Data Center Power Demand Shaping via Adaptive Load Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed generation systems in data centers face challenges in efficiently managing time-varying power demands due to their limited response speed, leading to increased energy storage costs, energy loss, and reduced battery lifespan, which complicates maintaining workload performance and environmental sustainability.
Innovation Solution
Implementing a power demand shaping technique that utilizes adaptive load tuning schemes, such as DGR Boost and UPS Boost, to coordinate power supply and load, allowing DG systems to efficiently follow power demand fluctuations and reduce reliance on utility grids, thereby enhancing performance and extending battery life.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If distributed generation systems are used to provide local power, then energy savings and environmental benefits are achieved, but the limited response speed of DG systems cannot handle moment-to-moment load power demand
Solution Approach 1:
The patent segments the power management function into two distinct components: a fast-responding energy storage system (batteries) that handles moment-to-moment load fluctuations, and a slower distributed generation system that provides sustained power over longer periods. This segmentation allows each component to operate within its optimal response characteristics, resolving the contradiction between energy efficiency and response speed.
2Speed
If large energy storage elements are used to handle load power demand, then moment-to-moment power demand is met, but total cost of ownership increases and roundtrip energy loss reaches up to 25%
Solution Approach 1:
The patent implements partial action by having the energy storage system provide only the incremental power needed to bridge the response gap between load demand and DG capability, rather than using oversized storage capacity. The storage system operates partially to supplement DG during transition periods, achieving fast response without the excessive energy losses associated with large-scale storage systems operating at full capacity.
3Productivity
If frequent battery discharging activities are used to follow load changes, then load power demand is met, but battery lifetime is degraded and stored energy is quickly depleted
Solution Approach 1:
The patent establishes continuous useful action by maintaining a baseline level of stored energy in the battery system that continuously supports load following activities. Rather than allowing the battery to fully deplete during each load cycle, the system continuously replenishes and maintains energy reserves, ensuring that discharging activities remain within sustainable limits that preserve battery lifetime while maintaining productive load following capability.
4Loss of energy
If load tuning techniques are used to track time-varying renewable power budget, then power management is achieved, but job turnaround time is extended and instantaneous throughput cannot be maintained
Solution Approach 1:
The patent applies preliminary action by having the energy storage system pre-charged during periods of excess renewable generation and low demand, storing energy in advance before load peaks occur. This preliminary energy accumulation allows the system to respond immediately to sudden load increases without waiting for DG systems to ramp up, thereby maintaining fast job turnaround times while still utilizing renewable energy sources.
Data Source
AI summary
Embodiments relate to a method and system for power demand shaping (PDS) so as to manage power generation and use. Specific embodiments relate to data center power demand shaping to achieve high-performance low-overhead data center operation. Specific embodiments can incorporate standard (utility power) energy sources, renewable energy sources, or a combination of standard (utility power) energy sources and renewable energy sources. Embodiments of the subject PDS techniques can incorporate trimming the data center load power so as to allow DG systems to follow the power demand efficiently and/or incorporate two adaptive load tuning schemes that can boost data center performance and enable near-oracle operation during power demand trimming process. To implement a cross-layer power optimization scheme, embodiments of the subject invention relate to a power management module that can reside between front-end distributed generation and back-end computing facilities to provide a coordinated tuning between the supply and load.