A multi-agent game method considering uncertainty of data center demand response adjustment capability

Abstract

The present application relates to the technical field of data center demand response, solves the technical problem of uncertain data center load regulation capacity, and particularly relates to a multi-agent game method considering uncertain data center demand response regulation capacity, comprising: constructing a data center user load optimization model; solving a microgrid operator optimization model to obtain a microgrid operator optimization strategy, and solving a non-data center user load optimization model to obtain a non-data center user load optimization strategy; inputting the non-data center user optimization strategy and the microgrid optimization strategy into the data center user load model for solving. The present application supports the participation of data centers in demand response through demand side resources of microgrid parks and non-data center user load regulation, solves the problem of uncertain data center load regulation capacity, thereby improving the flexibility of the power system, ensuring the safe and stable operation of the power system, and promoting renewable energy power consumption.

Description

Technical Field

[0001] This invention relates to the field of data center demand response technology, and in particular to a multi-agent game method that considers the uncertainty of data center demand response adjustment capabilities. Background Technology

[0002] In research and applications related to demand response in data centers, existing technologies mainly analyze and model the uncertainties affecting data center load regulation capabilities, employing algorithms such as stochastic optimization, robust optimization, and fuzzy optimization for control, ensuring that the formulated load optimization strategies can still be implemented normally under the influence of uncertainties. Alternatively, they improve prediction accuracy by forecasting data center workload and load, ensuring that the formulated load optimization strategies have small errors compared to actual execution, minimizing deviations caused by uncertainties. Another approach is to use real-time methods to optimize and adjust data center load. However, none of these methods consider addressing the uncertainty of data center load regulation capabilities through resource coordination within the microgrid or region where the data center is located. This results in optimization strategies failing to be implemented normally under uncertainties, seriously affecting the safe and stable operation of the power system. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a multi-agent game theory method that considers the uncertainty of data center demand response adjustment capabilities. This method solves the technical problem that traditional methods have not considered resource coordination through the micronet or region where the data center is located to address the uncertainty of data center load adjustment capabilities.

[0004] To address the aforementioned technical problems, this invention provides the following technical solution: a multi-agent game theory method considering the uncertainty of data center demand response adjustment capabilities, comprising the following steps:

[0005] S1. Construct a data center user load optimization model. The data center user load optimization model includes the data center optimization objective function, data center new workload constraints, data center workload model, and data center server quantity constraint model.

[0006] S2. Construct a microgrid operator optimization model and solve the model to obtain the microgrid operator optimization strategy, including the discharge power of the energy storage device. Charging power of energy storage devices and the output of the gas turbine

[0007] S3. Construct a non-data center user load optimization model to solve for the non-data center user load optimization strategy, including adjustable power. And whether to participate in demand response

[0008] S4. Construct a two-level iterative multi-agent game solving algorithm to obtain the optimal unit subsidy. and subsidies for the best unit Corresponding microgrid operator optimization strategies and non-data center user load optimization strategies.

[0009] Furthermore, in step S1, the data center optimization objective function is:

[0010]

[0011] In the above formula, C DC_DR Let T be a variable representing the additional operating costs incurred during the response period due to the increased workload; DR The constant represents the demand response period; , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; Δt is a variable representing the energy consumption of the data center due to handling new workloads; Δt is a constant representing the time interval.

[0012] New workload constraints in the data center include: latency constraints for new interactive workloads in the data center, and processing time constraints for new batch workloads in the data center.

[0013] The data center workload model is as follows:

[0014]

[0015] The data center server quantity constraint model is as follows:

[0016]

[0017] In the above formula, N t The constant represents the number of servers in the data center; Let be a variable representing the number of servers that the data center plans to activate for newly added batch processing workloads during time period t; Let be a variable representing the number of servers activated by the data center during time period t for the newly added interactive workload plan;

[0018]

[0019] In the above formula, N MAX This is a constant representing the total number of servers in the data center; The data center is processing the number of servers activated for workloads as originally planned.

[0020] Furthermore, the energy consumption of the data center due to processing the increased power demand. The following relationship must be satisfied:

[0021]

[0022] In the formula, γ is a constant, representing the energy consumption coefficient of the data center; N t Let λ be a variable representing the number of servers activated by the data center for new workload plans during time period t; t Let P be a variable representing the new workload that the data center needs to handle during time period t; PEAK P is a constant representing the peak power of servers in the data center. IDLE is a constant representing the idle power of the servers in the data center; μ is a constant representing the unit rate at which the data center processes workloads.

[0023] Furthermore, the processing latency of the newly added interactive workloads in the data center meets the service level agreement constraints, namely:

[0024]

[0025] In the above formula, T is a constant representing the maximum processing latency that a data center can tolerate during demand response periods for interactive workloads; TOL is a variable representing the processing latency reserved by the data center for interactive workloads; Let be a variable representing the number of servers activated by the data center for the new interactive workload plan during time period t; This is a constant representing the interactive workload that the data center needs to process after adding new workloads in time period t;

[0026] The processing time for newly added batch processing workloads in the data center must meet the following constraints:

[0027]

[0028] In the above formula, is a constant, representing the i-th batch processing workload added by the data center during the demand response period; Let be a variable, representing the workload that the data center needs to process for the i-th batch during time period t; is a constant representing the time when the i-th batch workload initiates a computation request; is a constant, representing the time when the calculation of the workload of the i-th batch is completed; Let i be a variable representing the batch workload that the data center needs to process after adding new workloads in time period t; i∈I, where I represents the total amount of batch workload. Let be a variable representing the number of servers that the data center plans to activate for new batch processing workloads during time period t.

[0029] Furthermore, in step S2, the microgrid operator optimization model includes: the microgrid operator optimization objective function, the microgrid power balance model, the energy storage operation model, and the distributed power source operation model.

[0030] Furthermore, the optimization objective function of the microgrid operator is:

[0031]

[0032] In the above formula, C TOTAL Let be a variable representing the increased operating costs incurred by microgrid operators during demand response; C(·) represents the cost function, where This indicates the costs incurred due to the participation of non-data center users in demand response; This indicates the cost incurred in adjusting the output of the energy storage device; This indicates the cost incurred from adjusting the output of distributed power sources;

[0033] The power balance model for microgrids is as follows:

[0034]

[0035] In the above formula, Let be a variable representing the energy consumption of the data center due to handling new workloads;

[0036] The energy storage operation model is as follows:

[0037]

[0038] In the above formula, Both are constants, representing the charging and discharging power of the energy storage during the planned discharge in time period t; ΔQ is a constant, representing the maximum charge / discharge power of the energy storage system during the originally planned discharge period t; t Let η be a variable, representing the change in the energy storage device's charge / discharge power after adjusting the charging / discharging power during time period t; CH η is a constant representing the charging efficiency of the energy storage device; DCH Q is a constant representing the discharge efficiency of the energy storage device; MAX Q is a constant representing the maximum capacity of the energy storage device; MIN is a constant, representing the minimum capacity of the energy storage device; is a constant, representing the amount of energy stored in time period t after the energy storage is charged and discharged according to the original plan;

[0039] The distributed power supply operation model is as follows:

[0040]

[0041] In the above formula, PDMIN P is a constant representing the minimum output of the gas turbine. DMAX is a constant, representing the maximum processing capacity of the gas turbine.

[0042] Furthermore, in step S3, the non-data center user load optimization model includes a non-data center user optimization objective function and a non-data center user incentive model.

[0043] The optimization objective function for non-data center users is:

[0044]

[0045] In the above formula, C NDC θ is a variable representing the additional operating costs for non-mobile users; θ is a parameter, ranging from 0 to 1, representing the unit demand response subsidy ratio allocated by the microgrid operator to non-data center users. , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; π is a variable representing the load adjustment amount of non-data center users estimated by the microgrid operator in time period t; π is a constant representing the additional cost coefficient for non-data center users to temporarily participate in demand response. Auxiliary variables introduced;

[0046] The incentive model for non-data center users is as follows:

[0047]

[0048]

[0049] In the above formula, x m,n,t is a 0-1 variable, indicating whether user m is in the nth gear during time period t; M is a constant, representing a maximum value; is a constant, representing the adjustable lower limit of power when user m is in the nth gear during time period t; is a constant, representing the adjustable power limit when user m is in the nth gear during time period t; is a constant, representing the lower limit of the incentive given to user m at the nth level during time period t; is a constant, representing the lower limit of the incentive given to user m at the nth level during time period t; These are auxiliary variables introduced.

[0050] Furthermore, in step S4, the specific process includes the following steps:

[0051] S41. Initialize variable k = 0, where k is a variable used to find the optimal unit subsidy and represents the kth iteration calculation.

[0052] S42, Randomly initialize unit subsidies Let the variable Based on the data center user load optimization model, the energy consumption is obtained by solving the data center optimization objective function.

[0053] S43, Subsidies from the unit Substituting into step S3, based on the non-data center user load optimization model, we solve for the non-data center user optimization objective function to obtain the non-data center user load optimization strategy. and

[0054] S44. Based on the non-data center user load optimization strategy and Calculate Substituting the microgrid operator optimization model from step S2, let Using variables, we solve the microgrid operator optimization objective function to obtain... Microgrid operator optimization strategy And let k = k + 1,

[0055] in,

[0056]

[0057] S45. Determining Unit Subsidies With the best unit subsidy Is the difference less than ∈, where ∈ is a local minimum?

[0058] If so, output the optimal unit subsidy. and subsidies for the best unit Corresponding microgrid operator optimization strategies and non-data center user load optimization strategies;

[0059] If not, return to step S42 and recalculate.

[0060] By employing the above technical solution, the present invention provides a multi-agent game theory method that considers the uncertainty of data center demand response adjustment capabilities, which has at least the following beneficial effects:

[0061] 1. This invention addresses the uncertainty of data center load regulation capacity by coordinating resources within the microgrid or region where the data center is located. It can eliminate the impact of uncertainties in data center load regulation capacity, ensuring that microgrid operator optimization strategies and non-data center user load optimization strategies can still be implemented normally even when affected by uncertainties. This achieves power consumption pattern optimization, continuously improves grid stability, and enhances the utilization level of renewable energy in the power system.

[0062] 2. This invention supports data centers' participation in demand response by adjusting demand-side resources in microgrid parks and loads from non-data center users, thereby solving the uncertainty problem of data center load adjustment capabilities, improving power system flexibility, ensuring the safe and stable operation of the power system, and promoting the consumption of renewable energy power.

[0063] 3. This invention can realize demand-side resource control for data centers, microgrids, and non-data center users in a specific region. On the one hand, it can solve the problem of changes in demand response adjustment capability of data centers due to workload uncertainty. By adjusting the load of other non-data center users, it can support the normalized participation of data centers in demand response. On the other hand, it models the user load adjustment capability and demand response unit subsidy. Attached Figure Description

[0064] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0065] Figure 1 This is a flowchart of the multi-agent game theory method of the present invention;

[0066] Figure 2 This is a schematic diagram illustrating the relationship between load regulation and incentive magnitude for non-data center users according to the present invention. Detailed Implementation

[0067] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0068] In the context of "East-to-West Computing," data center computing power, through the optimized allocation of computing resources, deeply participates in power system operation planning, aiming to improve the efficiency of power operation across different regions. Demand response is a crucial aspect of power grid planning and operation. Data centers' participation in demand response not only provides them with additional economic subsidies to optimize electricity consumption patterns but also sustainably improves grid stability and enhances the power system's utilization of renewable energy.

[0069] Research and applications related to data centers' participation in demand response can be categorized into three models: First, modeling the distribution characteristics of data center workloads to construct workload scheduling and load regulation models, and proposing corresponding methods, devices, and systems to support optimized data center load participation in demand response; second, modeling the distribution network and microgrid resources in the data center's region, using planning theory to optimize the data center load day-ahead and intraday, and proposing corresponding methods, devices, and systems to support optimized data center load participation in demand response; and third, predicting the data center load and workload before demand response, constructing a predictive analysis model, forming a load demand response regulation capacity evaluation model, and proposing corresponding methods, devices, and systems to support optimized data center load participation in demand response.

[0070] All three models take into account the uncertainty of the load adjustment capability of data centers during demand response. The basic approach is to analyze and model the uncertain factors affecting the load adjustment capability of data centers, and use algorithms such as stochastic optimization, robust optimization, and fuzzy optimization for control.

[0071] When data center loads participate in demand response, the regulation capacity is uncertain, which may cause the load regulation amount to fail to meet the threshold specified by the power grid, thus affecting the normal operation of the power grid demand response.

[0072] Existing technologies primarily analyze and model the uncertainties affecting data center load regulation capabilities, employing algorithms such as stochastic optimization, robust optimization, and fuzzy optimization for control, ensuring that the formulated load optimization strategies can still be implemented normally under the influence of uncertainties. Alternatively, they improve prediction accuracy by forecasting data center workload and load, ensuring that the error between the formulated load optimization strategy and the actual execution is small, minimizing deviations caused by uncertainties. Another approach is to use real-time methods to optimize and adjust data center load. However, none of these methods consider addressing the uncertainty of data center load regulation capabilities through resource coordination within the micronet or region where the data center is located.

[0073] In this embodiment:

[0074] Demand response refers to measures primarily based on economic incentives to guide electricity users to voluntarily adjust their electricity consumption behavior according to the needs of the power system, thereby addressing situations such as short-term power supply and demand tensions and difficulties in the consumption of renewable energy. This aims to achieve peak shaving and valley filling, improve the flexibility of the power system, ensure the safe and stable operation of the power system, and promote the consumption of renewable energy.

[0075] Demand-side resources refer to power resources that are widely distributed on the user side, such as adjustable loads, distributed power sources, and new energy storage, which can be aggregated, optimized, and participate in the operation and regulation of the power system.

[0076] To address the uncertainty of data center load regulation capabilities through resource coordination within the micronet or region where the data center is located, please refer to... Figure 1 This embodiment proposes a multi-agent game theory method that considers the uncertainty of data center demand response adjustment capabilities. This method can eliminate the impact of uncertainties in data center load adjustment capabilities, ensuring that microgrid operator optimization strategies and non-data center user load optimization strategies can still be implemented normally under the influence of uncertainty. This achieves power consumption pattern optimization, sustainably improves grid stability, and enhances the power system's utilization of renewable energy. The method includes the following steps:

[0077] S1. Construct a data center user load optimization model. The data center user load optimization model includes the data center optimization objective function, data center new workload constraints, data center workload model, and data center server quantity constraint model.

[0078] The objective function for data center optimization is:

[0079] The data center optimization objective function, which aims to minimize additional operating costs for data center operators, is as follows:

[0080]

[0081] In the above formula, C DC_DR Let T be a variable representing the additional operating costs incurred during the response period due to the increased workload; DR The constant represents the demand response period; , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; Δt is a variable representing the energy consumption of the data center due to handling new workloads; Δt is a constant representing the time interval.

[0082] Among them, the energy consumption of data centers in handling new power demand. The following relationship must be satisfied:

[0083]

[0084] In the formula, γ is a constant, representing the energy consumption coefficient of the data center; N t Let λ be a variable representing the number of servers activated by the data center for new workload plans during time period t; t Let P be a variable representing the new workload that the data center needs to handle during time period t; PEAK P is a constant representing the peak power of servers in the data center. IDLE is a constant representing the idle power of the servers in the data center; μ is a constant representing the unit rate at which the data center processes workloads.

[0085] New workload constraints in the data center include: latency constraints for new interactive workloads and duration constraints for new batch workloads, as follows:

[0086] The latency of newly added interactive workloads in the data center meets the service level agreement constraints, namely:

[0087]

[0088] In the above formula, T is a constant representing the maximum processing latency that a data center can tolerate during demand response periods for interactive workloads; TOL is a variable representing the processing latency reserved by the data center for interactive workloads; Let be a variable representing the number of servers activated by the data center for the new interactive workload plan during time period t; is a constant representing the interactive workload that the data center needs to handle after adding new workloads in time period t.

[0089] The processing time for newly added batch processing workloads in the data center must meet the following constraints:

[0090]

[0091] In the above formula, is a constant, representing the i-th batch processing workload added by the data center during the demand response period; Let be a variable, representing the workload that the data center needs to process for the i-th batch during time period t; is a constant representing the time when the i-th batch workload initiates a computation request; is a constant, representing the time when the calculation of the workload of the i-th batch is completed; Let i be a variable representing the batch workload that the data center needs to process after adding new workloads in time period t; i∈I, where I represents the total amount of batch workload. Let be a variable representing the number of servers that the data center plans to activate for new batch processing workloads during time period t;

[0092] Data center workload model:

[0093] The data center can calculate the batch processing workload and interactive workload that need to be processed during the demand response period, and the amount of new workload λ that the data center needs to process. t The constraints are:

[0094]

[0095] The data center server quantity constraint model is as follows:

[0096]

[0097] In the above formula, N t The constant represents the number of servers in the data center; Let be a variable representing the number of servers that the data center plans to activate for newly added batch processing workloads during time period t; Let be a variable representing the number of servers activated by the data center during time period t for newly added interactive workload plans.

[0098]

[0099] In the above formula, N MAX This is a constant representing the total number of servers in the data center; The data center is processing the number of servers activated for workloads as originally planned.

[0100] This invention enables demand-side resource control for data centers, microgrids, and non-data center users in a specific region. On the one hand, it can solve the problem of changes in demand response adjustment capabilities of data centers due to workload uncertainty, and support the normalized participation of data centers in demand response through load adjustment by other non-data center users. On the other hand, it models the user load adjustment capabilities and demand response unit subsidies.

[0101] S2. Construct a microgrid operator optimization model and solve the microgrid operator optimization model to obtain the microgrid operator optimization strategy. in, Let be a variable, representing the magnitude of the discharge power adjusted by the energy storage device during time period t; Let be a variable, representing the amount of charging power adjusted by the energy storage device during time period t; The output of the gas turbine is adjusted during time period t. In this embodiment, the microgrid operator optimization objective function is used as the optimization objective, and the microgrid power balance model, energy storage operation model and distributed power source operation model are used as constraints. The microgrid operator optimization model can be solved by conventional mixed integer linear programming or other methods. This method is an existing technology and will not be described in detail here.

[0102] The microgrid operator optimization model includes: the microgrid operator optimization objective function, the microgrid power balance model, the energy storage operation model, and the distributed power source operation model.

[0103] In this embodiment, the microgrid operator's optimization objective function is:

[0104] The optimization goal for microgrid operators is to minimize additional operating costs, namely:

[0105]

[0106] In the above formula, C TOTAL Let be a variable representing the increased operating costs incurred by microgrid operators during demand response; C(·) represents the cost function, where This indicates the costs incurred due to the participation of non-data center users in demand response; This indicates the cost incurred in adjusting the output of the energy storage device; This indicates the cost incurred in adjusting the output of distributed power sources.

[0107] Specifically, cost The definition is as follows:

[0108]

[0109] In the above formula, θ is a parameter that takes a value between 0 and 1, representing the unit demand response subsidy ratio allocated by the microgrid operator to non-data center users; , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; Let be the variable representing the estimated load adjustment amount for non-data center users during time period t; q DCH is a constant, representing the unit discharge cost of the energy storage device; Let q be a variable, representing the magnitude of the discharge power adjusted by the energy storage device during time period t; CH is a constant, representing the unit discharge cost of the energy storage device; Let be a variable, representing the amount of charging power adjusted by the energy storage device during time period t; This refers to the output of the gas turbine adjusted during time period t. Δt represents the planned output of the gas turbine during time period t; a and b are cost coefficients of the gas turbine; Δt is a constant representing the time interval.

[0110] The power balance model for microgrids is as follows:

[0111] Microgrid operators need to ensure the power balance of the microgrid where the data center is located, thus satisfying the constraints of the microgrid power balance model:

[0112]

[0113] In the above formula, Let be a variable representing the energy consumption of the data center due to handling new workloads.

[0114] The energy storage operation model is as follows:

[0115] Energy storage within the microgrid park must meet the following operational constraints:

[0116]

[0117] In the above formula, Both are constants, representing the charging and discharging power of the energy storage during the planned discharge in time period t; ΔQ is a constant, representing the maximum charge / discharge power of the energy storage system during the originally planned discharge period t; t Let η be a variable, representing the change in the energy storage device's charge / discharge power after adjusting the charging / discharging power during time period t; CH η is a constant representing the charging efficiency of the energy storage device; DCH Q is a constant representing the discharge efficiency of the energy storage device; MAX Q is a constant representing the maximum capacity of the energy storage device; MIN is a constant, representing the minimum capacity of the energy storage device; is a constant, representing the amount of energy stored during time period t after the energy storage has been charged and discharged according to the original plan.

[0118] The distributed power supply operation model is as follows:

[0119] Distributed power generation mainly uses small gas turbines. The output constraints of the gas turbines after adjusting their power generation capacity are as follows:

[0120]

[0121] In the above formula, P DMIN P is a constant representing the minimum output of the gas turbine. DMAX is a constant, representing the maximum processing capacity of the gas turbine.

[0122] This embodiment supports data center participation in demand response by adjusting demand-side resources in the microgrid park and loads from non-data center users, thus addressing the uncertainty of data center load adjustment capabilities, improving power system flexibility, ensuring the safe and stable operation of the power system, and promoting the consumption of renewable energy power.

[0123] S3. Construct a non-data center user load optimization model to solve for the non-data center user load optimization strategy. and in, Let m be a variable, representing the adjustable power of user m during time period t; The variable is 0-1, indicating whether user m participates in demand response during time period t. In this embodiment, the optimization objective function for non-data center users is taken as the optimization objective, and the incentive model for non-data center users is taken as the constraint. The non-data center user load optimization model can be solved by conventional mixed integer linear programming or other methods. This method is an existing technical means and will not be described in detail here.

[0124] The non-data center user load optimization model includes the non-data center user optimization objective function and the non-data center user incentive model.

[0125] The optimization objective function for non-data center users is:

[0126] The optimization goal for non-data center users is to determine their own pricing based on the data center pricing provided by the microgrid operator, and to decide whether to participate in demand response load reduction to help the data center complete its load reduction task. That is:

[0127]

[0128] In the above formula, C NDC θ is a variable representing the additional operating costs for non-mobile users; θ is a parameter, ranging from 0 to 1, representing the unit demand response subsidy ratio allocated by the microgrid operator to non-data center users. , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; Let be a variable representing the load adjustment amount of non-data center users estimated by the microgrid operator during time period t; Let m be a variable, representing the adjustable power of user m during time period t; π is a 0-1 variable, representing whether user m participates in demand response during time period t; π is a constant, representing the additional cost coefficient for non-data center users to temporarily participate in demand response.

[0129] Because the objective function for non-data center users involves the multiplication of two variables, i.e. This can easily increase the difficulty of solving the problem. Therefore, in order to reduce the difficulty of solving the load optimization strategy for non-data center users, it is necessary to transform the objective function for non-data center users. After transformation, it becomes:

[0130]

[0131] In the above formula, M is a constant, representing a maximum value.

[0132] The optimization goal for non-data center users is to maximize C NDC The final optimization objective function for non-data center users is expressed as follows:

[0133]

[0134] In the above formula, C NDC Let be a variable representing the additional operating costs for non-mobile users. θ is a parameter, ranging from 0 to 1, representing the proportion of demand response subsidies allocated by the microgrid operator to non-data center users per unit. , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; π is a variable representing the load adjustment amount of non-data center users estimated by the microgrid operator in time period t; π is a constant representing the additional cost coefficient for non-data center users to temporarily participate in demand response. These are auxiliary variables introduced.

[0135] The incentive model for non-data center users is as follows:

[0136] The load regulation for non-data center users is related to the incentive magnitude; please refer to [reference needed]. Figure 2 The excitation in the diagram is divided into multiple levels. Once the excitation reaches a certain range, the load regulation capacity also reaches the corresponding range.

[0137] The relationship between data center user load regulation and incentives is as follows:

[0138]

[0139] In the above formula, x m,n,t is a 0-1 variable, indicating whether user m is in the nth gear during time period t; M is a constant, representing a maximum value; is a constant, representing the adjustable lower limit of power when user m is in the nth gear during time period t; q is a constant, representing the adjustable power limit when user m is in the nth gear during time period t; m,t is a constant, representing the incentive given to user m during time period t; is a constant, representing the lower limit of the incentive given to user m at the nth level during time period t; is a constant, representing the lower limit of the incentive given to user m at the nth level during time period t.

[0140] S4. Construct a two-level iterative multi-agent game solving algorithm to obtain the optimal unit subsidy. and subsidies for the best unit The corresponding microgrid operator optimization strategy and non-data center user load optimization strategy. The specific process includes the following steps:

[0141] S41. Initialize variable k = 0, where k is a variable used to find the optimal unit subsidy and represents the kth iteration calculation.

[0142] S42, Randomly initialize unit subsidies Let the variable Based on the data center user load optimization model, the energy consumption is obtained by solving the data center optimization objective function.

[0143] S43, Subsidies from the unit Substituting into step S3, based on the non-data center user load optimization model, we solve for the non-data center user optimization objective function to obtain the non-data center user load optimization strategy. and

[0144] S44. Based on the non-data center user load optimization strategy and Calculate Substituting the microgrid operator optimization model from step S2, let Using variables, we solve the microgrid operator optimization objective function to obtain... Microgrid operator optimization strategy And let k = k + 1,

[0145] in,

[0146]

[0147] S45. Determining Unit Subsidies With the best unit subsidy Is the difference less than ∈, where ∈ is a local minimum? If so, output the optimal unit subsidy. and subsidies for the best unit The corresponding microgrid operator optimization strategy and non-data center user load optimization strategy; if not, return to step S42 and recalculate.

[0148] This embodiment can realize demand-side resource control for data centers, microgrids, and non-data center users in a specific region. On the one hand, it can solve the problem of changes in demand response adjustment capability of data centers due to workload uncertainty. By adjusting the load of other non-data center users, it can support the normalized participation of data centers in demand response. On the other hand, it models the user load adjustment capability and demand response unit subsidy.

[0149] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0151] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A multi-agent game theory method considering the uncertainty of data center demand response adjustment capabilities, characterized in that, The method includes the following steps: S1. Construct a data center user load optimization model. The data center user load optimization model includes the data center optimization objective function, data center new workload constraints, data center workload model, and data center server quantity constraint model. S2. Construct a microgrid operator optimization model and solve the microgrid operator optimization model to obtain the discharge power of the energy storage device. The charging power of the energy storage device and the output of the gas turbine Microgrid operator optimization strategies; The microgrid operator optimization model includes a microgrid operator optimization objective function, a microgrid power balance model, an energy storage operation model, and a distributed power source operation model. The objective function for the microgrid operator is: In the above formula, Let be a variable representing the increased operating costs incurred by microgrid operators during demand response; Represents the cost function, where This indicates the costs incurred due to the participation of non-data center users in demand response; This indicates the cost incurred in adjusting the output of the energy storage device; This indicates the cost incurred from adjusting the output of distributed power sources; The microgrid power balance model is as follows: In the above formula, Let be a variable representing the energy consumption of the data center due to handling new workloads; The energy storage operation model is as follows: In the above formula, , All are constants, representing the time periods during which energy storage is performed as originally planned. The charging and discharging power during discharge; , The constant indicates that energy storage will proceed as planned during the specified time period. The maximum charge / discharge power for discharging; Let be a variable, representing the energy storage device during the time period. Changes in charge level after adjusting charging and discharging power; is a constant, representing the charging efficiency of the energy storage device; is a constant, representing the discharge efficiency of the energy storage device; is a constant, representing the maximum capacity of the energy storage device; is a constant, representing the minimum capacity of the energy storage device; The constant represents the energy storage capacity during the time period after it is charged and discharged according to the original plan. The amount of electricity; The distributed power supply operation model is as follows: In the above formula, is a constant, representing the minimum output of the gas turbine; The constant represents the maximum output of the gas turbine; S3. Construct a non-data center user load optimization model, and solve the non-data center user load optimization model to obtain adjustable power. And whether to participate in demand response Non-data center user load optimization strategies; S4. Construct a two-level iterative multi-agent game solving algorithm to obtain the optimal unit subsidy. and subsidies for the best unit Corresponding microgrid operator optimization strategies and non-data center user load optimization strategies.

2. The multi-agent game method according to claim 1, characterized in that, In step S1, the objective function for data center optimization is: In the above formula, Let be a variable representing the additional operating costs incurred due to the new workload during the demand response period; The constant represents the demand response period; , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; Let be a variable representing the energy consumption of the data center due to handling new workloads; The constant represents the time interval; New workload constraints in the data center include: latency constraints for new interactive workloads in the data center, and processing time constraints for new batch workloads in the data center. The data center workload model is as follows: The data center server quantity constraint model is as follows: In the above formula, The constant represents the number of servers in the data center; Let be a variable, representing the time period of the data center. The number of servers to be enabled for the newly added batch processing workload plan; Let be a variable, representing the time period of the data center. The number of servers enabled for the new interactive workload plan; In the above formula, This is a constant representing the total number of servers in the data center; The data center is processing the number of servers activated for workloads as originally planned.

3. The multi-agent game method according to claim 2, characterized in that, The data center consumes energy due to handling increased power demand. The following relationship must be satisfied: In the formula, is a constant, representing the energy consumption coefficient of the data center; Let be a variable, representing the time period of the data center. The number of servers to be activated for the new workload plan; Let be a variable, representing the time period of the data center. The new workload that needs to be handled; This is a constant representing the peak power of the servers in the data center; This is a constant representing the idle power of servers in the data center; is a constant representing the unit rate at which the data center processes workloads.

4. The multi-agent game method according to claim 2, characterized in that, The latency of the newly added interactive workloads in the data center meets the service level agreement constraints, namely: In the above formula, This is a constant representing the maximum processing latency that the data center can tolerate during demand response for interactive workloads. is a variable representing the processing latency reserved by the data center for interactive workloads; Let be a variable, representing the time period of the data center. The number of servers enabled for the newly added interactive workload plan; The constant represents the time period of the data center. Interactive workloads that need to be handled after adding new workloads; The processing time for newly added batch processing workloads in the data center must meet the following constraints: In the above formula, The constant represents the number of new data centers added during the demand response period. Individual batch processing workloads; Let be a variable, representing the time period of the data center. Need to process the first Batch processing workload; Let be a constant, representing the th The time it takes for a batch processing workload to initiate a calculation request; Let be a constant, representing the th The time when the batch processing workload calculation ends; Let be a variable, representing the time period of the data center. The batch processing workload that needs to be processed after adding new workloads; , Indicates the total workload of batch processing; Let be a variable, representing the time period of the data center. The number of servers to be activated for the newly added batch processing workload plan.

5. The multi-agent game method according to claim 1, characterized in that, In step S3, the non-data center user load optimization model includes the non-data center user optimization objective function and the non-data center user incentive model; The optimization objective function for non-data center users is: In the above formula, Let be a variable representing the additional operating costs for non-data center users; The parameter, with a value between 0 and 1, represents the proportion of unit demand response subsidy allocated by the microgrid operator to non-data center users. , where is a variable representing the unit subsidy given by the data center operator to the microgrid operator; Let be the variable, representing the microgrid operator's estimated time period. Load regulation for non-data center users; This is a constant representing the additional cost coefficient for non-data center users temporarily participating in demand response; Auxiliary variables introduced; The incentive model for non-data center users is as follows: In the above formula, 0-1 variables represent Time period users Is it in the first One gear; A constant represents a maximum value; A constant, representing Time period users In the first The adjustable lower limit of power in each gear setting; A constant, representing Time period users In the first The adjustable power limit for each gear setting; A constant, representing Time period given to users No. The lower limit of incentives for each tier; A constant, representing Time period given to users No. The incentive cap for each tier.

6. The multi-agent game method according to claim 1, characterized in that, In step S4, the specific process includes the following steps: S41. Initialize variables , Let be a variable used to find the optimal unit subsidy, representing the th . The calculation is performed in several iterations; S42, Randomly initialize unit subsidies Let the variable Based on the data center user load optimization model, the energy consumption is obtained by solving the data center optimization objective function. ; S43, Subsidies from the unit Substituting into step S3, based on the non-data center user load optimization model, we solve for the non-data center user optimization objective function to obtain the non-data center user load optimization strategy. and ; S44. Based on the non-data center user load optimization strategy and Calculate , Substituting the microgrid operator optimization model from step S2, let Using variables, we solve the microgrid operator optimization objective function to obtain... Microgrid operator optimization strategy , , and order , ; in, S45. Determining Unit Subsidies With the best unit subsidy Is the difference less than ,in, It is a local minimum; If so, output the optimal unit subsidy. and subsidies for the best unit Corresponding microgrid operator optimization strategies and non-data center user load optimization strategies; If not, return to step S42 and recalculate.

7. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the multi-agent game method according to any one of claims 1 to 6.

8. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements the multi-agent game method as described in any one of claims 1 to 6.

Images