Graph embedding based retrieval has become one of the most popular techniques in the information retrieval community and search engine industry. The classical paradigm mainly relies on the flat Euclidean geometry. In recent years, hyperbolic (negative curvature) and spherical (positive curvature) representation methods have shown their superiority to capture hierarchical and cyclic data structures respectively. However, in industrial scenarios such as e-commerce sponsored search platforms, the large-scale heterogeneous query-item-advertisement interaction graphs often have multiple structures coexisting. Existing methods either only consider a single geometry space, or combine several spaces manually, which are incapable and inflexible to model the complexity and heterogeneity in the real scenario. To tackle this challenge, we present a web-scale Adaptive Mixed-Curvature ADvertisement retrieval system (AM-CAD) to automatically capture the complex and heterogeneous graph structures in non-Euclidean spaces. Specifically, entities are represented in adaptive mixed-curvature spaces, where the types and curvatures of the subspaces are trained to be optimal combinations. Besides, an attentive edge-wise space projector is designed to model the similarities between heterogeneous nodes according to local graph structures and the relation types. Moreover, to deploy AMCAD in Taobao, one of the largest e-commerce platforms with hundreds of million users, we design an efficient two-layer online retrieval framework for the task of graph based advertisement retrieval. Extensive evaluations on real-world datasets and A/B tests on online traffic are conducted to illustrate the effectiveness of the proposed system.

With the increasing amount and type of connected microgrids in the near-term future power networks, how to optimize the operation of a multi-microgrid system efficiently and reliably has become essential for taking full advantage of the complex systems. In this paper, we propose a multi-timescale coordinated optimization strategy for hybrid three-phase/single-phase multi-microgrids. With the consideration of the economy of microgrids and three-phase unbalance constraints of multi-microgrids, a strategy of collecting-distributing fuzzy modified adaptive particle swarm economic optimization is provided. In this day-ahead strategy, the system is constructed as a bi-layer rolling optimizing structure, where the lower layer sets objectives for economic optimization, and the upper layer balances constrained tie-line power of single-phase microgrids based on the optimized results obtained from lower layer. On this basis, by taking day-ahead optimized tie-line power as baseline values, we obtain the real-time power of energy storage systems using an improved nondominated sorting genetic algorithm, which achieves distributed real-time optimization for the multi-microgrid. Simulation results verify that the proposed strategy is feasible.

To manage the microgrid effectively, it is important to get the power information among multiple distributed generation (DG) units. However, when the droop control method is adopted, as the corresponding DG units cannot be simply modeled as VF or PQ buses, the conventional power flow algorithm may become inapplicable. To solve this issue, a time domain iteration (TDI) based power flow algorithm is hereby proposed. Firstly, a microgrid model is prepared, containing a network model and several DG unit models. Then, the proposed TDI is executed for power flow calculation, which mimics the real-time operation of microgrids. In each iteration, the DG unit models input voltages and currents to the network model; then the network model changes its state accordingly and feeds related parameters back to those DG unit models. As the DG unit models simulate the behavior of actual DG units, the proposed algorithm is not limited to the droop control governed microgrids. Moreover, as the TDI has definite physical meaning, convergence problems can be avoided. Finally, the validity of the proposed power flow algorithm is verified through the Matlab simulation results from an 8-bus microgrid system.

For an islanded microgrid, load power should be properly shared among multiple distributed generation (DG) units, not only for the fundamental power but also for the harmonic power. In this paper, the operation of microgrids under imbalanced and nonlinear load conditions is studied, and a dynamic consensus based power controller is proposed for DG units to realize enhanced power sharing performance. Firstly, the fundamental and low order harmonic powers of a DG unit are detected by using second-order generalized integrator (SOGI) modules. Then, the obtained local power information is transmitted to the neighbor DG units to help tuning their virtual impedances at fundamental and selected harmonic frequencies. After this, accurate fundamental and harmonic power sharing performance will be achieved. In the proposed control strategy, DG units only communicate with their neighbor units, thus the “plug and play” function will be reserved. Compared with those centralized control methods, the proposed strategy with distributed consensus protocol can simplify the maintenance and expansion of the system, making the microgrid more flexible. Moreover, as the knowledge of the detailed network is not required, it is convenient to be applied in practice. Simulation results are presented to demonstrate the proposed method.

The paper presents a microgrid energy management strategy (MEMS) for a Medium-Voltage Isolated microgrid, which is China's 10kv island microgrid demonstration project to accommodate the increasing power needs of customers and solve problems of island power supply. When the SOC is restored in security zone, the energy storage system (ESS) would increase the sensitivity and adjust the maximum charging or discharging power to avoid repetitive fluctuations. According to the system's control algorithm, when the microgrid load rate is normal, or above the upper limit, or below the lower limit, the MEMS system can start or stop some micro sources, charge or discharge battery, to maintain the stability of the system's voltage and frequency and to confirm the microgrid stable and effective. The effectiveness of the proposed strategy is verified by experimental results.

In an autonomous microgrid without connection to the interconnected grid to support the frequency and voltage, it is more complex to control and manage. Thus in order to investigate the dynamic stability of the medium-voltage (MV) microgrid in Dongao Island, a detailed small-signal state-space model of the autonomous MV mixed-source microgrid containing diesel generator set (DGS), grid-supporting battery energy storage system (BESS), squirrel cage induction generator (SCIG) wind turbine and network is developed. Sensitivity analysis is carried out to reveal the dynamic stability margin of the MV microgrid and identify the permissible variations range of the droop coefficients. Finally, the theoretical analysis is verified with time-domain simulation and experimental results.