With the proliferation of distributed energy resources (DERs), voltage regulation in active distribution networks (ADNs) has been facing a great challenge. This paper derives an asynchronous distributed voltage contro...
详细信息
With the proliferation of distributed energy resources (DERs), voltage regulation in active distribution networks (ADNs) has been facing a great challenge. This paper derives an asynchronous distributed voltage control strategy based on the partial primal-dual gradient algorithm, where both active and reactive power of DERs are considered. Different types of asynchrony due to imperfect communication or other practical limits, such as random time delays and non-identical sampling/computation rates, are fitted into a unified analytic framework. The proposed asynchronous algorithm is then converted into a fixed-point problem by leveraging the operator splitting method, which leads to the convergence proof. Moreover, an online implementation method is introduced to make the controller adjustable to time-varying environments. Finally, numerical experiments are carried out on a rudimentary 8-bus and the IEEE-123 bus system to verify the effectiveness of the proposed method. (c) 2020 Elsevier Ltd. All rights reserved.
Distribution systems are becoming more complex with the integration of numerous distributed energy resources (DERs), such as distributed generators (DGs), distributed energy storages (DESs), and controllable end-users...
详细信息
Distribution systems are becoming more complex with the integration of numerous distributed energy resources (DERs), such as distributed generators (DGs), distributed energy storages (DESs), and controllable end-users (EUs). Aggregators (Aggs) are emerging to bridge the gap between the distribution system operator (DSO) and EUs. This creates a hierarchical "DSO-Agg-EU" architecture that requires effective and efficient coordination among massive agents. However, this architecture faces challenges such as communication asynchrony and solution inexactness among different agents, which may undermine the performance of energy management or even cause unexpected failures. To address these challenges, this paper proposes an error-tolerant and asynchronous algorithm based on dual decomposition for hierarchically distributed energy management (HDEM). The algorithm is theoretically guaranteed to converge to a near-optimal value and can be further accelerated by Nesterov's method. The algorithm is tested on the IEEE 123-bus distribution system and demonstrates high effectiveness and efficiency under asynchrony and inexactness.
We consider the problem of link scheduling in multihop wireless networks subject to interference constraints. We consider a graph-based representation it wireless networks where scheduled links adhere to the K-hop lin...
详细信息
We consider the problem of link scheduling in multihop wireless networks subject to interference constraints. We consider a graph-based representation it wireless networks where scheduled links adhere to the K-hop link interference model. For this scheduling problem. we develop an asynchronous and distributed greedy heuristic. Furthermore, we show that this distributed heuristic computes the exact same schedule as the well-known centralized greedy link scheduling algorithm.
The problem of data reversal in discretized adjoint problems is often solved using checkpointing, trading, memory usage with computations and data movement. The authors present a useful model to design and implement a...
详细信息
The problem of data reversal in discretized adjoint problems is often solved using checkpointing, trading, memory usage with computations and data movement. The authors present a useful model to design and implement an asynchronous two-level checkpointing method with parameterizable values for current and future system configurations. They also evaluate the benefits of new supercomputing hardware through the implementation of an asynchronous algorithm that takes advantage of the fast NVLINK interconnect and Non-Volatile Memory Express (NVMe) memory. They show that the new hardware combined with an asynchronous approach is able to run bigger simulations faster than current generation hardware.
In this paper, we study the problem of decentralized learning in sensor networks in which local learners estimate and reach consensus to the quantity of interest inferred globally while communicating only with their i...
详细信息
In this paper, we study the problem of decentralized learning in sensor networks in which local learners estimate and reach consensus to the quantity of interest inferred globally while communicating only with their immediate neighbours. The main challenge lies in reducing the communication cost in the network, which involves inter-node synchronisation and data exchange. To address this issue, a novel asynchronous broadcast-based decentralized learning algorithm is proposed. Furthermore, we prove that the iterates generated by the developed decentralized method converge to a consensual optimal solution (model). Numerical results demonstrate that it is a promising approach for decentralized learning in sensor networks. [GRAPHICS] . The execution model on a decentralized sensor network and the workflow of asynchronous computing.
This letter studies the circular formation control problem from a perspective of distributed optimization. Two types of algorithms are developed, i.e., a synchronous Jacobi iteration method and its asynchronous varian...
详细信息
This letter studies the circular formation control problem from a perspective of distributed optimization. Two types of algorithms are developed, i.e., a synchronous Jacobi iteration method and its asynchronous variant, both of which are proven to achieve the convergence with exponential rates. As distinct from the existing continuous-time approaches, a step-size is imposed into our discrete-time methods and its influence on the convergence rate is analyzed explicitly. According to the analysis, the optimal choice of step-size is provided with respect to the convergence rate. Extensive numerical results are finally present to demonstrate the theoretical findings in this letter.
Solving L2-regularized empirical risk minimization (e.g., linear SVMs and logistic regression) using multiple cores has become an important research topic. Among all the existing algorithms, Parallel asynchronous Stoc...
详细信息
ISBN:
(纸本)9781509054732
Solving L2-regularized empirical risk minimization (e.g., linear SVMs and logistic regression) using multiple cores has become an important research topic. Among all the existing algorithms, Parallel asynchronous Stochastic dual Co-Ordinate DEscent (PASSCoDe) demonstrates superior performance compared with other methods. Although PASSCoDe is fast when it converges, the algorithm has been observed to diverge on several cases especially when a relatively large number of threads are used. This is mainly due to the delayed parameter access problem-the parameters used for the current update may be delayed and are not the latest ones. In theory, the algorithm converges only when the delay is small enough, but in practice the delay depends on the underlying parallel computing environment and cannot be guaranteed. In this work, we propose a simple and computational efficient way to fix the convergence problem of PASSCoDe. Instead of allowing all worker threads to conduct asynchronous updates wildly, we add periodic check points to the procedure, where all workers need to stop and refine the current solution at each check point. The resulting "semi-asynchronous" algorithm is guaranteed to converge for any problem even when PASSCoDe diverges, and for the cases where PASSCoDe converges they have almost identical speed.
Tensor decomposition is widely used in machine learning, recommendation systems, and social networks. Large real-world tensors require parallel algorithms running on distributed memory systems. Parallel algorithms suf...
详细信息
ISBN:
(数字)9781665498562
ISBN:
(纸本)9781665498562
Tensor decomposition is widely used in machine learning, recommendation systems, and social networks. Large real-world tensors require parallel algorithms running on distributed memory systems. Parallel algorithms suffer two major performance bottlenecks: load imbalance and communication cost, which are difficult to overcome due to the inherent tradeoff among the multiple types of computations and communications, especially for irregular sparse tensors. Previous work predominately focuses on balancing the load within the tensor-related computation, resulting in imbalance for multiple matrix-only computations and increased communication costs. It also extensively uses collective communication operations and bulk-synchronous computations by interleaving stages of global communication and stages of local computation, failing to hide the communication cost. In this paper, we present a novel algorithm BALA-CPD, which achieves the best overall workload balance, and effectively overlaps communication and computation for the popular distributed Canonical Polyadic Decomposition (CPD) algorithms. BALA-CPD uses a workload and data partition scheme that prioritizes the load balance for all the matrix-only computations and all the communications. When necessary, BALA-CPD adjusts to mitigate the load imbalance for the tensor-related computation. Departing from the bulk-synchronous approaches, BALA-CPD breaks down computation and communication in consecutive stages, and masks the communication costs by a combination of one-sided asynchronous communication and a fine-grained interleaving of communication and computation. We implement BALA-CPD and evaluate it on a 64-node cluster with 1280 processors. Experimental results show BALA-CPD is scalable and outperforms the state-of-the-art distributed implementations by up to 1.8x on 1280 processors.
There are many algorithms to solve large sparse linear systems in parallel;however, most of them acquire synchronization and thus are lack of scalability. In this paper, we propose a new distributed numerical algorith...
详细信息
ISBN:
(纸本)9781595939739
There are many algorithms to solve large sparse linear systems in parallel;however, most of them acquire synchronization and thus are lack of scalability. In this paper, we propose a new distributed numerical algorithm, called Directed Transmission Method (DTM). DTM is a fully asynchronous, scalable and continuous-time iterative algorithm to solve the arbitrarily-large sparse linear system whose coefficient matrix is symmetric-positive-definite (SPD). DTM is able to be freely running on the heterogeneous parallel computer with arbitrary number of processors, which might be manycore microprocessors, clusters, grids, clouds, and the Internet. We proved that DTM is convergent by making use of the final value theorem of Laplacian Transformation. Numerical experiments show that DTM is efficient.
In this paper we address the problem of asynchronous distributed principal component analysis. We provide several algorithms coping with different situations according to the underlying graph structure. A general enou...
详细信息
ISBN:
(纸本)9781467320665
In this paper we address the problem of asynchronous distributed principal component analysis. We provide several algorithms coping with different situations according to the underlying graph structure. A general enough framework allows us to analyze all these algorithms at the same time. Convergence is proved with probability 1 under suitable assumptions, and numerical experiments illustrate their good behavior.
暂无评论