Multi-data center multi-computing power collaborative optimization method and system based on computing power and refrigeration system comprehensive energy consumption cost, and storage medium

Abstract

The invention discloses a multi-data center multi-computing power collaborative optimization method and system based on computing power and refrigeration system comprehensive energy consumption cost, and a storage medium. The method comprises the following steps: S1, uniformly dividing a total conventional computing power resource and a total intelligent computing power resource which need to be scheduled into a plurality of sub-computing power resources; s2, sequentially allocating and starting a data center for each conventional sub-computing power resource; S2.1, calculating the cost of each data center after the conventional sub-computing power resource is started, and selecting to start the data center with the minimum cost; s2.2, repeating the step S2.1 until the total conventional computing power resource needing to be scheduled is reached; s3, sequentially allocating and starting a data center for each intelligent sub-computing power resource: S3.1, solving an optimal operation period deployment scheme of the intelligent sub-computing power resource in each data center through a genetic algorithm, calculating the cost after starting the intelligent sub-computing power resource according to the optimal operation period deployment scheme, and selecting to start the intelligent sub-computing power resource with the minimum cost; s3.2, repeating the step S3.1 until the total intelligent computing power resource needing to be scheduled is reached; according to the method, the computing power resource operation cost can be minimized.

Description

TECHNICAL FIELD

[0001] The application relates to data center computing power scheduling, in particular to a multi-data center multi-computing power collaborative optimization method, system and storage medium based on comprehensive energy consumption cost of computing power and refrigeration systems. BACKGROUND

[0002] The computing power services of the existing data center can be divided into two categories according to the scheduling type: uninterruptible computing power and interruptible computing power. The uninterruptible computing power is the conventional computing power, which is characterized by running continuously after being called once and cannot be interrupted. The interruptible computing power is the intelligent computing power, which is characterized by large computing capacity but controllable computing period, and a complete computing power service can be divided into computing power services in different time periods and deployed and run in different data centers.

[0003] The main energy consumption of the data center includes computing power energy consumption and refrigeration system energy consumption, which is closely related to the total computing power size called by the server. However, the energy consumption of the same computing power called by the same type of server in different data centers is roughly the same, but the refrigeration system energy consumption corresponding to the difference in outdoor temperature and humidity environment is not the same. Secondly, the electricity price of different data centers at different times is not the same. The existing multi-data center scheduling different computing power resources cannot reasonably plan according to the different time-of-use electricity prices corresponding to different regions, and the scheduled computing power resources can meet the requirements of the computing power service, but cannot meet the requirement of the lowest economic cost at the same time. SUMMARY

[0004] The purpose of the application is to provide a multi-data center multi-computing power collaborative optimization method, system and storage medium based on comprehensive energy consumption cost of computing power and refrigeration systems, which can minimize the running cost of computing power resources.

[0005] Technical scheme: The multi-data center multi-computing power collaborative optimization method based on comprehensive energy consumption cost of computing power and refrigeration systems comprises the following steps:

[0006] S1, divide the total conventional computing power resources and the total intelligent computing power resources to be scheduled into multiple conventional sub-computing power resources and intelligent sub-computing power resources to be scheduled respectively;

[0007] S2, allocate a data center to each conventional sub-computing power resource in turn:

[0008] S2.1, calculate the total power consumption cost of each data center after starting the conventional sub-computing power resource, and select the data center with the minimum total power consumption cost;

[0009] S2.2, repeat S2.1 until the sum of the started conventional sub-computing power resources reaches the total conventional computing power resources to be scheduled;

[0010] S3, allocate a data center to each intelligent sub-computing power resource in turn:

[0011] S3.1, solving the optimal runtime period deployment scheme of the intelligent sub-computing power resource in each data center by a genetic algorithm, according to which the total power consumption cost after enabling the intelligent sub-computing power resource is calculated according to the time-of-use electricity price of each data center, and the data center with the minimum total power consumption cost is selected to be enabled;

[0012] S3.2, repeating S3.1 until the sum of the enabled intelligent sub-computing power resources reaches the total intelligent computing power resource to be scheduled.

[0013] Based on the above technical solutions, step S1 realizes fine-grained scheduling by dividing the total computing power to be scheduled into multiple sub-computing power resources, avoids single data center overload, and improves resource utilization; step S2 re-calculates the total data center cost after each sub-unit is scheduled for the calling of the conventional sub-computing power resource, dynamically reflects the superimposed effect of the deployed computing power on energy consumption, and compared with the traditional one-time allocation, the calculated cost is more accurate, which can ensure the minimization of the calling cost of the conventional computing power resource; step S3 considers the feature that the runtime period of the intelligent sub-computing power resource can be split, considers that the original electricity price is different due to the difference in regions of different data centers, and the electricity price is also different at different time periods, and on this basis, the genetic algorithm is used to solve the optimal runtime period deployment scheme of the intelligent sub-computing power resource in each data center, which can realize the lowest running cost of the intelligent sub-computing power resource in each data center, and then the data center with the lowest cost is selected to enable the intelligent sub-computing power resource, which can ensure the lowest calling cost of the intelligent sub-computing power resource, and the calling of the intelligent sub-computing power resource also adopts the sequential calling mode, which can dynamically reflect the superimposed effect of the deployed computing power on energy consumption, improve the accuracy of cost calculation, and further reduce the calling cost of the final computing power resource. In summary, the method can minimize the running cost of multiple computing power resources in multiple data centers.

[0014] As a preferred, the calculation formula of the fitness in step (2) is

[0015]

[0016] Wherein, F l is the fitness of the lth individual, is the total power consumption cost of the ith data center after enabling the mth intelligent sub-computing power resource according to the lth individual scheme;

[0017] The calculation formula of the fitness is

[0018]

[0019] Wherein, a total power consumption cost of all the conventional sub-computing resources enabled in the i th data center at the time when the m th intelligent sub-computing resource is enabled in the i th data center, a total power consumption cost of all the intelligent sub-computing resources enabled in the i th data center at the t th time period after the m th intelligent sub-computing resource is enabled in the i th data center according to the l th individual scheme;

[0020] a total power consumption cost of all the conventional sub-computing resources enabled in the i th data center at the time when the m th intelligent sub-computing resource is enabled in the i th data center,

[0021]

[0022] wherein, TOU(i, t) is a time-of-use electricity price of the i th data center at the t th time period, a total power consumption cost of all the intelligent sub-computing resources enabled in the i th data center at the t th time period after the m th intelligent sub-computing resource is enabled in the i th data center according to the l th individual scheme;

[0023] a total power consumption cost of all the conventional sub-computing resources enabled in the i th data center at the time when the m th intelligent sub-computing resource is enabled in the i th data center,

[0024]

[0025] wherein, a total power consumption cost of all the intelligent sub-computing resources enabled in the i th data center at the t th time period after the m th intelligent sub-computing resource is enabled in the i th data center according to the l th individual scheme;

[0026] The above method, when calculating the total power consumption cost of the intelligent sub-computing resource, not only considers the time-of-use electricity price of different data centers, but also considers the total power consumption cost of the conventional sub-computing resource enabled in the data center, dynamically reflects the superimposed influence of the deployed resource on energy consumption, and improves the accuracy of cost calculation.

[0027] As preferred, when any individual corresponding operation period deployment scheme causes the intelligent sub-computing resource to consume intelligent computing power exceeding the maximum intelligent computing power that the data center can provide at any time period, the following penalty term is added to the fitness, and the fitness after adding the penalty term is

[0028]

[0029] wherein, F l is the fitness after adding the penalty term.

[0030] The addition of the penalty term can greatly reduce the fitness of the individual whose operation deployment scheme causes the intelligent sub-computing resource to exceed the corresponding computing power upper limit that the data center can provide, and further reduce the probability of being selected, can greatly reduce the probability of being selected as the optimal scheme, and ensure that the scheme finally solved by the genetic algorithm is feasible.

[0031] As preferred, each data center enabled conventional sub-computing resource and intelligent sub-computing resource cannot exceed the maximum value of the corresponding computing power that the data center can provide when the computing resource is enabled.

[0032] Through the above-mentioned limitation, it can be avoided that the upper limit of the computing power provided by the data center is ignored due to the consideration of the running cost alone, resulting in the situation that the final cost is the lowest but the actual calling is impossible, because in practice, there may be some data centers in the region with lower electricity price, leading to the situation that the computing power resource is concentrated in these data centers only considering the total power consumption cost.

[0033] The multi-data center multi-computing power collaborative optimization system based on the computing power and refrigeration system comprehensive energy consumption cost of the application comprises:

[0034] The computing power resource division module is used to divide the total conventional computing power resource and the total intelligent computing power resource into multiple conventional sub-computing power resources and intelligent sub-computing power resources to be dispatched respectively.

[0035] The conventional computing power scheduling module is used to confirm the data center enabled for each conventional sub-computing power resource in turn, comprising the following sub-modules:

[0036] The cost calculation sub-module is used to calculate the total power consumption cost after the data center enables the conventional sub-computing power resource, and select the data center with the minimum total power consumption cost.

[0037] The conventional computing power iteration sub-module is used to repeat the operation of the cost calculation sub-module to confirm the data center enabled for the next conventional sub-computing power resource, until the sum of all enabled conventional sub-computing power resources reaches the total conventional computing power resource to be dispatched.

[0038] The intelligent computing power scheduling module is used to confirm the data center enabled for each intelligent sub-computing power resource in turn, comprising the following sub-modules:

[0039] The optimization deployment sub-module is used to solve the optimal running time period deployment scheme of the intelligent sub-computing power resource in each data center by genetic algorithm, and calculate the total power consumption cost after the data center enables the intelligent sub-computing power resource according to the time-of-use electricity price of each data center, and select the data center with the minimum total power consumption cost.

[0040] The intelligent computing power iteration sub-module is used to repeat the operation of the optimization deployment sub-module to confirm the data center enabled for the next intelligent sub-computing power resource, until the sum of all enabled intelligent sub-computing power resources reaches the total intelligent computing power resource to be dispatched.

[0041] The computer-readable storage medium storing one or more programs of the present application includes one or more programs including instructions that, when executed by a computing device, cause the computing device to perform any of the above methods.

[0042] Advantages: BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 Flowchart of the genetic algorithm of the present application. DETAILED DESCRIPTION

[0044] As shown in the figure, the multi-data center multi-computing power collaborative optimization method based on computing power and refrigeration system comprehensive energy consumption cost of the present application includes the following steps:

[0045] S1, divide the total conventional computing power resources and the total intelligent computing power resources to be dispatched into multiple portions of conventional sub-computing power resources and intelligent sub-computing power resources to be dispatched, respectively.

[0046] S2, allocate each conventional sub-computing power resource to a data center in turn:

[0047] S2.1, calculate the total power consumption cost of each data center after enabling the conventional sub-computing power resource, and select the data center with the minimum total power consumption cost. Because the main energy consumption of the data center comes from the power consumption of calling computing power resources and the power consumption of the refrigeration system, the power consumption cost mentioned in the present application refers to the comprehensive power consumption cost (i.e. comprehensive energy consumption cost) of computing power and refrigeration system.

[0048] The total power consumption cost calculation formula of the conventional sub-computing power resource in the data center is

[0049]

[0050] Among them, is the total power consumption cost of the i th data center after enabling the n th conventional sub-computing power resource, and COST U,n (i, t) is the power consumption cost of the i th data center after enabling the n th conventional sub-computing power resource at time period t.

[0051] COST U,n (i, t) is calculated by the formula

[0052] COST U,n (i, t)TOU(i, t)EU n (i, t)

[0053] Among them, TOU(i, t) is the time-of-use electricity price of the i th data center at time period t, and EU n(i, t) is the power consumption of all enabled regular sub-computing power resources of the ith data center at time t after the nth regular sub-computing power resource is enabled.

[0054] EU n The calculation formula of (i, t) is

[0055]

[0056] wherein, VACU n (i, t) is the regular computing power value consumed by all enabled regular sub-computing power resources of the ith data center at time t after the nth regular sub-computing power resource is enabled. With the increase of the enabled regular sub-computing power resources, the consumed regular computing power value will be nonlinearly superimposed. Because the more enabled, the greater the increase of the consumed computing power value, the more accurate the cost calculated considering these, and these specific values can be obtained according to the past running data of each data center or estimated according to the hardware configuration of the data center (the hardware configuration includes the refrigeration system in addition to the computing power hardware); CCPRU(i) is the regular computing power value provided by the unit power consumption of the ith data center.

[0057] S2.2, repeat S2.1 to allocate the next regular sub-computing power resource to the data center until the sum of the enabled regular sub-computing power resources reaches the total regular computing power resource to be scheduled;

[0058] S3, in turn, allocate each intelligent sub-computing power resource to the enabled data center:

[0059] S3.1, solve the optimal running time period deployment scheme of the intelligent sub-computing power resource in each data center by genetic algorithm, and calculate the total power consumption cost after the intelligent sub-computing power resource is enabled according to the time-of-use electricity price of each data center, and select the data center with the minimum total power consumption cost to enable;

[0060] The genetic algorithm to solve the optimal running time period deployment of the intelligent sub-computing power resource in any data center includes the following steps:

[0061] (1) initialize the population, randomly generate N intelligent sub-computing power resource running time period deployment schemes in the data center as individuals, and the value of N can be set according to the actual situation.

[0062] The intelligent sub-computing power resource running time period deployment scheme in the data center is

[0063]

[0064] wherein, is the running time period deployment scheme of the mth intelligent sub-computing power resource in the ith data center, is the running instruction value of the mth smart sub-computing resource in the DCI period of the ith data center, and DCI is the running time set for the smart sub-computing resource every day, 0 indicates not running, and 1 indicates running;

[0065] The lth individual in the population The expression of the lth individual in the population is

[0066]

[0067] wherein the value of each element is randomly set and satisfies that the number of elements with a value of 1 is not greater than the actual required running time TPI of the smart sub-computing resource, TPI < DCI, and DCI is greater than TPI, that is, the daily set running time of the smart sub-computing resource is greater than its actual required running time, so as to ensure that it can complete the operation task, but when solving the optimal running deployment scheme, as long as the running period deployment scheme of the smart sub-computing resource is set according to TPI, the running cost can be minimized.

[0068] (2) Calculate the fitness of each individual, and the fitness is the reciprocal of the total power consumption cost of enabling the smart sub-computing resource in the data center according to the deployment scheme of the individual;

[0069] The calculation formula of the fitness is

[0070]

[0071] wherein F l is the fitness of the lth individual, is the total power consumption cost of the ith data center after enabling the mth smart sub-computing resource according to the lth individual scheme;

[0072] The calculation formula of the total power consumption cost of the ith data center after enabling the mth smart sub-computing resource according to the lth individual scheme is

[0073]

[0074] wherein, is the total power consumption cost of all the conventional sub-computing resources already enabled in the data center when the mth smart sub-computing resource is enabled in the ith data center, is the total power consumption cost of all the smart sub-computing resources already enabled in the data center at time t after the mth smart sub-computing resource is enabled in the ith data center according to the lth individual scheme;

[0075] The calculation formula of the total power consumption cost of all the smart sub-computing resources already enabled in the data center at time t after the mth smart sub-computing resource is enabled in the ith data center according to the lth individual scheme is

[0076]

[0077] wherein TOU(i, t) is the time-of-use electricity price of the ith data center at time t, the power consumption of all enabled intelligent sub-computing power resources of the i th data center in the t th time period after the m th intelligent sub-computing power resource is enabled in the i th data center according to the l th individual scheme;

[0078] The calculation formula is

[0079]

[0080] wherein, the intelligent computing power value consumed by all enabled intelligent sub-computing power resources of the i th data center in the t th time period after the m th intelligent sub-computing power resource is enabled in the i th data center according to the l th individual scheme, and the intelligent computing power value consumed by the intelligent sub-computing power resources enabled in the same data center in the same time period also increases nonlinearly with the increase of the intelligent sub-computing power resources, and the specific value can be obtained according to the past operation data of the data center or estimated through the hardware parameters of the data center (the hardware includes not only the computing power hardware but also the refrigeration system) ; C C P R I ( i ) is the intelligent computing power value provided by the unit power consumption of the i th data center.

[0081] When the running time period deployment scheme corresponding to any individual causes the intelligent computing power consumed by all enabled intelligent sub-computing power resources to exceed the maximum intelligent computing power that can be provided by the data center in any time period, the following penalty term is added to the fitness, and the fitness after adding the penalty term is

[0082]

[0083] wherein, F l is the fitness after adding the penalty term.

[0084] (3) The probability of being selected is calculated according to the fitness of all individuals, and N individuals are repeatedly selected from the population as a preliminary new population according to the probability;

[0085] The probability P l of being selected of the l th individual is calculated according to the following formula:

[0086]

[0087] The original population is the parent population, and the individuals of the preliminary new population are randomly and repeatedly selected from the parent population according to the size of the probability of being selected of each individual as the individuals of the preliminary new population. The individual with the greater probability of being selected is more likely to be repeatedly selected.

[0088] (4) The individuals of the preliminary new population are subjected to crossover or mutation operation, and then the individuals are repaired to obtain a final new population;

[0089] The crossover refers to randomly selecting two individuals, and exchanging random elements in the two individuals (vectors) in any pair; the mutation is to randomly extract an individual, and transform the random element value in the individual, that is, transform the element value between 0 and 1, and the original 0 becomes 1, and the original 1 becomes 0.

[0090] After crossover and mutation, a new individual can be generated, but the number of elements with a value of 1 in the new individual can be greater than TPI, which can cause additional running time and increase the cost, at this time, the individual needs to be repaired, that is, the number of elements to be repaired is obtained according to the actual number of elements with a value of 1-TPI, and the value of the number of elements with a value of 1 in the individual is randomly selected and replaced by 0; on the contrary, that is, the number of elements with a value of 1 in the new individual is less than TPI, and the total power consumption cost calculated according to the individual is smaller, but the actual calculation task cannot be completed, if it is not repaired, the optimal deployment scheme finally solved is also not practical, at this time, the number of elements to be repaired = TPI-number of elements with a value of 1, and the value of the number of elements with a value of 0 in the individual is randomly selected and replaced by 1, so the individual repair is completed.

[0091] (5) Calculate the fitness of each individual in the final new population, and take the individual with the maximum fitness as the output value of this iteration;

[0092] (6) Return to step (2) until the number of iterations reaches the threshold value or the fitness change corresponding to the iteration output value is less than the threshold value for M consecutive times, and output the optimal deployment scheme, that is, the output value (individual corresponding running deployment scheme) of the last iteration is taken as the optimal running period deployment scheme of the intelligent sub-computing power resource in the data center.

[0093] The threshold value of the number of iterations, the threshold value to be less than the fitness change corresponding to the iteration output value for M consecutive times, and M can be set according to the actual situation.

[0094] S3.2, repeat S3.1 until the sum of the enabled intelligent sub-computing power resources reaches the total intelligent computing power resource to be scheduled.

[0095] The scheduling principle that steps S2 and S3 also need to comply with is that the enabled conventional sub-computing power resource and intelligent sub-computing power resource of each data center cannot exceed the maximum value of the corresponding computing power that can be provided when the computing power resource is enabled.

[0096] The multi-data center multi-computing power collaborative optimization system based on the comprehensive energy consumption cost of computing power and refrigeration system according to the application comprises:

[0097] The computing power resource division module is used for dividing the total conventional computing power resource and the total intelligent computing power resource to be scheduled into multiple portions of conventional sub-computing power resources and intelligent sub-computing power resources to be scheduled, respectively.

[0098] Conventional computing power scheduling module: for confirming each data center enabled by each conventional sub-computing power resource in turn, including the following sub-modules:

[0099] Cost calculation sub-module: for calculating the total power consumption cost of each data center after enabling the conventional sub-computing power resource, and selecting the data center with the minimum total power consumption cost for enabling;

[0100] Conventional computing power iteration sub-module: for repeating the operation of the cost calculation sub-module to confirm the data center enabling the next conventional sub-computing power resource until the sum of all enabled conventional sub-computing power resources reaches the total conventional computing power resource to be scheduled;

[0101] Intelligent computing power scheduling module: for confirming each data center enabled by each intelligent sub-computing power resource in turn, including the following sub-modules:

[0102] Optimized deployment sub-module: for solving the optimal runtime period deployment scheme of the intelligent sub-computing power resource in each data center by a genetic algorithm, calculating the total power consumption cost of each data center after enabling the intelligent sub-computing power resource according to the time-of-use electricity price of each data center, and selecting the data center with the minimum total power consumption cost for enabling;

[0103] Intelligent computing power iteration sub-module: for repeating the operation of the optimized deployment sub-module to confirm the data center enabling the next intelligent sub-computing power resource until the sum of all enabled intelligent sub-computing power resources reaches the total intelligent computing power resource to be scheduled.

[0104] The computer-readable storage medium storing one or more programs includes one or more programs including instructions, which when executed by a computing device, cause the computing device to perform any of the above methods.

Claims

1. A multi-data center multi-computing power collaborative optimization method based on computing power and refrigeration system comprehensive energy consumption cost, characterized in that, The method comprises the following steps: S1, the total conventional computing power resource and the total intelligent computing power resource to be dispatched are divided into multiple conventional sub-computing power resources and intelligent sub-computing power resources to be dispatched respectively; S2, each conventional sub-computing power resource is sequentially assigned to a data center for use: S2.1, the total power consumption cost of each data center after the conventional sub-computing power resource is used is calculated, and the data center with the minimum total power consumption cost is selected for use; S2.2, S2.1 is repeated until the sum of the used conventional sub-computing power resources reaches the total conventional computing power resource to be dispatched; S3, each intelligent sub-computing power resource is sequentially assigned to a data center for use: S3.1, the optimal operation time period deployment scheme of the intelligent sub-computing power resource in each data center is solved by a genetic algorithm, the total power consumption cost of the data center after the intelligent sub-computing power resource is used is calculated according to the time-of-use electricity price of each data center, and the data center with the minimum total power consumption cost is selected for use; S3.2, S3.1 is repeated until the sum of the used intelligent sub-computing power resources reaches the total intelligent computing power resource to be dispatched.

2. The method of claim 1, wherein: The total power consumption cost calculation formula of the conventional sub-computing power resource in the data center in the step S2.1 is wherein, COST is the total power consumption cost of the ith data center after enabling the nth regular sub-computing resource, U,n (i, t) is the power consumption cost of the ith data center after enabling the nth regular sub-computing resource at time period t; COST U,n The formula for COST(i,t) is U,n (i,t) = TOU(i,t) EU n (i,t) wherein TOU(i, t) is the time-of-use electricity price of the ith data center at time period t, EU n (i, t) is the power consumption of all the enabled regular sub-computing resources of the ith data center at time period t after the nth regular sub-computing resource is enabled.

3. The method of claim 2, wherein: The EU n The calculation formula of (i,t) is VACU n (i, t) is the value of the conventional computing power consumed by the nth conventional sub-computing power resource enabled by the ith data center at the t period after all the conventional sub-computing power resources enabled by the data center, and CCPRU(i) is the value of the conventional computing power provided by the unit power consumption of the ith data center.

4. The method of claim 1, wherein: The optimal operation time period deployment of the intelligent sub-computing power resource in any data center solved by the genetic algorithm in the step S3.1 comprises the following sub-steps: (1) initialize the population, and randomly generate N intelligent sub-computing power resource operation time period deployment schemes in the data center as individuals; (2) calculate the fitness of each individual, and the fitness is the inverse of the total power consumption cost of the data center after the intelligent sub-computing power resource is used according to the deployment scheme of the individual; (3) calculate the probability of being selected according to the fitness of all individuals, and select N individuals from the population according to the probability to obtain a preliminary new population; (4) perform the crossover or mutation operation on the individuals in the preliminary new population, and then perform individual repair to obtain a final new population; (5) calculate the fitness of each individual in the final new population, and the individual with the maximum fitness is taken as the output value of this iteration; (6) return to step (2) until the iteration number reaches a threshold value or the fitness corresponding to the iteration output value changes less than a threshold value for M consecutive times, and output the optimal deployment scheme.

5. The method of claim 4, wherein: The operation time period deployment scheme of the intelligent sub-computing power resource in the data center in the step (1) is wherein, is the runtime period deployment scheme of the mth intelligent sub-computing resource in the i th data center, is the DCI period running instruction value of the mth intelligent sub-computing resource in the i th data center, and DCI is the daily running time set for the intelligent sub-computing resource, is 0, indicating not running, and 1, indicating running; The expression for the lth individual in the population is The expression for the lth individual in the population is wherein the value of each element is randomly set and satisfies that the number of elements with a value of 1 is not greater than the actual required operation time TPI of the intelligent sub-computing power resource, and TPI < DCI.

6. The method of claim 5, wherein: The calculation formula of the fitness in the step (2) is wherein F l is the fitness of the ith individual, is the total power consumption cost of the mth intelligent sub-computing resource in the ith data center after being enabled by the lth individual scheme. The calculation formula is: wherein, is the total power consumption cost of all the regular sub-computing resources enabled in the i-th data center at the time when the m-th intelligent sub-computing resource is enabled in the i-th data center, is the total power consumption cost of all the enabled intelligent sub-computing resources in the i-th data center at the t-th time period after the m-th intelligent sub-computing resource is enabled in the i-th data center according to the l-th individual scheme. The calculation formula is wherein TOU(i, t) is the time-of-use electricity price of the ith data center at time period t, is the power consumption of all enabled intelligent sub-computing resources of the ith data center at time period t after the mth intelligent sub-computing resource of the data center is enabled according to the lth scheme. The calculation formula is: wherein, is the intelligent computing power value consumed by the mth intelligent sub-computing power resource of the ith data center in the t time period after all the enabled intelligent sub-computing power resources of the data center, and CCPRI(i) is the intelligent computing power value provided by the unit power consumption of the ith data center.

7. The method of claim 5, wherein: When the operation time period deployment scheme corresponding to any individual causes all used intelligent sub-computing power resources to consume intelligent computing power exceeding the maximum intelligent computing power that the data center can provide in any time period, a penalty term is added to the fitness, and the fitness with the penalty term is where F l is the fitness after the addition of the penalty term.

8. The method of claim 1, wherein: The conventional sub-computing power resource and the intelligent sub-computing power resource used by each data center cannot exceed the maximum value of the corresponding computing power that the data center can provide when the computing power resource is used.

9. A multi-data center multi-computing power collaborative optimization system based on computing power and refrigeration system comprehensive energy consumption cost, characterized in that, The system comprises: a computing power resource division module for dividing the total conventional computing power resource and the total intelligent computing power resource to be dispatched into multiple conventional sub-computing power resources and intelligent sub-computing power resources to be dispatched respectively; The conventional computing power scheduling module is configured to confirm each data center enabling each conventional sub-computing power resource in sequence, and includes the following sub-modules: The cost calculation sub-module is configured to calculate the total power consumption cost of each data center after enabling the conventional sub-computing power resource, and select the data center with the minimum total power consumption cost for enabling; The conventional computing power iteration sub-module is configured to repeat the operation of the cost calculation sub-module to confirm the data center enabling the next conventional sub-computing power resource until the sum of all enabled conventional sub-computing power resources reaches the total conventional computing power resource to be scheduled. The intelligent computing power scheduling module is configured to confirm each data center enabling each intelligent sub-computing power resource in sequence, and includes the following sub-modules: The optimal deployment sub-module is configured to solve the optimal runtime period deployment scheme of the intelligent sub-computing power resource in each data center by using a genetic algorithm, calculate the total power consumption cost of each data center after enabling the intelligent sub-computing power resource according to the time-of-use electricity price of each data center, and select the data center with the minimum total power consumption cost for enabling; The intelligent computing power iteration sub-module is configured to repeat the operation of the optimal deployment sub-module to confirm the data center enabling the next intelligent sub-computing power resource until the sum of all enabled intelligent sub-computing power resources reaches the total intelligent computing power resource to be scheduled.

10. A computer-readable storage medium storing one or more programs, the one or more programs comprising instructions for: The one or more programs including instructions, when executed by a computing device, cause the computing device to perform any of the methods of claims 1-8. The one or more programs including instructions, when executed by a computing device, cause the computing device to perform any of the methods of claims 1-8.

Images