This paper presents a method to achieve optimal active and reactive power contributions from each energy storage system in an unbalanced distribution network to minimize power loss, while ensuring network current and voltage constraints are satisfied. By modelling loads as either constant current or constant impedance, the AC optimal power-flow is transformed into a noniterative convex optimization problem. The application of capacity constraints, voltage constraints, and energy storage constraints in an unbalanced three-phase four-wire system is considered, addressing specific issues pertaining to unbalanced networks such as voltage unbalance and neutral voltage displacement. The proposed method is then used to demonstrate optimized dispatch of energy storage systems in a suitable four-wire unbalanced distribution test network. The contribution of losses in the neutral wire to the total losses is also determined for a test system under a range of operating conditions and various neutral earthing systems, highlighting the importance of considering this in a typical unbalanced distribution network.

No abstract

In the NeurIPS 2018 Artificial Intelligence for Prosthetics challenge, participants were tasked with building a controller for a musculoskeletal model with a goal of matching a given time-varying velocity vector. Top participants were invited to describe their algorithms. In this work, we describe the challenge and present thirteen solutions that used deep reinforcement learning approaches. Many solutions use similar relaxations and heuristics, such as reward shaping, frame skipping, discretization of the action space, symmetry, and policy blending. However, each team implemented different modifications of the known algorithms by, for example, dividing the task into subtasks, learning low-level control, or by incorporating expert knowledge and using imitation learning.

No abstract

No abstract