Precisely forecasting wind speed is essential for wind power producers and grid operators. However, this task is challenging due to the stochasticity of wind speed. To accurately predict short-term wind speed under uncertainties, this paper proposed a multi-variable stacked LSTMs model (MSLSTM). The proposed method utilizes multiple historical meteorological variables, such as wind speed, temperature, humidity, pressure, dew point and solar radiation to accurately predict wind speeds. The prediction performance is extensively assessed using real data collected in West Texas, USA. The experimental results show that the proposed MSLSTM can preferably capture and learn uncertainties while output competitive performance.

Efficient job scheduling on data centers under heterogeneous complexity is crucial but challenging since it involves the allocation of multi-dimensional resources over time and space. To adapt the complex computing environment in data centers, we proposed an innovative Advantage Actor-Critic (A2C) deep reinforcement learning based approach called A2cScheduler for job scheduling. A2cScheduler consists of two agents, one of which, dubbed the actor, is responsible for learning the scheduling policy automatically and the other one, the critic, reduces the estimation error. Unlike previous policy gradient approaches, A2cScheduler is designed to reduce the gradient estimation variance and to update parameters efficiently. We show that the A2cScheduler can achieve competitive scheduling performance using both simulated workloads and real data collected from an academic data center.

Single-phase PWM rectifier, grid connected PV inverter system, static synchronous compensator (STATCOM) all can be implemented by an H-bridge inverter and a large electrolytic dc capacitor to absorb the ripple power pulsating at twice the line frequency (2ω ripple power). This paper proposed a generalized active power decoupling method for Hbridge with minimum voltage and current stress. By adding another phase leg to control an ac capacitor, the 2ω ripple power can be absorbed by the capacitor and theoretically 2ω ripples to the dc capacitor can be eliminated completely. The H-bridge and the addition phase leg can be analyzed together as an unbalanced three phase system. By adopting SVPWM control and choosing the optimum ac capacitance and the capacitor voltage reference, the voltage and current stress of the switches can be minimized to the same as the conventional H-bridge. The size of capacitor is reduced by 10 times compared to the conventional H-bridge system. Simulation and experimental results are shown to prove the effectiveness of the proposed active power decoupling method.

A six-switch solid state variable capacitor (SSVC) with minimum dc capacitance is proposed. A variable ac capacitor (with capacitance variable from 0 to C ac ) is conventionally realized by an H-bridge inverter and a large electrolytic dc capacitor with capacitance of 10 times the ac capacitance value, C ac to absorb the 2ω dc ripple. The proposed SSVC consists of a six-switch converter with two ports. One port connected to an ac capacitor with fixed capacitance, C ac can reduce the dc capacitance to the minimum just for absorbing switching ripples. The fixed ac capacitor controlled by one port absorbs the 2ω component and theoretically can eliminate 2ω ripples to the dc capacitor completely. Therefore, no electrolytic capacitors would be needed. Theoretical analysis of the SSVC is provided. Simulation results are shown to prove the effectiveness of the proposed SSVC with minimum dc capacitance. I.978-1-4799-5776-7/14/$31.00 ©2014 IEEE

A new solid state variable capacitor (SSVC) with minimum dc capacitance is proposed. A variable ac capacitor (with capacitance variable from 0 to C ac ) is traditionally implemented by an H-bridge inverter and a large electrolytic dc capacitor with capacitance of 20 times the ac capacitance value, C ac to absorb the 2ω dc ripple. The proposed SSVC consists of an H-bridge and an additional phase leg connected to an ac capacitor with fixed capacitance, C ac and can reduce the dc capacitance to the minimum just for absorbing switching ripples. The fixed ac capacitor controlled by the additional phase leg absorbs the 2ω component and theoretically can eliminate 2ω ripples to the dc capacitor completely. Therefore, no electrolytic capacitors would be needed. Theoretical analysis of the SSVC is provided. Simulation and experimental results are shown to prove the effectiveness of the proposed SSVC with minimum dc capacitance.