A multi-resource hierarchical regulation method for green electricity direct connection data center

Abstract

The application discloses a kind of green electricity direct connection data center's multi-resource hierarchical regulation method, belong to energy system technical field.It includes: S1, the dispatching period is discrete into multiple continuous time periods, and determine the predicted power and reference load power of current time period;S2, whether the current time period belongs to surplus green electricity time period according to the predicted power and reference load power of current time period, if yes, execute step S3;Otherwise, execute step S4;S3, the flexible task quantity of remaining time period is migrated to surplus green electricity time period, judge whether there is still surplus surplus green electricity after migration current time period;If still there is surplus surplus green electricity, surplus surplus green electricity is converted into cooling capacity and / or cold storage capacity;S4, reduce the flexible task quantity of current time period, and release cold storage capacity.The method can solve the problem of multi-resource disorder competition in the prior art, difficult to reflect the coupled operation characteristics of data center and lack of hierarchical regulation sequence for surplus green electricity consumption.

Description

Technical Field

[0001] This invention relates to a multi-resource hierarchical regulation method for direct green electricity connection to a data center, belonging to the field of energy system technology. Background Technology

[0002] Data centers, as digital infrastructure characterized by both high computing power and high energy consumption, face the dual pressures of increasing computing power and achieving green and low-carbon development. Flexible task scheduling, cooling, and battery energy storage have become key means to improve energy efficiency and promote the consumption of green electricity. However, this field involves the functional division of various regulatory resources, and how to achieve collaborative optimization to meet computing power and security constraints remains a key technical problem that needs to be solved.

[0003] Existing technologies mostly employ optimized control methods for data center battery energy storage charging and discharging, flexible computing task migration and scheduling methods, independent optimization operation methods for cooling systems or cold storage systems, and joint optimization methods that integrate flexible tasks, cold storage systems, and energy storage systems into a single objective function for weighted solution. Among these, optimized control methods for data center battery energy storage charging and discharging typically establish mathematical programming models based on real-time information such as grid electricity prices, renewable energy output, and energy storage state of charge, aiming to minimize electricity purchase costs or maximize peak-valley arbitrage, and solve for charging and discharging commands for each time period under constraints such as capacity, power, and lifetime. Flexible computing task migration and scheduling methods identify delayable or migrateable non-real-time tasks, combine multi-node load, heat dissipation efficiency, and electricity price differences, construct integer programming or heuristic search models, and decide the task execution location and time under constraints of task deadlines and transmission bandwidth to achieve flexible load redistribution in time and space. Independent optimization operation methods for cooling systems utilize model predictive control or reinforcement learning to adjust actuators such as chillers, pumps, and fans to cool energy. Minimizing energy consumption satisfies temperature safety constraints, while the cold storage system, based on cold load and electricity price forecasts, establishes a dynamic model for cold storage / discharge and solves the cold storage plan. This allows for cold storage during off-peak hours and cold release during peak hours to reduce operating costs. A joint optimization method integrates flexible tasks, the cold storage system, and the energy storage system into a single objective function for weighted solution. Through state variables, decision variables, and cross-system coupling relationships, a comprehensive weighted objective is constructed, including grid power purchase costs, battery life loss, cooling and cold storage energy consumption costs, and task delay penalties. This objective is then solved uniformly under the physical constraints of each subsystem and coupling constraints such as power supply and power allocation. Finally, linked battery charging and discharging, cold storage operation, and task migration scheduling instructions are generated and executed collaboratively by the data center management system.

[0004] While the above methods can improve the operational economy and energy efficiency of data centers to some extent, they generally have the following shortcomings: First, they fail to clearly distinguish the functional division of flexible tasks, cooling systems, and battery energy storage during the adjustment process, which can easily lead to disorderly competition among multiple resources; second, they do not fully consider the transmission impact of IT load changes on cooling load demand, making it difficult to truly reflect the coupled operation characteristics of data centers in the "computing-cooling-power" process; and third, they lack a hierarchical adjustment sequence for the consumption of surplus green electricity, resulting in insufficient utilization of green electricity, increased redundancy of battery energy storage, and insufficient engineering interpretability of the adjustment results. Summary of the Invention

[0005] This invention provides a multi-resource hierarchical regulation method for direct green electricity connection to data centers, which can solve the problems in existing technologies such as disorderly competition among various resources, difficulty in accurately reflecting the coupled operation characteristics of data centers' "computing-cooling-power" systems, and lack of a hierarchical regulation sequence for accommodating surplus green electricity. Specifically, it includes the following steps:

[0006] S1. Discretize the scheduling cycle into multiple consecutive time periods and determine the predicted power and baseline load power for the current time period;

[0007] S2. Determine whether the current period belongs to the surplus green electricity period based on the predicted power and the baseline load power of the current period. If yes, proceed to step S3; otherwise, proceed to step S4.

[0008] S3. Migrate the flexible task load of the remaining time periods to the surplus green electricity time period, and after migration, determine whether there is still surplus green electricity in the current time period; if there is still surplus green electricity, convert the surplus green electricity into cooling capacity and / or cooling storage capacity.

[0009] S4. Reduce the amount of flexible tasks in the current period and release the stored cooling capacity.

[0010] Furthermore, after converting the surplus green electricity into cooling capacity and / or cold storage capacity, the process also includes:

[0011] The remaining surplus green electricity is used to charge the battery;

[0012] Alternatively, after reducing the amount of flexible tasks in the current period and releasing the stored cooling capacity, it also includes:

[0013] Battery discharge.

[0014] Specifically, before supplying surplus green electricity to the battery for charging, the process also includes:

[0015] Determine the binary variable of the battery charge / discharge state, and control the battery charging based on the binary variable of the battery charge / discharge state.

[0016] Alternatively, before the battery discharges, it also includes:

[0017] Determine the binary variable of the battery's charge / discharge state, and control the battery discharge based on the binary variable of the battery's charge / discharge state.

[0018] Specifically, determining whether the current period belongs to a surplus green electricity period based on the predicted power and baseline load power of the current period includes:

[0019] Determine the photovoltaic power forecast and wind power forecast for the current time period, and determine the available green electricity power for the current time period based on the photovoltaic power forecast and wind power forecast.

[0020] Determine the baseline load power for the current time period;

[0021] When the available green electricity is greater than or equal to the baseline load power, the current time period is determined to be a period of surplus green electricity.

[0022] Specifically, determining the baseline load power for the current time period includes:

[0023] Determine the base cooling load power for the current period based on the base cooling load and the performance coefficient of the refrigeration equipment for the current period.

[0024] Determine the rigid load power for the current time period;

[0025] The baseline load power for the current time period is determined based on the baseline cooling load power and the rigid load power for the current time period.

[0026] Furthermore, after migrating the flexible workload from other time periods to the surplus green electricity periods, it also includes:

[0027] Determine the net increase in the amount of flexible tasks during the period of surplus green electricity.

[0028] Specifically, converting the surplus green electricity into cooling capacity and / or cold storage capacity includes:

[0029] Determine the total load power for the current time period;

[0030] The cooling load power for the current period is determined based on the basic cooling load item, the load ratio factor, and the total load power.

[0031] The cooling power and / or cold storage power for the current period are determined based on the cooling balance relationship and the cooling load power.

[0032] Furthermore, in S3, the remaining surplus green electricity is converted into cold storage capacity, including:

[0033] Determine the binary variable of the cold storage and release status of the cold storage system for the current period, and convert the remaining surplus green electricity into cold storage capacity based on the binary variable of the cold storage and release status;

[0034] Alternatively, the stored cold energy can be released in S4, including:

[0035] Determine the binary variable of the cold storage and release status of the cold storage system for the current period, and release the stored cold amount according to the binary variable of the cold storage and release status.

[0036] Furthermore, after S3 or S4, the following is also included:

[0037] Establish the power balance relationship for the current time period;

[0038] Determine the power allocation relationship between the current green energy utilization rate and the green energy curtailment power;

[0039] The amount of green electricity used, the amount of green electricity abandoned, and the amount of electricity purchased by the grid during the current period are calculated based on the power balance relationship and the power distribution relationship.

[0040] The beneficial effects of this invention include: First, it determines the time period type based on predicted power and baseline load power. Then, during periods of surplus green electricity, it migrates flexible tasks first, and then determines the cooling capacity and cold storage capacity. During periods of insufficient green electricity, it first reduces the amount of flexible tasks and then releases the cold storage capacity, clarifying the priority of flexible tasks and the cold storage system and avoiding disorderly competition among multiple resources. After adjusting the flexible tasks, it makes decisions on cooling, cold storage, and cold release, ensuring that the cold-side adjustment is always based on changes in the total load, truly reflecting the coupled operation characteristics of the data center's "computing-cooling-power". At the same time, it also constructs an adjustment sequence oriented towards the consumption of surplus green electricity. When there is a surplus of green electricity, it prioritizes the migration and consumption of green electricity through flexible tasks, and then uses cold storage to achieve cross-time period transfer of cooling capacity. When there is a shortage of green electricity, it prioritizes reducing the amount of flexible tasks executed, and then releases the cold storage capacity to reduce cooling power consumption. This sequence makes the priority hierarchy of the adjustment strategy clear and has good engineering feasibility. Finally, it uses battery energy storage as residual compensation, effectively reducing unnecessary battery activation and wear. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating a multi-resource hierarchical adjustment method for direct green electricity connection to a data center, as provided in an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0043] This invention provides a multi-resource hierarchical regulation method for green electricity direct connection data centers, applicable to green electricity direct connection data centers including a power supply side, a load side, a cooling side, an energy storage side, and a control side. The power supply side includes photovoltaic power generation units, wind power generation units, and a public grid interface; the load side includes rigid loads, flexible tasks, and auxiliary loads; the cooling side includes cooling equipment, a cold storage system, and cooling load demand units; the energy storage side includes a battery energy storage system; and the control side includes an energy management and coordinated regulation platform. This platform receives photovoltaic predicted power data, wind power predicted power data, rigid load data, flexible task data, equipment status data, and constraint parameters for each time period, and outputs flexible task migration results, cooling equipment operation results, cold storage system cold storage and release results, battery energy storage charging and discharging results, and grid power purchase results. In practical applications, without changing the core regulation logic of this invention, other adjustable load units, other cooling-side equipment, or other auxiliary energy storage units can be further connected to the system. As long as their regulation methods still conform to the hierarchical coordinated regulation concept described in this invention, they can all be considered extended implementations of this invention.

[0044] The specific method for multi-resource hierarchical regulation of green electricity direct connection to data centers is as follows: Figure 1 As shown, it includes the following steps:

[0045] S1. Discretize the scheduling cycle into multiple consecutive time periods and determine the predicted power and baseline load power for the current time period;

[0046] In this embodiment of the invention, the scheduling period is discretized into A series of consecutive time periods, numbered as follows: The duration of each time period is The current time period can be any time period.

[0047] S2. Determine whether the current period belongs to the surplus green electricity period based on the predicted power and the baseline load power of the current period. If yes, proceed to step S3; otherwise, proceed to step S4.

[0048] Specifically, determining whether the current period belongs to a surplus green electricity period based on the predicted power and baseline load power of the current period includes:

[0049] S21. Determine the photovoltaic power forecast and wind power forecast for the current time period, and determine the available green electricity power for the current time period based on the photovoltaic power forecast and wind power forecast.

[0050] In this embodiment of the invention, the current time period is defined as a time period. Time period The total usable green electricity is shown in formula (1):

[0051] (1)

[0052] in, For time period The total available green electricity For time period The predicted photovoltaic power, For time period The predicted wind power output.

[0053] S22. Determine the baseline load power for the current time period;

[0054] Specifically, the basic cooling load power is determined based on the basic cooling load and the coefficient of performance of the refrigeration equipment for the current period.

[0055] Determine the rigid load power for the current time period;

[0056] Determine the baseline load power for the current period based on the baseline cooling load power and rigid load power for the current period.

[0057] In this embodiment of the invention, time period The baseline load power At least include time periods rigid load power Auxiliary load power and basic cooling load power Among them, the basic cooling load power The basic refrigeration power can be regulated by the base refrigeration power before migration and cold storage. Coefficient of performance of refrigeration equipment To be determined together, namely:

[0058] (2)

[0059] Therefore, time period The baseline load power It can be represented as:

[0060] (3)

[0061] S23. When the available green electricity is greater than or equal to the baseline load power, the current time period is determined to be a period of surplus green electricity.

[0062] In an embodiment of the present invention, when At that time, the current time period The period is identified as a surplus of green electricity; when At that time, the current time period The periods identified as having insufficient green electricity are thus defined as periods with surplus green electricity. and periods of insufficient green electricity The surplus green electricity time period is used as a priority adjustment criterion for subsequent flexible task migration, early production of cold storage, and energy storage to absorb surplus green electricity.

[0063] In practical applications, in addition to comparing the available green electricity power in the current period with the benchmark load power, we can also combine the green electricity forecast trend, rolling forecast results or window average value of several future periods for joint judgment. As long as we can distinguish between periods of surplus green electricity and periods of insufficient green electricity, they are all equivalent alternatives of the present invention.

[0064] S3. Migrate the flexible task load of the remaining time periods to the surplus green electricity time periods. After migration, determine whether there is still surplus green electricity in the current time period. If so, execute step S31.

[0065] In an embodiment of the present invention, it is set that For time period Flexible tasks in time period The actual amount of tasks performed, of which, Must belong to a time period Allow migration window For any given time period The conservation relationship of flexible task quantity is expressed as:

[0066] (4)

[0067] in, For time period Flexible workload. Therefore, time period Flexible task migration volume Represented as:

[0068] (5)

[0069] Time period Flexible workload The tasks are migrated to the surplus green electricity period, thus concentrating flexible tasks on the surplus green electricity period.

[0070] Furthermore, to measure how much flexible task migration has occurred, after migrating the amount of flexible tasks to the surplus green electricity periods, the following steps are also included:

[0071] Determine the net increase in flexible task volume during the surplus green electricity period. Net increase in flexible task volume. As shown in formula (6), this is used to measure the degree to which flexible tasks migrate to periods of surplus green electricity:

[0072] (6)

[0073] In specific migration operations, a migration matrix approach can be used. For example, if the original task queue contains tasks that can be executed early, those tasks are prioritized for transfer to surplus green electricity periods. If the original task queue allows for delayed execution, the flexible task migration volume is postponed to subsequent surplus green electricity periods, provided that the task completion constraints are guaranteed. In practical applications, in addition to using the migration matrix approach described above to describe flexible task migration, task priority queues, task cluster migration, batch processing migration, or rolling scheduling methods can also be used. As long as the early or delayed execution of flexible tasks can be achieved within the allowed window, they all fall within the scope of protection of this invention.

[0074] Step S31: Convert the remaining surplus green electricity into cooling capacity and / or cold storage capacity. This can be achieved by increasing the cooling power of the refrigeration equipment and / or the cold storage power of the cold storage system, thereby realizing the advance production and cross-time period transfer of cooling capacity.

[0075] Converting the surplus green electricity into cooling capacity and / or cold storage capacity includes:

[0076] S311. Determine the total load power for the current time period. ;

[0077] (7)

[0078] In practical applications, total load power Data center equipment processing capacity constraints must be met:

[0079] (8)

[0080] in, For time period of Maximum power capacity that the equipment can handle.

[0081] S312. Determine the cooling load power for the current period based on the basic cooling load item, the load ratio factor, and the total load power. ;

[0082] (9)

[0083] in, For time period cooling load power, This is the load ratio coefficient between total load power and cooling load power. This refers to the basic cooling load item; when the execution timing of flexible tasks changes. As it changes, the corresponding cooling load power It is also updated simultaneously.

[0084] In practical applications, in addition to using the linear mapping relationship shown in formula (9) to calculate the cooling load power based on the total load power, piecewise linear relationship, empirical fitting relationship, lookup table relationship or data-driven prediction model can also be used. As long as the corresponding cooling load power can be calculated based on the total load power, it falls within the protection scope of this invention.

[0085] S313. Determine the cooling power and cold storage power for the current period based on the cooling balance relationship and the cooling load power.

[0086] In an embodiment of the present invention, it is set that For time period The cooling capacity provided by the refrigeration equipment For time period The cold storage capacity of the cold storage system For time period The cold release power of the cold storage system, during the time period The cooling balance relationship is as follows:

[0087] (10)

[0088] The above relationship indicates that the immediate cooling power of the refrigeration equipment and the cold release power of the cold storage system together meet the current cooling load power and the additional cooling power absorbed by the cold storage system.

[0089] It can also calculate time periods Power consumption of refrigeration equipment Calculate using the following formula:

[0090] (11)

[0091] in, For time period Coefficient of performance of refrigeration equipment.

[0092] It can also calculate the state variables of the cold storage system for the current period. The following formula can be used for recursion:

[0093] (12)

[0094] in, The self-loss coefficient of the cold storage system. For time period The state parameters of the cold storage system in the previous period. To improve cold storage efficiency, For cooling efficiency. Time period. The cold storage capacity satisfies the following capacity constraints:

[0095] (13)

[0096] in, The maximum cold storage capacity of a cold storage system is determined by the system's volume, medium, and other physical parameters.

[0097] In practical applications, the cold storage system can be a cold water cold storage system, an ice cold storage system, or other cold storage devices capable of storing and releasing cold energy across time periods; the cold storage system can be a single cold storage unit, or multiple cold storage units connected in parallel or partitioned. This invention does not impose specific limitations in this regard.

[0098] In this embodiment of the invention, when the time period During periods of surplus green electricity, and when surplus green electricity still exists after the flexible task relocation, increase the cooling capacity of the refrigeration equipment. and the cold storage capacity of the cold storage system This is to enable the advance production and cross-period transfer of cooling capacity.

[0099] After converting the surplus green electricity into cooling capacity and / or cold storage capacity in step S31, the method further includes:

[0100] The remaining surplus green electricity is used to charge the battery.

[0101] In an embodiment of the present invention, it is set that For time period Battery charging power, For time period Battery discharge power, For time period If the battery state of charge is determined, then the battery state of charge assessment model is as shown in formula (14):

[0102] (14)

[0103] in, For time period The state of charge of the battery. Improve battery charging efficiency. For the battery discharge efficiency, and the battery state of charge satisfying the following:

[0104] (15)

[0105] in, This is the minimum permissible state of charge for the battery, preventing over-discharge; This is the maximum state of charge allowed for the battery, preventing overcharging.

[0106] S4. Reduce the amount of flexible tasks in the current period and release the cold storage capacity to reduce the power consumption of instantaneous cooling.

[0107] Furthermore, by reducing the amount of flexible tasks during the current period and releasing the stored cooling capacity, if the period... When a power shortage still exists, the battery discharges to compensate for the power shortage.

[0108] In practical applications, to avoid simultaneous cold storage and cold release within the same time period, this embodiment of the invention also introduces a binary variable representing the cold storage and cold release status of the cold storage system during the current time period. This is used to apply a mutual exclusion constraint between energy storage and energy release. That is, converting the surplus green electricity into energy storage includes:

[0109] Determine the binary variable of the cold storage and release status of the cold storage system in the current period, and convert the remaining surplus green electricity into cold storage capacity based on the binary variable of the cold storage and release status, that is, increase the cold storage power of the cold storage system.

[0110] Alternatively, release the stored cold energy, including:

[0111] Determine the binary variable of the cold storage and release status of the cold storage system for the current period, and release the stored cold amount according to the binary variable of the cold storage and release status.

[0112] The binary variable representing the cold storage / release state is used to characterize the operating state of the cold storage system in the current time period. When the value is 1, the cold storage system is allowed to store cold and its release is limited; when... When the value is 0, the cold storage system is allowed to release cold and cold storage is limited. As shown in formulas (16) and (17):

[0113] (16)

[0114] (17)

[0115] in, This represents the upper limit of the maximum cold storage capacity of the cold storage system. The maximum cooling power of the cold storage system is determined by factors such as the heat exchange capacity of the cold storage system and the flow rate of the pipeline.

[0116] Similarly, to avoid simultaneous charging and discharging of the battery energy storage system within the same time period, this embodiment of the invention introduces a binary variable representing the battery charging and discharging state during the current time period. And apply charging and discharging mutual exclusion constraints. That is, before providing excess green electricity to the battery for charging, it also includes:

[0117] Determine the binary variable of the battery charge / discharge state, and control the battery charging based on the binary variable of the battery charge / discharge state.

[0118] Alternatively, before the battery discharges, it also includes:

[0119] Determine the binary variable of the battery's charge / discharge state, and control the battery discharge based on the binary variable of the battery's charge / discharge state.

[0120] The binary variable representing the battery charge / discharge state is used to characterize the operating state of the battery energy storage system in the current time period. When the value is 1, battery charging is allowed and battery discharge is limited; when... When the value is 0, battery discharge is allowed and battery charging is limited.

[0121] As shown in formulas (18) and (19):

[0122] (18)

[0123] (19)

[0124] in, Maximum charging power for the battery. This represents the battery's maximum discharge power.

[0125] Therefore, the battery energy storage system undertakes the task of balancing the residuals on the electric side after the load side and cold side adjustments.

[0126] It should be noted that the battery energy storage system can be a single energy storage unit or an energy storage cluster composed of multiple energy storage units; the energy storage technology can be lithium-ion batteries, sodium-ion batteries, or other energy storage devices with fast charging and discharging capabilities. This invention does not impose specific limitations in this regard.

[0127] Furthermore, after S3 or S4, it also includes:

[0128] Establish the power balance relationship for the current time period;

[0129] In an embodiment of the present invention, it is set that For time period Green electricity utilization power, For time period The power purchased by the power grid during the time period The power balance equation is:

[0130] (20)

[0131] Determine the power allocation relationship between the current green energy utilization rate and the green energy curtailment power;

[0132] In this embodiment of the invention, the green electricity utilization power and the green electricity abandonment power satisfy the following:

[0133] (twenty one)

[0134] (twenty two)

[0135] in, For time period The power of green abandonment.

[0136] Based on the above power balance and power allocation relationships, the current period's green electricity utilization, green electricity curtailment, and grid-purchased electricity can be obtained. When an unbalanced load gap still exists after adjustments by flexible tasks, cold storage systems, and battery energy storage, electricity will be purchased from the grid. Compensation will be provided.

[0137] In this embodiment of the invention, after determining the above-mentioned multi-resource hierarchical adjustment method, different priorities can be used for calculation when adjusting based on this method.

[0138] First, the optimization objective is to minimize the amount of green space abandoned throughout the entire time period.

[0139] (twenty three)

[0140] After obtaining the minimum green energy abandonment, the adjustment method corresponding to the minimum green energy abandonment for the entire time period is set as the first candidate method set; then, the second-level optimization objective is to maximize the net increase of flexible tasks during the surplus green energy period within the first candidate method set.

[0141] (twenty four)

[0142] The adjustment method corresponding to the maximum net increase of the flexible task during the surplus green electricity period is set as the second candidate method set; then, in the second candidate method set, the maximum cold storage power of the cold storage system during the surplus green electricity period is taken as the third-level optimization objective:

[0143] (25)

[0144] in, The duration of the surplus green electricity period.

[0145] Setting a third layer of optimization objectives is to further enhance the ability of the cold storage system to absorb surplus green electricity, while ensuring that green electricity is minimized and flexible tasks are prioritized.

[0146] The adjustment method corresponding to the maximum cold storage power of the cold storage system during surplus green electricity periods is set as the third candidate method set.

[0147] Furthermore, in the third candidate method set, one or more combinations of the following are used as the fourth-level optimization objectives: electricity purchase cost, battery throughput, flexible task delay penalty, peak electricity purchase, cold storage operation cost, and smoothness index. Among them, peak electricity purchase refers to the maximum power value purchased from the grid in a single scheduling cycle; cold storage operation cost refers to the comprehensive cost of energy consumption and losses incurred during the cold storage system's cold release and storage process; and the smoothness index refers to the quantitative value that measures the degree of fluctuation of battery power or purchased power between time periods.

[0148] The cost of purchasing electricity from the power grid can be expressed as:

[0149] (26)

[0150] Battery throughput can be expressed as:

[0151] (27)

[0152] Flexible task delay penalties can be constructed based on the deviation between the original execution time and the actual execution time of the flexible task, for example:

[0153] (28)

[0154] in, For time period The task during the time period The corresponding delay penalty coefficient during execution.

[0155] Through the above-mentioned hierarchical solution process, an adjustment result is formed that prioritizes "minimizing green abandonment, then accommodating flexible tasks, then prioritizing cold storage, and finally adjusting in conjunction with battery and economic indicators".

[0156] In practical applications, in addition to the lexicographical optimization method, which solves multiple optimization objectives sequentially according to their priority from high to low, phased optimization, master-slave optimization, rule-driven priority adjustment, or rolling optimization can also be used. As long as the adjustment logic is the same as that of this invention, they are all equivalent alternatives to this invention.

[0157] In the simulation or optimization of scheduling cycles, it is usually necessary to set the battery state of charge or cold storage state to return to a preset state at the end of the cycle. However, this setting may produce non-engineering abnormal adjustment behaviors at the end of the scheduling cycle. For example, the battery may suddenly charge at high power in the last period just to pull the battery state of charge back to a specified value; or the cold storage system may force high-power cold storage or cold release at the end of the period just to return the cold storage state to the initial value. These charging, cold storage, or cold release actions are not actually required by the current load, green energy consumption, or system operation, but are "sudden adjustments" generated by the model to meet mathematical constraints.

[0158] Therefore, in order to avoid abnormal adjustment behaviors that do not meet actual operating requirements, such as sudden charging, sudden discharging, forced cold storage, or forced cold release, due to the forced fulfillment of the battery state of charge or cold storage state return requirements at the end of the scheduling cycle, soft constraints on the terminal state of the battery energy storage and cold storage system can be set.

[0159] For battery energy storage, the battery end-state deviation is expressed as:

[0160] (29)

[0161] The terminal state deviation of the cold storage system is expressed as:

[0162] (30)

[0163] in, Decomposition variables for positive deviation of battery terminal state Decompose the negative deviation variables of the battery terminal state. For the positive deviation decomposition variables of the terminal state of the cold storage system, The negative deviation decomposition variable represents the terminal state of the cold storage system. This deviation decomposition variable is introduced into the final objective function as a penalty term, and a maximum allowable deviation range is limited to avoid unnecessary high-power charging and discharging or forced cold storage / release behaviors during the end-of-period state-forced regression.

[0164] After completing the above multi-layer solution, the hourly collaborative adjustment results are output. The output results include at least: the flexible task load for each time period. Time period Flexible tasks in time period Actual workload Total load power Cooling load power Cooling power of refrigeration equipment The cold storage capacity of the cold storage system The cooling power of the cold storage system Cold storage state Battery charging power Battery discharge power Battery state of charge Green electricity utilization power Green abandonment power and the power purchased by the power grid .

[0165] In practical applications, this invention preferably operates using a rolling adjustment method. Specifically, after each control cycle, the photovoltaic power output prediction, wind power output prediction, flexible task queue, load data, and equipment status data for the next rolling window are reacquired, along with the battery state of charge at the end of the previous cycle. and cold storage state As the initial state for the next cycle, the above steps are repeated to achieve continuous implementation of green electricity direct connection to data center multi-resource coordinated adjustment.

[0166] This invention first determines the time period type based on predicted power and baseline load power. Then, during periods of surplus green electricity, it migrates flexible tasks first, and then determines the cooling capacity and cold storage capacity. During periods of insufficient green electricity, it first reduces the amount of flexible tasks and then releases the cold storage capacity, clarifying the priority of flexible tasks and the cold storage system and avoiding disorderly competition among multiple resources. After the adjustment of flexible tasks, the decision on cooling, cold storage, and cold release is made, ensuring that the cold-side adjustment is always based on the change in total load, truly reflecting the coupled operation characteristics of the data center's "computing-cooling-power". At the same time, it also constructs an adjustment sequence oriented towards the consumption of surplus green electricity. When green electricity is surplus, it prioritizes the migration and consumption of green electricity through flexible tasks, and then realizes the cross-time period transfer of cooling capacity through cold storage. When green electricity is insufficient, it prioritizes reducing the amount of flexible tasks executed, and then releases the cold storage capacity to reduce cooling power consumption. This sequence makes the priority hierarchy of the adjustment strategy clear and has good engineering feasibility. Finally, battery energy storage is used as residual compensation, effectively reducing unnecessary battery activation and wear.

Claims

1. A multi-resource hierarchical adjustment method for direct green electricity connection to a data center, characterized in that, include: S1. Discretize the scheduling cycle into multiple consecutive time periods and determine the predicted power and baseline load power for the current time period; S2. Determine whether the current period belongs to the surplus green electricity period based on the predicted power and the baseline load power of the current period. If so, proceed to step S3. Otherwise, proceed to step S4; S3. Migrate the flexible task load of the remaining time periods to the surplus green electricity time periods, and determine whether there is still surplus green electricity in the current time period after migration. If there is still surplus green electricity, the surplus green electricity will be converted into cooling capacity and / or cooling storage capacity; the surplus green electricity will be supplied to the battery for charging. Converting the surplus green electricity into cooling capacity and / or cold storage capacity specifically includes: Determine the total load power for the current time period; The cooling load power for the current period is determined based on the basic cooling load item, the load ratio factor, and the total load power. ， in, For the current time period The cooling load power; This is the load ratio coefficient between total load power and cooling load power; For the current time period Total load power; Basic cooling load item; The cooling power and / or cold storage power for the current period are determined based on the cooling balance relationship and the cooling load power. S4. Reduce the amount of flexible tasks in the current period and release the stored cold energy; discharge the battery.

2. The method according to claim 1, characterized in that, Before supplying surplus green electricity to the battery for charging, the following steps are also included: Determine the binary variable of the battery charge / discharge state, and control the battery charging based on the binary variable of the battery charge / discharge state. Alternatively, before the battery discharges, it also includes: Determine the binary variable of the battery's charge / discharge state, and control the battery discharge based on the binary variable of the battery's charge / discharge state.

3. The method according to claim 1, characterized in that, Determining whether the current period belongs to a surplus green electricity period based on the predicted power and baseline load power of the current period includes: Determine the photovoltaic power forecast and wind power forecast for the current time period, and determine the available green electricity power for the current time period based on the photovoltaic power forecast and wind power forecast. Determine the baseline load power for the current time period; When the available green electricity is greater than or equal to the baseline load power, the current time period is determined to be a period of surplus green electricity.

4. The method according to claim 3, characterized in that, The baseline load power for the current time period is determined as follows: Determine the base cooling load power for the current period based on the base cooling load and the performance coefficient of the refrigeration equipment for the current period. Determine the rigid load power for the current time period; The baseline load power for the current period is determined based on the baseline cooling load power and the rigid load power for the current period.

5. The method according to claim 1, characterized in that, After migrating the flexible workload from other time periods to the surplus green electricity periods, the following is also included: Determine the net increase in the amount of flexible tasks during the period of surplus green electricity.

6. The method according to claim 1, characterized in that, S3 converts the surplus green electricity into cold storage capacity, including: Determine the binary variable of the cold storage and release status of the cold storage system for the current period, and convert the remaining surplus green electricity into cold storage capacity based on the binary variable of the cold storage and release status; Alternatively, the stored cold energy can be released in S4, including: Determine the binary variable of the cold storage and release status of the cold storage system for the current period, and release the stored cold amount according to the binary variable of the cold storage and release status.

7. The method according to claim 1, characterized in that, Following S3 or S4, the following is also included: Establish the power balance relationship for the current time period; Determine the power allocation relationship between the current green energy utilization rate and the green energy curtailment power; The amount of green electricity used, the amount of green electricity abandoned, and the amount of electricity purchased by the grid during the current period are calculated based on the power balance relationship and the power distribution relationship.

Images