Federated learning (FL) is a distributed training technique for machine learning (ML) models that ensures ownership of training data for the devices or users. Data ownership is guaranteed when the devices train the ma...
详细信息
Federated learning (FL) is a distributed training technique for machine learning (ML) models that ensures ownership of training data for the devices or users. Data ownership is guaranteed when the devices train the machine model. The attribution of responsibility for the distributed training of the model causes variations in the training efficiency based on the characteristics or behaviors of these users. Among the user characteristics that can interfere with federated training is mobility. The mobility of users may prevent the user from completing the training by losing connection with other devices on the network, causing a client dropout. This work introduces a specific FL coordination algorithm to guarantee training efficiency in scenarios with mobility named MoFeL. To analyze its efficiency, we performed simulation experiments using machine models trained by a convolutional neural network from an image classification application. Simulation results show that MoFeL performs FL even in scenarios with intense user mobility, while other traditional training coordination algorithms cannot do so.
At present, most of the existing optimal reconfiguration methods for power system active distribution network (ADN) aim to change the network topology without considering the Adjustable load. In ADN, the efficiency of...
详细信息
ISBN:
(纸本)9781665423144
At present, most of the existing optimal reconfiguration methods for power system active distribution network (ADN) aim to change the network topology without considering the Adjustable load. In ADN, the efficiency of the original reconstruction method can't meet the requirements when the proportion of the Adjustable load is increasing. Based on the consideration of the Adjustable load, this paper establishes the fast non dominated sorting genetic algorithm (NSGA-II) based on the coevolution and the dual track system coevolution group evolution model based on the coevolution total learning particle swarm optimization (CE-MOCLPSO), and solves the problem of the active distribution network reconfiguration and the simultaneous optimization of the Adjustable load. In this paper, based on the IEEE33 node distribution system design, several simulation examples are analyzed. The results show that the active distribution network reconfiguration considers the scheduling of Adjustable load, effectively improves the operation quality of the distribution network, and verifies the efficiency of the reconfiguration scheme is higher than the original two-stage reconfiguration method.
To account for endogeneity effects in, e.g., energy demand, modern bottom-up energy system models are often linked to a top-down model describing the macroeconomic system. Solving such linked models involves iterative...
详细信息
To account for endogeneity effects in, e.g., energy demand, modern bottom-up energy system models are often linked to a top-down model describing the macroeconomic system. Solving such linked models involves iteratively passing solutions from one model to the other and vice versa until convergence is reached, which can be computationally demanding. This paper proposes a coordination algorithm that speeds up convergence for the linkage of the two models in case the bottom-up model is a linear program and the top-down model is a mixed-complementarity problem. The coordination algorithm uses duality theory to select optimal bases from previous iterations to predict the solution of the bottom-up model. If the predicted solution is correct, which is shown to be equivalent to the predicted solution vector being non-negative, the bottom-up problem need not be solved in that iteration, resulting in a time gain. Numerical experiments on an energy system design problem illustrate that our coordination algorithm correctly predicts the bottom-up solution in most iterations, resulting in a significant reduction in overall computation time.
The use of battery energy storage systems (BESS) is one of the methods employed in solving the major challenge of overvoltage, experienced on low voltage (LV) distribution networks with high penetration of photovoltai...
详细信息
The use of battery energy storage systems (BESS) is one of the methods employed in solving the major challenge of overvoltage, experienced on low voltage (LV) distribution networks with high penetration of photovoltaics (PV). The overvoltage problem limits the penetration levels of PV into the LV network, and the benefits that could be gained. This paper presents a smart scheme for the coordination of multiple battery energy storage systems (BESS) in such networks. An approximate method was adopted for the evaluation of network voltage sensitivity, and the coordination algorithm was developed based upon this. Through the efficient selection, coordination and timing of charge and discharge operations of the BESS, the scheme maintains bus voltages within statutory ranges during both periods of high PV power generation and high network load demand. The scheme also prevents sudden voltage rise, which usually occurs in such networks immediately a BESS gets fully charged. Simulations were carried out on a real LV distribution network and results demonstrate the effectiveness of this approach.
Presently, most of the normal distribution system configuration is a radial type. When a fault occurs in the radial system, one or more loads experience an outage for a few minutes until the fault is cleared. But if t...
详细信息
Presently, most of the normal distribution system configuration is a radial type. When a fault occurs in the radial system, one or more loads experience an outage for a few minutes until the fault is cleared. But if the distribution system is configured with a loop type, then the power can be supplied to all customers without any outage. So, for improving reliability of power supply, the research of the loop distribution system has been performed recently. Previously, At the KEPCO PT Center (PT: Power Testing), several tests like single fault, communication failure, CB open failure, etc, were performed to verify performances of the coordination control algorithm for loop distribution. However, there are some limitations in this field test for overall its functions. For example, a few tests such as a three-phase fault, line to line fault and the Closed-loop Control Device (CCD) internal fault are hard to fulfill in field test. Therefore, it should be tested by Real Time Digital Simulator (RTDS). The RTDS is a fully digital power system simulator in real-time and hence can make an environment similar to physical power system. In this paper, the validity of the coordination algorithm for a loop distribution system is verified with RTDS which is connected with the protection relays called CCD.
This paper is concerned with assessing the contribution of grid-scale storage to generation capacity adequacy. Results are obtained for a utility-scale exemplar involving the Great Britain power system. All stores are...
详细信息
This paper is concerned with assessing the contribution of grid-scale storage to generation capacity adequacy. Results are obtained for a utility-scale exemplar involving the Great Britain power system. All stores are assumed, for the purpose of capacity adequacy assessment, to be centrally controlled by the system operator, with the objective of minimising the Expected Energy Not Served over the peak demand season. The investigation is limited to stores that are sufficiently small such that discharge on one day does not restrict their ability to support adequacy on subsequent days. We argue that for such stores, the central control assumption does not imply loss of generality for the results. Since it may be the case that stores must take power export decisions without the benefit of complete information about the state of the system, a methodology is presented for calculating bounds on the value of such information for supporting generation adequacy. A greedy strategy is proven to be optimal for the case where decisions can be made immediately after a generation shortfall event has occurred, regardless of the decision maker's risk aversion. The adequacy contribution of multiple stores is examined, and algorithms for coordinating their responses are presented. (c) 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license.
The scheduling of nodes in a wireless Ad Hoc Sensor network is getting them to alternate between the sleeping and active mode. If this process of adjusting the wake/sleep schedule of all nodes that is, a topology mana...
详细信息
ISBN:
(纸本)9781538611043
The scheduling of nodes in a wireless Ad Hoc Sensor network is getting them to alternate between the sleeping and active mode. If this process of adjusting the wake/sleep schedule of all nodes that is, a topology management mechanism, is maintained in an optimal manner, further energy can be saved, which will have a direct impact on prolonging the lifetime of the network. So, in this paper a distributed power saving coordination algorithm for multi-hop ad hoc wireless networks based on learning automata without significantly diminishing the quality of services of the network such as capacity or connectivity of the network is proposed such that all nodes in the network that are equipped with learning automata don't need to be synchronized with each other. Learning automata abilities such as low computational load, usability in distributed environments with ambiguous information, and adaptability to changes via low environmental feedbacks are all, factors that can provide the mentioned optimal manner for nodes;and cause better fitness with local techniques in ad hoc wireless networks. The proposed protocol, SpanLAQ, consists of two phases: coordinator announcement and coordinator withdrawal which are based on learning automata to ensure fairness, and to make local decisions on whether a node is going to sleep or joining to a forwarding backbone as a coordinator. So, using learning automata with passing of time causes a decrease in energy consumption and an improvement of the network lifetime in SpanLAQ protocol with an 802.11 network in power saving mode, in comparison to other similar protocols such as Span and without any topology management protocols. Simulation results with a practical energy model also show that the above result is being achieved with some improvements in capacity, connectivity and communication latency.
In this work a distributed control method is proposed to mitigate the unbalance in a low voltage grid based on an approximated three-phase power flow. As the power flow is both explicit and linear with respect to the ...
详细信息
ISBN:
(纸本)9781538609439
In this work a distributed control method is proposed to mitigate the unbalance in a low voltage grid based on an approximated three-phase power flow. As the power flow is both explicit and linear with respect to the apparent power, it can be substituted into the objective function of the controllers leading to a separable optimization problem. The resulting subproblems are coupled in the costs which are related to the negative-sequence voltage formulated as a linear function of the apparent power. By exchanging nodal measurements via a communication bus, the controllers adjust their set-points in order to mitigate the unbalance with a Jacobi-like coordination algorithm. The effectiveness of the approach is demonstrated in a simulation study with an unbalanced low voltage grid.
Exploration is a fundamental research problem in robotics, and multi-robot exploration has been widely studied in this field. However, unexplored islands exist in most of the bidding-based coordination algorithms duri...
详细信息
Exploration is a fundamental research problem in robotics, and multi-robot exploration has been widely studied in this field. However, unexplored islands exist in most of the bidding-based coordination algorithms during the exploration process, which directly affects the exploration time. Meanwhile, some unnecessary overlapping also affects the exploration time. To overcome these problems, we propose a multi-robot coordinated strategy that includes bidding and rules for the unknown environment exploration. The proposed algorithm has a hierarchical architecture. The upper level is the emotion generation system, which can generate emotional states through the perception of input stimuli from the environment. In the lower level, the robots perform bidding activities and walk rules according to the emotional states. The simulation results of 640 voyages demonstrate that the unexplored-island problem is resolved, and our algorithm has a lower overlap rate than the bidding-based coordination algorithm. Thus, the exploration time is efficiently reduced. Moreover, our algorithm has a good ability of detecting static and dynamic obstacles.
In this work we study a multi-agent coordination problem in which agents are only able to communicate intermittently with the cloud. To minimize the amount of required communication, we are interested in developing a ...
详细信息
ISBN:
(纸本)9781467386838
In this work we study a multi-agent coordination problem in which agents are only able to communicate intermittently with the cloud. To minimize the amount of required communication, we are interested in developing a self-triggered algorithm that determines when communication with the cloud is necessary. Unlike the overwhelming majority of similar works that propose distributed event- and/or self-triggered control laws, this work doesn't assume agents can be `listening' continuously. In other words, when an event is triggered by one agent, neighboring agents will not be aware of this until the next time they establish communication with the cloud. Each time an agent communicates with the cloud, it must determine the next time to reestablish communication while guaranteeing the completion of some global task. We show that our self-triggered coordination algorithm guarantees that the system asymptotically reaches the set of desired states. Simulations illustrate our results.
暂无评论