Abstract as per original application
(English/Chinese): |
We consider a combinatorial optimization problem to find a minimum-cost topology of a networked system, where the cost of each feasible topology is defined by an underlying continuous optimization of network resource allocation and flows. We focus on such a problem, called optimal topology and power flow (OTPF), in power distribution networks, which jointly optimizes the on/off status of switches on power lines and the generations, loads, and power flows, for more reliable, economical, and sustainable operation. The essential tradeoff between scalability (by which we mean a mild increase of computational burden with the network size) and optimality (a lower cost without violating physical and operational constraints) makes it hard to improve both attributes of an algorithm to solve OTPF, especially considering the practical three-phase unbalanced nonlinear alternating-current (AC) power flow models, the uncertainties of renewable generations and loads, and the key reliability indices, e.g., voltage stability, that do not have closed-form expressions.
We will tackle these challenges to develop OTPF solution algorithms with improved scalability and optimality. The proposed research is planned as three tasks:
(1) Develop a topology-informed switch opening and exchange algorithm based on convex relaxation to AC power flow, to solve deterministic OTPF problems with provable suboptimality bounds.
(2) Extend the algorithm in (1) with a topology-informed scenario clustering method, to solve stochastic and robust OTPF problems under uncertainties of renewable generations and loads, with improved computational efficiency.
(3) Merge deep-learning-based prediction of voltage stability indices into the topology-informed algorithms, to solve OTPF problems considering voltage stability enhancement. We shall validate scalability and optimality of the proposed algorithms through software simulations of practical power distribution network models with up to 11,000 nodes.
我們考慮一類組合優化問題來找到網絡化系統的最小成本拓撲結構，其中每個可行拓撲結構的成本由其底層的網絡資源分配和流量的連續優化來定義。我們關注配電網中的這樣一個問題，稱為最優拓撲和潮流（OTPF），對電力線路上的開關狀態以及發電、負載和功率流動進行聯合優化，以實現更可靠、經濟、可持續的運行。可擴展性（這裡意為計算負擔隨著網絡規模溫和增加）和最優性（在不違反物理和運行約束的情況下降低成本）之間的基本權衡關係使得同時改進一個OTPF算法的這兩種屬性變得十分困難，特別是考慮到實際的三相不平衡非線性交流潮流模型、可再生能源發電和負載的不確定性以及沒有閉式表達的關鍵可靠性指標（如電壓穩定性）。我們將應對這些挑戰，開發具有更高可擴展性和最優性的 OTPF 算法。我們提出的研究計劃分為三個任務：（1）開發一種基於交流潮流凸鬆弛的拓撲知情的開關斷開和交換算法，以求解確定性 OTPF 問題並證明次優邊界。（2）利用拓撲知情的場景聚類方法擴展（1）中的算法，以更高的計算效率求解可再生能源發電與負載不確定性下的隨機及魯棒OTPF問題。（3）將基於深度學習的電壓穩定性指標預測融入上述拓撲知情的算法，以解決電壓穩定性增強的OTPF問題。我們將通過對多達 11,000 節點的實際配電網模型進行軟件仿真來驗證所提出算法的可擴展性和最優性。
|
