We present a novel boundary-aware face alignment algorithm by utilising boundary lines as the geometric structure of a human face to help facial landmark localisation. Unlike the conventional heatmap based method and regression based method, our approach derives face landmarks from boundary lines which remove the ambiguities in the landmark definition. Three questions are explored and answered by this work: 1. Why using boundary? 2. How to use boundary? 3. What is the relationship between boundary estimation and landmarks localisation? Our boundaryaware face alignment algorithm achieves 3.49% mean error on 300-W Fullset, which outperforms state-of-the-art methods by a large margin. Our method can also easily integrate information from other datasets. By utilising boundary information of 300-W dataset, our method achieves 3.92% mean error with 0.39% failure rate on COFW dataset, and 1.25% mean error on AFLW-Full dataset. Moreover, we propose a new dataset WFLW to unify training and testing across different factors, including poses, expressions, illuminations, makeups, occlusions, and blurriness. Dataset and model will be publicly available at https://wywu. github.io/projects/LAB/LAB.html

Reactive power outputs of DGs are used along with capacitor banks to regulate distribution network voltage. However, reactive power capability of a DG is limited by the inverter ratings and real power outputs of the DG. In order to achieve optimal power flow, minimize power losses, and minimize switching of capacitor banks, a day-ahead coordinated dispatch method of reactive power is proposed. Forecast errors of DG real power in every period are used to estimate the probability distribution of DGs reactive power capacity. Considering different output characteristics and constraints of reactive power sources, a dynamic preliminary-coarse-fine adjustment method is designed to optimize DG and shunt compensator outputs, decrease the switching cost, and reduce loss. The preliminary optimization obtains initial values, and multiple iterations between the coarse and fine optimizations are used to achieve a coordinated result. Simulations studies are performed to verify the proposed method.

Transfer capacities of urban distribution networks need to be increased to fulfill the increasing load demands and to accommodate distributed generation (DG). However, there are limited spaces to build new substations and lines, and curtailment of DGs may happen due to voltage violation during DG outputs fluctuation. This paper proposes and analyzes various methods to convert some existing ac MV lines to dc lines in order to form a hybrid ac/dc distribution network, based on which transfer capacities of lines can be increased, and flexible power shift can be achieved through a voltage source converter between two lines. The increases of transfer capacities are quantified. Optimal operation to fully utilize the increased capacities is achieved, in which losses are minimized in day-ahead scheduling, and node voltages are regulated real-time within security ranges based on limited measurement. Not only reactive power but also real power optimization are designed to maximize load supply and DG accommodation. A sensitivity method is proposed considering relatively large r/x ratio of an MV distribution network, which is effective for the real-time voltage regulation. Simulations are performed to verify the proposed method.

No abstract

No abstract