Method and system for operation scheduling of cold source master and slave based on multi-load demand

Abstract

This application relates to the technical field of cold source equipment scheduling in computing centers, and discloses a method and system for scheduling the operation of cold source hosts and slaves based on multi-load demand, including: S1, predicting the load curve of the cooling system within the analysis period based on the outdoor ambient temperature of the data center, pending task data, and historical data; S2, dividing the analysis period into time periods according to the range of the load curve, performing optimization analysis on the equipment combination method in each time period, and selecting the optimization analysis results according to the principle of minimum energy consumption; S3, performing secondary screening on the selected optimization analysis results based on equipment loss data to obtain the candidate equipment combination schemes for each time period; S4, arbitrarily combining the candidate equipment combination schemes for all time periods, sorting the several groups of arbitrarily combined schemes according to the equipment switching rate, and selecting the best equipment combination scheme for each time period based on the equipment switching rate sorting results.

Description

Technical Field

[0001] This application relates to the technical field of scheduling cold source equipment in computing centers, and in particular to a method and system for scheduling the operation of cold source master and slave machines based on multi-load demand. Background Technology

[0002] During the operation of a computing center, a cooling system is required to maintain the ambient temperature for its operation. The equipment in the cooling system is usually redundant. On the one hand, this ensures the stable operation of the computing center under extreme operating conditions. On the other hand, by setting up a master and slave machine, the equipment can be scheduled and rotated according to actual needs, thereby reducing the energy consumption during equipment operation.

[0003] In existing technologies, there are various methods for scheduling the operation of chiller main units and slave units. A relatively simple method is to start and stop the chiller units in a balanced manner according to their operating time, which ensures the balance of operation time for each device. Another method follows the principle of minimum energy consumption. The system calculates the total energy consumption under different combinations based on the current cooling load and the performance curve (COP Map) of each unit, and selects the combination with the lowest energy consumption to operate. This method can ensure low energy consumption during equipment operation. The above methods can achieve equipment scheduling, but it is difficult to simultaneously meet the requirements of low energy consumption, balanced operation, and low switching rate during the operation of chiller equipment. If the principle of minimum energy consumption is strictly followed, it may reduce the overall lifespan and cause equipment "oscillation". Therefore, how to simultaneously meet the requirements of low energy consumption, balanced operation, and low switching rate through the operation scheduling of chiller main units and slave units is the fundamental problem that this invention aims to solve. Summary of the Invention

[0004] In order to simultaneously meet the requirements of low energy consumption, balanced operation and low switching rate by scheduling the operation of the chiller master and slave, this application provides a method and system for scheduling the operation of chiller master and slave based on multi-load demand.

[0005] Firstly, this application provides a method for scheduling the operation of a chiller main unit and its slave units based on multi-load demand, employing the following technical solution:

[0006] The operation scheduling method for chiller mainframes and slave units based on multi-load demand includes:

[0007] S1. Based on the outdoor ambient temperature of the data center, the data of tasks to be processed, and historical data, predict the load curve of the cooling system within the analysis period.

[0008] S2. Divide the analysis period into time periods according to the range of the load curve, perform optimization analysis on the equipment combination method in each time period, and select the optimization analysis results according to the principle of minimum energy consumption.

[0009] S3. Based on the equipment loss data, the selected optimization analysis results are further filtered to obtain the equipment combination schemes to be selected for each time period.

[0010] S4. Arbitrarily combine the candidate device combination schemes for all time periods, sort the several groups of schemes according to the device switching rate, select the best device combination scheme for each time period based on the device switching rate sorting result, set the device in the best device combination scheme as the master, and set the remaining devices as slaves.

[0011] Optionally, step S1 includes the following process:

[0012] S11. Obtain the start time, end time, and computational load of each pending task data. Accumulate the computational load of all pending task data according to the start time and end time to obtain the pending task load curve.

[0013] S12. Calculate the average temporary task load for each time period based on historical data;

[0014] S13. Determine the load cooling coefficient based on the range of outdoor ambient temperature of the data center, accumulate the average value of temporary task load on the task load curve to be processed according to the corresponding time period, and determine the load curve of the cooling system within the analysis period based on the task load curve to be processed and the load cooling coefficient.

[0015] Optionally, step S2 includes the following process:

[0016] S21. Obtain the x-coordinate of the intersection of the preset load step value and the load curve, and divide the analysis period according to the x-coordinate;

[0017] S22. Using the minimum total system energy consumption as the objective function and the equipment combination method as the variable, establish constraints, including:

[0018] The equipment load rate is greater than the preset value;

[0019] The maximum load of the equipment combination is greater than or equal to the maximum value of the load curve for the corresponding time period × p, where 1.23 < p < 1.56;

[0020] The temperature difference between chilled water supply and return water is greater than 3°C;

[0021] Set the number of iterations and the minimum total system energy consumption threshold, and optimize the equipment combination scheme based on the particle swarm optimization algorithm. Stop the iteration when the set number of iterations is reached, and obtain the top N equipment combination schemes in the iteration process in order of total system energy consumption from low to high.

[0022] Determine whether the total system energy consumption of the Nth device combination scheme is less than or equal to the minimum total system energy consumption threshold:

[0023] If so, the top N equipment combination schemes will be used as the optimization analysis results;

[0024] If not, the equipment combination scheme with total system energy consumption less than or equal to the minimum total system energy consumption threshold will be selected as the optimization analysis result.

[0025] Optionally, step S3 includes the following process:

[0026] S31. Calculate the loss rate of each device based on the loss data;

[0027] S32. Calculate the average loss rate of all devices, and determine the selection weight of the devices based on the range of the difference between the loss rate and the average loss rate. The smaller the difference between the loss rate and the average loss rate, the greater the selection weight.

[0028] S33. Calculate the average selection weight of all devices in each device combination scheme and the energy consumption difference between the system total energy consumption threshold and the system total energy consumption.

[0029] For each equipment combination scheme, the average selection weight and energy consumption difference are normalized. The normalized average selection weight and energy consumption difference are then weighted and summed according to preset weights to obtain the selection priority value. The equipment combination schemes are then screened a second time in descending order of selection priority value.

[0030] Optionally, the loss calculation process for each device includes:

[0031] Feature extraction is performed on the loss data to obtain the total runtime and high-load runtime of the equipment;

[0032] The additional life loss coefficient is determined based on the range of high load running time and the range of the corresponding maximum load percentage. The additional life loss time is obtained by multiplying the high load running time by the additional life loss coefficient.

[0033] The ratio of the sum of total runtime and additional life loss time to the rated life of the equipment is used as the wear rate of each device.

[0034] Optionally, step S4 includes the following process:

[0035] For any combination of several schemes, obtain the number of devices started (N1) and stopped (N2) during the device combination switching process in adjacent time periods for each scheme;

[0036] Obtain the number of devices NQ in the previous time period and the number of devices NH in the next time period in the corresponding candidate device combination schemes for adjacent time periods in chronological order.

[0037] The stopping rate of equipment in the adjacent time period is determined by the ratio of the number of stopped devices N2 to the number of devices in the previous time period NQ.

[0038] The activation rate of equipment in the adjacent time period is determined by the ratio of the number of devices activated N1 to the number of devices activated in the subsequent time period NH.

[0039] The average stop rate and average start rate of all adjacent time periods are obtained respectively. The average stop rate and average start rate are weighted and summed according to the preset weights to obtain the device switching rate. The sum of the weights corresponding to the average stop rate and average start rate is 1.

[0040] The solution with the lowest equipment switching rate is taken as the best solution, and the best equipment combination solution for each time period is obtained based on the best solution.

[0041] Optionally, the loss data is updated at the end of each analysis cycle.

[0042] Secondly, this application provides an operation scheduling system for chiller main units and slave units based on multi-load demand, adopting the following technical solution:

[0043] A scheduling system for chiller hosts and slaves based on multi-load demand, wherein the system operates any one of the above-described scheduling methods for chiller hosts and slaves based on multi-load demand, including:

[0044] The load prediction module is used to predict the load curve of the cooling system within the analysis period based on the outdoor ambient temperature of the data center, the data of tasks to be processed, and historical data.

[0045] The optimization analysis module is used to divide the analysis period into time periods according to the range of the load curve, perform optimization analysis on the equipment combination method in each time period, and select the optimization analysis results according to the principle of minimum energy consumption.

[0046] The secondary screening module is used to perform secondary screening on the selected optimization analysis results based on the equipment loss data to obtain the candidate equipment combination scheme for each time period.

[0047] The switching rate filtering module is used to arbitrarily combine the candidate device combination schemes for all time periods, sort the several groups of schemes according to the device switching rate, select the best device combination scheme for each time period based on the device switching rate sorting results, set the device in the best device combination scheme as the master, and set the remaining devices as slaves.

[0048] In summary, this application includes at least one of the following beneficial technical effects:

[0049] This invention predictively judges the load status within the analysis period, and combines the optimization analysis, secondary screening, and switching rate screening process of equipment combination schemes. It can select the scheme with the lowest possible switching rate while ensuring low energy consumption and balanced use. It achieves the simultaneous satisfaction of low energy consumption, balanced operation, and low switching rate requirements by scheduling the operation of the cold source host and slave units. Attached Figure Description

[0050] Figure 1 This is a flowchart of the operation scheduling system of the cold source host and slave in this invention.

[0051] Figure 2 This is a framework diagram of the operation scheduling method of the cold source master and slave in this invention.

[0052] Figure 3 This is a flowchart of step S1 in this invention.

[0053] Figure 4 This is a flowchart of step S2 in this invention.

[0054] Figure 5 This is a flowchart of step S3 in this invention. Detailed Implementation

[0055] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0056] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0057] This application discloses a method and system for scheduling the operation of a chiller main unit and its slave units based on multi-load demand. Specific implementation methods are detailed in conjunction with the appendix. Figures 1-5 For detailed explanation, please refer to... Figure 2 The system includes a load prediction module, an optimization analysis module, a secondary screening module, and a switching rate screening module. These modules work together through physical connections and data transmission to jointly complete the operation scheduling of the cold source host and slave units.

[0058] In actual operation, the load prediction module is used as a reference. Figure 1 , Figure 3First, the load curve of the cooling system within the analysis period is predicted based on the outdoor ambient temperature of the data center, the data of tasks to be processed, and historical data. The outdoor ambient temperature can be obtained through IoT devices. The data of tasks to be processed includes periodic batch computing tasks, AI model training tasks, etc. The approximate computational range and start and end times of these tasks are fixed. Historical data is used to predict temporary tasks. In this embodiment, the analysis period is one day. Therefore, the process of predicting the load curve of the cooling system includes: first, obtaining the start time, end time, and computational load of each data of tasks to be processed. The computational load is obtained by selecting the average value according to the computational load range. The computational load of all data of tasks to be processed is accumulated according to the start time and end time to obtain the computational load of each data of tasks to be processed. The task load curve clearly reflects the fixed load status of the computing center. Then, based on historical data, the average temporary task load for each time period is statistically analyzed to predict the temporary task load. The load cooling coefficient is determined based on the range of outdoor ambient temperature in the data center. This process is achieved by comparing the temperature range in the test data with the load cooling coefficient. Therefore, the load cooling coefficient reflects a linear relationship between the computing center load and the cooling system load. The average temporary task load is accumulated on the task load curve for each corresponding time period. Based on the task load curve and the load cooling coefficient, the cooling system load curve for the analysis period is determined. Therefore, the cooling system load curve can predict the load status within the analysis period.

[0059] In the above scheme, the load of the computing center and the load of the cooling system are not completely linearly related, but in terms of the total amount of heat, the heat dissipation demand and the IT load are basically linear, so they are often regarded as linear relationships for the convenience of calculation.

[0060] Then, through the optimization analysis module, referring to Figure 1 The analysis period is divided into time periods based on the range of the load curve. Within each time period, an optimization analysis of the equipment combination is performed. The results of the optimization analysis are selected based on the principle of minimum energy consumption. Figure 4First, the abscissa of the intersection of the preset load step value and the load curve is obtained. The analysis period is divided according to the abscissa. Several load values ​​are selected as step values ​​based on empirical data. By dividing the time period corresponding to the load curves at the same step, the optimization analysis of the equipment combination method can be performed separately for each time period. The optimization algorithm in this embodiment adopts the particle swarm optimization algorithm. First, the minimum total system energy consumption is used as the objective function, and the equipment combination method is used as the variable. Constraints are established, including: the equipment load rate is greater than the preset value; the maximum load of the equipment combination is ≥ the maximum value of the load curve in the corresponding time period × p, 1.23 < p < 1.56; the temperature difference between chilled water supply and return water is greater than 3°C; finally, the number of iterations and the maximum value are set. The minimum total system energy consumption threshold is obtained by comparing the load curve for that period with the energy consumption correspondence in empirical data (by integrating the load interval and obtaining the minimum total system energy consumption threshold based on the integral value and the energy consumption correspondence). The device combination scheme is optimized using a particle swarm optimization algorithm. Iteration stops when a set number of iterations is reached. The top N device combination schemes are obtained in ascending order of total system energy consumption. It is then determined whether the total system energy consumption of the Nth device combination scheme is less than or equal to the minimum total system energy consumption threshold. If yes, the top N device combination schemes are used as the optimization analysis result; otherwise, the device combination schemes with total system energy consumption less than or equal to the minimum total system energy consumption threshold are selected as the optimization analysis result.

[0061] In the above scheme, theoretically there is a situation where the energy consumption of the top N equipment combinations is greater than the minimum total system energy consumption threshold. However, since the energy consumption correspondence is set according to the average data (not the minimum energy consumption data) during the process of obtaining the minimum total system energy consumption threshold, there will always be a better result in the optimization process. That is, there will always be equipment combinations whose total system energy consumption is less than or equal to the minimum total system energy consumption threshold.

[0062] Then, through the secondary screening module, referring to Figure 1 Based on the equipment loss data, the selected optimization analysis results are further filtered to obtain the candidate equipment combination schemes for each time period; refer to Figure 5First, the loss rate of each device is calculated based on the loss data. Then, the average loss rate of all devices is calculated. The selection weight of a device is determined based on the range of the difference between its loss rate and the average loss rate; the smaller the difference, the greater the selection weight. Specific weight settings are based on test data. The average selection weight of all devices in each device combination scheme and the energy consumption difference between the system's total energy consumption threshold and the system's total energy consumption are calculated. The average selection weight and energy consumption difference of each device combination scheme are normalized. The normalized average selection weight and energy consumption difference are then weighted and summed according to preset weights. These preset weights are set based on a balance between energy consumption and usage equilibrium to obtain selection priority values. The device combination schemes are then screened a second time according to the selection priority values ​​from largest to smallest. This process ensures that the screened device combination schemes have low energy consumption while maintaining a balance in the usage status of all devices.

[0063] In the above scheme, the loss degree calculation process for each device includes: extracting features from the loss data to obtain the total operating time and high-load operating time of the device; determining the additional life loss coefficient based on the range of high-load operating time and the corresponding range of maximum load percentage, which is based on the comparison relationship between different load percentage ranges and the additional life loss coefficient. This comparison relationship is set based on the fitting of historical operating data of multiple sets of scrapped cold source equipment. The high-load operating time is multiplied by the additional life loss coefficient to obtain the additional life loss time; the ratio of the sum of the total operating time and the additional life loss time to the rated life of the equipment is used as the loss degree of each device. Therefore, by using the loss degree of the equipment, the life status of the cold source equipment can be judged more accurately and objectively.

[0064] In addition, the loss data in this example is updated after each analysis cycle, which avoids the impact of loss data adjustments on the operation and scheduling of the cold source host and slave units.

[0065] Finally, through the switching rate filtering module, refer to Figure 1All available device combination schemes for all time periods are arbitrarily combined. These combinations are then sorted by device switching rate. The optimal device combination scheme for each time period is selected based on the switching rate ranking. Devices in the optimal combination scheme are designated as masters, and the remaining devices as slaves. First, for each arbitrary combination scheme, the number of devices started (N1) and stopped (N2) during the switching process between adjacent time periods are obtained. Then, the number of devices in the preceding time period (NQ) and the number of devices in the following time period (NH) are obtained in chronological order. The stopping rate of devices in the adjacent time period is determined by the ratio of the number of stopped devices (N2) to the number of devices in the preceding time period (NQ). The starting rate of devices in the adjacent time period is determined by the ratio of the number of started devices (N1) to the number of devices in the following time period (NH). The average stopping rate and average starting rate for all adjacent time periods are obtained, and the average stopping rate and average starting rate are weighted according to a preset weight. The average rates are weighted and summed. The preset weights here are set according to the degree of influence of the start-up and shutdown process of the cold source equipment on the "oscillation" phenomenon. It is necessary to ensure that the sum of the weights corresponding to the average shutdown rate and the average start-up rate is 1. Finally, the equipment switching rate is obtained, and the scheme with the lowest equipment switching rate is taken as the optimal scheme. Based on the optimal scheme, the optimal equipment combination scheme for each time period is obtained. Therefore, through this process, while ensuring low energy consumption and balanced use, the scheme with the lowest possible switching rate can be selected. Then, by scheduling the operation of the cold source host and slave, the requirements of low energy consumption, balanced operation and low switching rate can be met simultaneously.

[0066] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for operation scheduling of a master and slaves of a cold source based on multi-load demand, characterized in that, include: S1. Based on the outdoor ambient temperature of the data center, the data of tasks to be processed, and historical data, predict the load curve of the cooling system within the analysis period. S2. Divide the analysis period into time periods according to the range of the load curve, perform optimization analysis on the equipment combination method in each time period, and select the optimization analysis results according to the principle of minimum energy consumption. S3. Based on the equipment loss data, the selected optimization analysis results are further filtered to obtain the equipment combination schemes to be selected for each time period. S4. Arbitrarily combine the candidate equipment combination schemes for all time periods, sort the several groups of schemes according to the equipment switching rate, select the best equipment combination scheme for each time period based on the equipment switching rate sorting result, set the equipment in the best equipment combination scheme as the master, and set the other equipment as the slave. Step S1 includes the following process: S11. Obtain the start time, end time, and computational load of each pending task data. Accumulate the computational load of all pending task data according to the start time and end time to obtain the pending task load curve. S12. Calculate the average temporary task load for each time period based on historical data; S13. Determine the load cooling coefficient based on the range of outdoor ambient temperature of the data center, accumulate the average value of temporary task load on the task load curve to be processed according to the corresponding time period, and determine the load curve of the cooling system within the analysis period based on the task load curve to be processed and the load cooling coefficient. Step S2 includes the following process: S21. Obtain the x-coordinate of the intersection of the preset load step value and the load curve, and divide the analysis period according to the x-coordinate; S22. Using the minimum total system energy consumption as the objective function and the equipment combination method as the variable, establish constraints, including: The equipment load rate is greater than the preset value; Maximum load of the device combination > maximum value of the corresponding period load curve p, 1.23 < p < 1.56; The temperature difference between chilled water supply and return water is greater than 3°C; Set the number of iterations and the minimum total system energy consumption threshold, and optimize the equipment combination scheme based on the particle swarm optimization algorithm. Stop the iteration when the set number of iterations is reached, and obtain the top N equipment combination schemes in the iteration process in order of total system energy consumption from low to high. Determine whether the total system energy consumption of the Nth device combination scheme is less than or equal to the minimum total system energy consumption threshold: If so, the top N equipment combination schemes will be used as the optimization analysis results; If not, the equipment combination scheme with total system energy consumption less than or equal to the minimum total system energy consumption threshold will be selected as the optimization analysis result.

2. The operation scheduling method of a master and slaves of a cold source based on a multi-load demand according to claim 1, wherein Step S3 includes the following process: S31. Calculate the loss rate of each device based on the loss data; S32. Calculate the average loss rate of all devices, and determine the selection weight of the devices based on the range of the difference between the loss rate and the average loss rate. The smaller the difference between the loss rate and the average loss rate, the greater the selection weight. S33. Calculate the average selection weight of all devices in each device combination scheme and the energy consumption difference between the system total energy consumption threshold and the system total energy consumption. For each equipment combination scheme, the average selection weight and energy consumption difference are normalized. The normalized average selection weight and energy consumption difference are then weighted and summed according to preset weights to obtain the selection priority value. The equipment combination schemes are then screened a second time in descending order of selection priority value.

3. The operation scheduling method of a master and slaves of a cold source based on a multi-load demand according to claim 2, characterized in that, The loss calculation process for each device includes: Feature extraction is performed on the loss data to obtain the total runtime and high-load runtime of the equipment; The additional life loss coefficient is determined based on the range of high load running time and the range of the corresponding maximum load percentage. The additional life loss time is obtained by multiplying the high load running time by the additional life loss coefficient. The ratio of the sum of total runtime and additional life loss time to the rated life of the equipment is used as the wear rate of each device.

4. The operation scheduling method of a master and slaves of a cold source based on a multi-load demand according to claim 1, wherein, Step S4 includes the following process: For any combination of several schemes, obtain the number of devices started (N1) and stopped (N2) during the device combination switching process in adjacent time periods for each scheme; Obtain the number of devices NQ in the previous time period and the number of devices NH in the next time period in the corresponding candidate device combination schemes for adjacent time periods in chronological order. The stopping rate of equipment in the adjacent time period is determined by the ratio of the number of stopped devices N2 to the number of devices in the previous time period NQ. The activation rate of equipment in the adjacent time period is determined by the ratio of the number of devices activated N1 to the number of devices activated in the subsequent time period NH. The average stop rate and average start rate of all adjacent time periods are obtained respectively. The average stop rate and average start rate are weighted and summed according to the preset weights to obtain the device switching rate. The sum of the weights corresponding to the average stop rate and average start rate is 1. The solution with the lowest equipment switching rate is taken as the best solution, and the best equipment combination solution for each time period is obtained based on the best solution.

5. The operation scheduling method of a master and slaves of a cold source based on a multi-load demand according to claim 3, wherein, The loss data is updated at the end of each analysis cycle.

6. A system for scheduling operation of a master and slave of a cold source based on multi-load demand, characterized by, The system operates a scheduling method for a cooling source host and slave units based on multiple load demands, as described in any one of claims 1-5, including: The load prediction module is used to predict the load curve of the cooling system within the analysis period based on the outdoor ambient temperature of the data center, the data of tasks to be processed, and historical data. The optimization analysis module is used to divide the analysis period into time periods according to the range of the load curve, perform optimization analysis on the equipment combination method in each time period, and select the optimization analysis results according to the principle of minimum energy consumption. The secondary screening module is used to perform secondary screening on the selected optimization analysis results based on the equipment loss data to obtain the candidate equipment combination scheme for each time period. The switching rate filtering module is used to arbitrarily combine the candidate device combination schemes for all time periods, sort the several groups of schemes according to the device switching rate, select the best device combination scheme for each time period based on the device switching rate sorting results, set the device in the best device combination scheme as the master, and set the remaining devices as slaves.

Images