-II rapid prototyping unit.
In this paper, we address data-driven predictive control of Linear Time-Invariant (LTI) systems. Specifically, we demonstrate the direct learning of predictive laws from data, eliminating the need for system identific...
详细信息
ISBN:
(纸本)9798331517519;9798331517526
In this paper, we address data-driven predictive control of Linear Time-Invariant (LTI) systems. Specifically, we demonstrate the direct learning of predictive laws from data, eliminating the need for system identification prior to control. Indeed, a data-based system representation, which provides a more accurate description of the original system, is constructed to replace the traditional model for predicting future behaviors. This approach helps to reduce the computational burden. Furthermore, two separate techniques for data-driven predictive control are presented for stabilizing an unknown LTI system. In each approach, a state feedback control law is formulated at every time step to optimize an infinite horizon objective function with the goal of stabilizing the unknown system based on the available data. We exemplify our findings through two numerical examples, emphasizing key features of our approaches.
In the context of microgrids, Battery Energy Storage systems play a vital role in maintaining power system reliability and efficiency. Adequate control strategies are necessary for exploiting these resources appropria...
详细信息
ISBN:
(纸本)9798350370959;9798350370942
In the context of microgrids, Battery Energy Storage systems play a vital role in maintaining power system reliability and efficiency. Adequate control strategies are necessary for exploiting these resources appropriately. However, developing accurate models for complex energy systems is becoming quite challenging. This study presents the application of a Set-Membership data-driven approach in the controller design process for a Real Battery Energy Storage System with a peak capacity of 267kW, avoiding the requirement of a modeling step and directly deriving the controller from data. A comparison with an existing PID controller, tuned by a trial-and-error procedure, demonstrates that the data-drivencontroller significantly improves the system performance by reducing the step response time by up to 18%, the overshoots to less than 1%, and increasing the robustness of the loop to time-variant delays. This work highlights how the considered data-driven approach helps in the decision-making process to optimize the system's performance.
We present some preliminary ideas on a data-driven Model Predictive control framework for continuous-time systems. We use Chebyshev polynomial orthogonal bases to represent system trajectories and subsequently develop...
详细信息
暂无评论