Learning a Data Center Model for Efficient Demand Response

Data centers’ demand response (DR) program has been paid more and more attention recently. As an important component of data centers, multi-tenant data centers (also called “colocation”) play a significant role in the demand response, especially in the emergency demand response (EDR). In this paper, we focus on how the colocation can perform better in the EDR program. We formulate the “uncoordinated relationship” in the colocation which is the key problem affecting energy efficiency, and propose a reward system to motivate tenants to join the EDR program, and a truthful strategy is developed to ensure the authenticity of tenants’ information. For achieving the overall coordination, we integrate tenants’ resources to increase the colocation’s resource utilization and optimize the whole colocation’s energy efficiency, then devise two algorithms to solve the actual resource migration and integration problem. We analyze the complexity of allocation model and two algorithms. Experimental results show that our solution is practical and efficient.

Demand Response (DR) is often regarded as the killer application of the emerging smart grid. The ever-growing Internet population and energy consumption makes data centers auspicious candidates with great potential for DR. Besides commercial data centers and large server farms, small data centers scattered in academic or research environments consume considerable amount of energy. Unlike commercial data centers that are mission critical and must meet rigorous reliability and availability requirements, small data centers on campus are typically not mission critical. Such data centers are the most likely prospects for participating in DR programs. Different from the long term effort required for improving energy efficiency, DR programs expect participants to temporarily reduce their power demand on some short periods. The related information is still inadequate for organizations maintaining many small data centers, since the existing studies usually focus on energy efficiency issues for the large dedicated server farms. This study examines the power consumption recorded by smart meters from a small data center, and identifies the limitations and potential for small data centers participating in DR programs.

Power Management in Data Centers

This study examines data center characteristics, loads, control systems, and technologies to identify demand response (DR) and automated DR (Open Auto-DR) opportunities and challenges. The study was performed in collaboration with technology experts, industrial partners, and data center facility managers and existing research on commercial and industrial DR was collected and analyzed. The results suggest that data centers, with significant and rapidly growing energy use, have significant DR potential. Because data centers are highly automated, they are excellent candidates for Open Auto-DR. 'Non-mission-critical' data centers are the most likely candidates for early adoption of DR. Data center site infrastructure DR strategies have been well studied for other commercial buildings; however, DR strategies for information technology (IT) infrastructure have not been studied extensively. The largest opportunity for DR or load reduction in data centers is in the use of virtualization to reduce IT equipment energy use, which correspondingly reduces facility cooling loads. DR strategies could also be deployed for data center lighting, and heating, ventilation, and air conditioning. Additional studies and demonstrations are needed to quantify benefits to data centers of participating in DR and to address concerns about DR's possible impact on data center performance or quality of service and equipment life span.

In a re-structured environment, demand response (DR) programs are presented in a new style. The demand response program consists of demand-side management methods which answers to electricity price variations. With advent of electricity markets, the demand-side management programs was introduced into two categories: (1) energy efficiency program, (2) demand response program. This paper studies engagement of participants load in an electricity market having a grid with large and small loads interested to participate in the market, energy storage systems, micro-grids, and distributed generations. In the proposed scheme, it is supposed to have an aggregator which sends demand-side preferences including load curtailment, load shifting, onsite generation, and energy storage systems along with proposed value and prices and all respective system constraints to the independent system operator (ISO). In fact, load aggregators submit aggregated DR offers to the ISO in order to optimize final decisions on aggregators’ DR contributions in wholesale market. The main goal of this paper is to solve the operation problem considering demand-side, system, and pollution and reliability constraints. Compared to other methods, the results indicate that the explicit modeling of customer DR would provide ISOs with more flexible options for scheduling the available energy resources in day-ahead energy markets.

Demand response modeling considering Interruptible/Curtailable loads and capacity market programs

The rapid growth of cloud computing and data centers with skyrocketing energy consumption, together with the accelerating penetration of renewable energy sources, is creating both severe challenges and tremendous opportunities. Data centers offering large flexible loads in the grid, opens up a unique opportunity to smooth out the significant fluctuation and uncertainty of renewable generation and hence enable seamless integration. To take the market power of data centers into consideration, this paper proposes a bargaining solution to the market program for data center demand response when the load serving entity (LSE) has power supply deficiency. Specifically, due to the uncertainty of load flexibility of data centers incurred by the intermittent on-site renewable generation and dynamic service requests, there exists information asymmetry between the LSE and the data center, which complicates the design of the bargaining solution. Making use of the log-concavity of the (expected) utility functions, a computationally efficient method to implement the best response updates in the bargaining procedure is presented. Furthermore, it is shown analytically that the bid sequences of the LSE and the data center are guaranteed to converge and the final price clinched by the bargaining algorithm is indeed the Nash bargaining solution, which is proportionally fair. In addition, the proposed bargaining solution is compared with two other schemes, namely the Stackelberg game and the social welfare maximization schemes. Finally, extensive numerical experiments are conducted to validate the theoretical guarantees of the bargaining and to examine the impact of various model parameters. Empirical comparison indicates the fairness advantage of the bargaining approach over the other two schemes, especially when the load of the data center is not very flexible, highlighting the importance of information feedback embodied by the bargaining procedure.

To mitigate the substantial energy footprint of data centers and exploit their flexibility for low-carbon power scheduling, a low-carbon economic dispatch scheme is developed for virtual power plants (VPPs) that explicitly models the spatiotemporal distribution of data-center workloads. First, the flexibility of data center workload adjustment in temporal shifting and spatial migration is characterized and exploited, and the “source-load-storage” coordinated scheduling architecture, including wind power generation, photovoltaic power generation, energy storage systems and data center, is constructed. Second, the carbon trading mechanism and tenant satisfaction function are introduced to minimize the total operating cost, composed of energy storage, electricity transaction, and carbon trading costs. Considering the tenant satisfaction constraint, power-balance constraint, energy storage system constraint, and electricity transaction constraint, the VPP optimal scheduling model, including data center, is established. Finally, an example is analyzed based on the Institute of Electrical and Electronics Engineers 118-bus system and the operation data of an actual data center in Gansu Province. Simulation studies validate that the scheme leverages workload spatiotemporal shifting to curb system-level carbon emissions and total operating expenditure, while maintaining tenant quality of service and markedly enhancing renewable-energy utilization.

To accommodate the increasing presence of volatile and intermittent renewable energy sources in power generation, independent system operators (ISO) offer opportunities for demand side regulation service (RS) so as to stabilize the grid load. These power market features allow the demand side to earn monetary credits by modulating its power consumption dynamically following an RS signal broadcast by ISO. This paper studies the capacities and benefits of a major potential demand side, the data center, to provide RS. We propose a dynamic control policy that modulates the data center power consumption in response to ISO requests by leveraging server power capping techniques and various server power states. Results demonstrate that using our policy, data centers can provide fast reserves in quantities that are substantial proportions (around 50%) of their average energy consumption, with no major deterioration in quality of service (QoS). By doing so, data centers decrease their energy costs around 50%, while providing the ISOs and the society in general with cost effective demand side reserves that render massive renewable generation adoption affordable.

This paper focuses on alternative highly energy-efficient data center configurations with air-cooled servers that eliminate the use of chillers. In the first method, outdoor air is directly pumped through one side of an indirect air-to-air heat exchanger with air flow through the other side being passed through the data center to cool the air-cooled server racks. This method is called Indirect Air Side Economizer (IASE). In the second method, an external wet cooling tower or a dry cooler is used to reject the data center heat load to the ambient from a circulating liquid loop that is routed inside the data center for use inside a heat exchanger cooled rack. In this method, an air-to-liquid heat exchanger is used at the rack level to transfer the rack heat load dissipated by the servers from the exhaust air stream into the circulating water loop. In some cases, the external dry cooler may be enhanced to include evaporative cooling of outdoor air to reduce its' dry bulb temperature to be closer to the outdoor wet bulb temperature. These methods are called Dry Cooler with Rack Heat Exchanger (DRHx) and Evaporative Dry Cooler with Rack Heat Exchanger (EDRHx). This paper discusses these data center configurations in the context of their cooling power use and realizable energy savings compared with traditional chiller plant based cooling loops. Hour by hour outdoor air temperature data for a typical summer day in conjunction with simple transfer functions are used to predict the server inlet air temperatures with the use of different cooling configurations. Power use at the electronic equipment is also discussed.

Designing an optimal Demand Response (DR) program necessitates having information on the utility of users. However, that is not easily accessible since the users may not be willing to report their personal information. In this paper, we propose an incentive-based DR mechanism for Commercial and Industrial (C <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\&amp;$</tex-math> </inline-formula> I) customers, leveraging contract theory to address the challenges of incomplete information and ensure the participation of customers in the DR program. Designing a DR mechanism includes specifying an incentive reward, penalty, and the corresponding demand reduction as functions of the customer’s utility information. These functions are obtained through an optimization problem that maximizes the grid operator’s profit, ensures truthful reporting of personal information and participation of the customers in the DR program. After formulating the optimal mechanism, the main technical challenge is reformulating the obtained non-convex and computationally inefficient optimization problem to a tractable and convex one. The extensive numerical evaluations demonstrate the effectiveness of our proposed framework in comparison to existing benchmark schemes. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Imbalances between supply and demand, particularly during the hot summer months, can result in issues such as power outages. This problem is detrimental for both grid operators and customers alike. DR programs have been implemented in numerous countries to address these challenges. C <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\&amp;$</tex-math> </inline-formula> I customers, due to the schedulable nature of their loads, make for ideal targets for these programs. However, the main obstacle to implementing such programs is customer participation. Incentive rewards offered in these programs for load reduction are fixed for all customers, discouraging those with low rewards from participating and imposing high costs on grid operators for those with high rewards. On the other side, designing unique rewards and penalties for each customer necessitates information about their preferences, which grid operators cannot access. In this paper, we propose an incentive mechanism based on the contract between the grid operator and C <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\&amp;$</tex-math> </inline-formula> I customers that ensures customers truthfully report their preferences while simultaneously maximizing grid operator profit and guaranteeing the participation of all customers. The proposed DR program can be implemented for large non-residential customers in order to reduce the peak load of the grid.

A market-oriented incentive mechanism for emergency demand response in colocation data centers

Demand Response (DR) policies define the interactions between an energy supplier and its consumers and allow for customer energy regulation given a supplier request. Given the high flexibility and controllability of Data Centers (DC), they are promising candidates to participate in DR for power grid stabilization. In this work, we consider the setting where an energy supply deficit event occurs and must be addressed to avoid grid strain. We present two novel frameworks for DR, where a load aggregator offers price incentives to a set of consumer DCs so they can dynamically adjust their electricity consumption and provide DR to the grid via server usage reductions. Modeling DCs using realistic cost functions based on Quality of Service (QoS) requirements of the DC workloads, we present a data-driven inverse optimization method to estimate DC cost function parameters for precise and efficient pricing and provide an algorithm for solving the inverse problem. Experimental results on two test cases demonstrate the benefits of our proposed DR mechanisms for energy control.

Charging performance of Plug-in Hybrid Electric Vehicles (PHEVs) under various charging approaches brings new challenges for the distribution grid, such as feeder overloading and increment of loss. In this paper, the impact of charging PHEVs on the load profile is evaluated considering Demand Response (DR) programs in order to minimize the impact of charging PHEVs on load profile. In the proposed DR strategy, all consumers can determine and control their loads. Also, the proposed DR strategy can satisfy the function of reducing peak with regard to different PHEV penetration, while preserving consumer comfort levels. For performing DR programs, time of use and real time pricing models are used to evaluate the impacts of charging of PHEVs on distribution system operations. A 33-node test feeder has been studied to assess different penetration schemes of PHEVs including 11.3%, 35% and 45% PHEV penetration in residential loads. Moreover, uncertainties associated with PHEV charging performances are handled with Monte Carlo simulation. Results validate that the amount of peak loss, and Peak to Average Ratio (PAR) increase owing to PHEVs penetration growth. Also, the results illustrate that scheduling for the operation of PHEVs with DR program reduces the distribution losses and PAR.

This paper models a novel demand response (DR) trading strategy. In this model, the DR aggregator obtains the DR from the end-users via two types of DR programs, i.e. a time-of-use (TOU) program and an incentive-based DR program. Then, it offers this DR to the wholesale market. Three consumer sectors, namely residential, commercial and industrial, are included in this problem. The DR program is dependent on their corresponding load profiles during the studied time horizon. This paper uses a mixed-integer linear programming (MILP) problem and it is solved using the CPLEX solver through a stochastic programming approach in GAMS. The risk measure chosen to represent the load uncertainty of the users who are participating in the DR program is Conditional Value-at-Risk (CVaR). The proposed problem is simulated and assessed through a case study of a test system. The results indicate that the industrial loads play a major role in the power system and this directly affects the DR program. Moreover, the risk-averse decision-maker in this model favors a reduced participation in the DR programs when compared to a decision-maker who is risk-neutral, since the risk-averse decision maker prefers to be more secure against uncertainties. In other words, an increase in risk factor results in a decrease in the participation rate of the consumers in DR programs.