Abstract: The present disclosure provides for dynamically deactivating rectifiers to force remaining rectifiers to operate at or near their peak power efficiency. Rectifiers, for example rectifiers on racks of a data center, may operate according to an efficiency curve, based on its current load. Instead of distributing an AC power load across more rectifiers that operate sub-optimally on their efficiency curve, aspects of the disclosure provide for automatically deactivating some rectifiers by lowering voltage set-points. As power load to a rack decreases, the voltage of the current to a rectifier with a reduced voltage set-point falls below the set-point and turns off. Power is automatically redistributed to the remaining active rectifiers. The redistribution increases the power load onto the remaining rectifiers, allowing the rectifiers to perform more efficiently in converting AC power to DC power.

Abstract: Current imbalance may be detected and components reactively moved to correct the current imbalance. The components, such as rectifiers, machines, etc., may be moved from the most loaded phase to the least loaded phase. The imbalance may be detected at one or more power distribution units. Rebalancing may be performed using a model which preserves the number of components per rack, while limiting per-rack phase imbalance and minimizing imbalance among phases. Once the rebalancing has been computed, instructions for moving components according to the rebalancing may be generated.

Abstract: Methods, systems, apparatus and computer program products for implementing machine learning within control systems are disclosed. An industrial facility setting slate can be received from a machine learning system and a determination can be made as to whether to adopt the settings in the industrial facility setting slate. The machine learning model can be a neural network, e.g., a deep neural network, that has been trained, e.g., using reinforcement learning to predict a data setting slate that is predicted to optimize an efficiency of a data center.

Abstract: Methods, systems, apparatus and computer program products for implementing machine learning within control systems are disclosed. An industrial facility setting slate can be received from a machine learning system and a determination can be made as to whether to adopt the settings in the industrial facility setting slate. The machine learning model can be a neural network, e.g., a deep neural network, that has been trained, e.g., using reinforcement learning to predict a data setting slate that is predicted to optimize an efficiency of a data center.

Abstract: Current imbalance may be detected and components reactively moved to correct the current imbalance. The components, such as rectifiers, machines, etc., may be moved from the most loaded phase to the least loaded phase. The imbalance may be detected at one or more power distribution units. Rebalancing may be performed using a model which preserves the number of components per rack, while limiting per-rack phase imbalance and minimizing imbalance among phases. Once the rebalancing has been computed, instructions for moving components according to the rebalancing may be generated.

Abstract: Methods, systems, apparatus and computer program products for implementing machine learning within control systems are disclosed. An industrial facility setting slate can be received from a machine learning system and a determination can be made as to whether to adopt the settings in the industrial facility setting slate. The machine learning model can be a neural network, e.g., a deep neural network, that has been trained, e.g., using reinforcement learning to predict a data setting slate that is predicted to optimize an efficiency of a data center.