An algorithmic electric non-local mapping scheduling method and system based on space-time vector decoupling

Abstract

This application discloses a nonlocal mapping scheduling method and system for computing power based on spatiotemporal vector decoupling, including the following steps: real-time collection of power grid node information to form a node spatiotemporal dataset, screening of nodes with abnormal voltage, splitting the node spatiotemporal dataset, orthogonal processing of the strong correlation between computing power and power, sorting computing power demand and power supply according to urgency and abundance, and matching the optimal scheduling node according to spatial distance. In this invention, multi-dimensional spatiotemporal data of power grid nodes are collected in real time to complete the decoupling of spatial and electrical time-series attributes, accurately screen effective scheduling nodes, significantly reduce the data volume and computational load of scheduling operations, perform residual orthogonal processing on the strong correlation between computing power and power to eliminate collinearity interference and ensure scheduling accuracy, hierarchically sort computing power demand and power supply, and complete accurate matching of supply and demand by combining spatial distance, thereby reducing scheduling losses and improving the execution efficiency and configuration rationality of computing power collaborative scheduling.

Description

Technical Field

[0001] This application relates to the technical field of power dispatching, and in particular to a computational nonlocal mapping dispatching method and system based on spatiotemporal vector decoupling. Background Technology

[0002] The field of power dispatching technology covers core aspects such as power grid operation status monitoring, power output regulation, load distribution control, transmission power flow management, and collaborative allocation of computing resources. It achieves power grid supply and demand balance and safe and stable operation by collecting power grid electrical measurement data, performing optimization calculations, issuing dispatching instructions, and feeding back the execution status. Among them, traditional computing non-local mapping dispatching refers to the dispatching processing for the supply and demand mismatch between computing resources and power resources at different spatiotemporal nodes. It usually uses spatiotemporal coupling modeling, unified iteration of global variables, fixed-dimensional vector mapping, and linear constraint substitution to solve the problem to complete the construction of dispatching relationships and the generation of dispatching schemes.

[0003] Existing computing-coordinated scheduling adopts a spatiotemporal coupled modeling logic, binding the spatial attributes of power grid nodes with electrical temporal attributes. This approach fails to enable differentiated applications of the two types of attributes, easily leading to data redundancy and computational errors, increasing computational load, lengthening scheduling response cycles, and making it difficult to adapt to high-frequency computing power scheduling requirements. Furthermore, the scheduling relationship construction process fails to resolve the inherent correlation between computing power and power fluctuation characteristics, which can easily cause scheduling mapping errors due to mutual interference between characteristics, thus failing to guarantee the accuracy of scheduling decisions. At the same time, it does not prioritize computing power demand and power supply, nor does it conduct supply effectiveness assessments based on spatial dimension information, which can easily lead to supply-demand mismatches and high power transmission losses, making it difficult to balance scheduling execution efficiency with the rationality of resource allocation. Summary of the Invention

[0004] This application provides a computational nonlocal mapping scheduling method and system based on spatiotemporal vector decoupling to solve the above-mentioned technical problems.

[0005] In view of this, this application provides a nonlocal mapping scheduling method for computing based on spatiotemporal vector decoupling, comprising the following steps: S1: Real-time collection of electrical data, unique identifier codes and location data of each power grid node, while reading the collection time and summarizing them to form a node spatiotemporal dataset; S2: Based on the node spatiotemporal dataset, obtain the long-term average voltage and daily average voltage, calculate the difference between the historical average voltage and the current daily average voltage, and set a judgment threshold. When the difference is greater than the judgment threshold, split the location information and time voltage data in the node spatiotemporal dataset to obtain the node spatial set and the node time electrical set. S3: Set a strong correlation threshold, collect computing power fluctuation data of servers in each data center, calculate the fluctuation amplitude within one hour to obtain the computing power feature sequence, extract the voltage value sequence of the corresponding time period of the node time electrical concentration, calculate the Pearson correlation coefficient of the two sequences, and when the absolute value exceeds the strong correlation threshold, perform residual orthogonal processing on the computing power feature and power feature to obtain the orthogonal computing power feature set and power feature set. S4: Sort the computing power feature set in descending order according to the urgency of computing power supply demand to generate a set of computing power demand to be scheduled; sort the power feature set in descending order according to the power abundance to generate a set of power supply available for scheduling. S5: Based on the node space set, calculate the spatial distance between the first data center of the scheduling computing power demand set and each power grid node in the scheduling power supply set, then calculate the effective supply value, and send the computing power demand to the power grid node with the largest effective supply value.

[0006] Preferably, the specific steps of S1 are as follows: S101: Collect the analog current signal from the current transformer and the analog voltage signal from the voltage transformer at each power grid node, convert them into digital values ​​representing current and voltage values ​​via an analog-to-digital converter, and read the longitude coordinates, latitude coordinates, altitude, unique identifier code, and upstream and downstream node identifier code lists of the power grid node to form a node spatial attribution parameter set. S102: Based on the node space attribution parameter set, read the time value from the clock chip, and append the time value to the digital quantity representing the current value and the digital quantity representing the voltage value to obtain the electrical quantity with time stamp. S103: Call the information from the time-stamped electrical quantities and the node spatial attribution parameter set, and combine the two to form a node spatiotemporal dataset.

[0007] Preferably, the specific steps of S2 are as follows: S201: Based on the node spatiotemporal dataset, the arithmetic mean of the voltage values ​​of the power grid nodes at one thousand consecutive time points is selected as the short-time average voltage value, and the arithmetic mean of the voltage values ​​of the power grid nodes at twenty-four consecutive hours is selected as the daily average voltage value. S202: Calculate the absolute value of the difference between the short-time average voltage value and the daily average voltage value, set a judgment threshold, compare the absolute value with the set judgment threshold, if the absolute value of the difference is greater than the threshold, it is judged as exceeding the limit, otherwise it is not exceeding the limit, and a voltage fluctuation exceeding the limit mark is obtained.

[0008] Preferably, based on the voltage fluctuation exceeding limit marker in S202, when the marker is exceeding the limit, the longitude coordinate value, latitude coordinate value, altitude value, unique identifier code and upstream and downstream node identifier code list are extracted from the node spatiotemporal dataset and incorporated into the node spatial set, and the time-marked electrical measurement data are incorporated into the node time electrical set, thus obtaining the node spatial set and the node time electrical set.

[0009] Preferably, the specific steps of S3 are as follows: S301: Set a strong correlation threshold, collect the percentage values ​​of CPU usage and memory usage of servers in each data center, calculate the difference between the maximum and minimum CPU usage percentage within a continuous 60 minutes as the computing power load fluctuation amplitude, extract the effective voltage value of the corresponding 60-minute period from the node time electrical collection, and obtain the computing power fluctuation amplitude sequence and voltage sequence. S302: Calculate the Pearson correlation coefficient between the computing power fluctuation amplitude sequence and the voltage sequence values, compare the absolute value of the correlation coefficient with the strong correlation threshold, and obtain the correlation exceeding limit marker and the correlation coefficient value.

[0010] Preferably, in S302, when the correlation exceeds the limit marker and correlation coefficient value, residual operation is performed on the computing power load fluctuation amplitude sequence and the voltage effective value numerical sequence. The voltage-related components are removed from the computing power load fluctuation amplitude sequence to obtain an orthogonalized computing power feature set, and the computing power-related components are removed from the voltage effective value numerical sequence to obtain an orthogonalized power feature set, thus obtaining the orthogonalized computing power feature set and power feature set.

[0011] Preferably, the specific steps of S4 are as follows: S401: Call the computing power feature set, extract the values ​​representing the urgency of computing power supply demand, sort them from largest to smallest, and obtain the computing power demand set to be scheduled; S402: Call the power feature set, extract the values ​​representing the availability of power, arrange them in descending order of value, and obtain the dispatchable power supply set.

[0012] Preferably, the specific steps of S5 are as follows: S501: Based on the node spatial set, call the longitude and latitude coordinates of the first data center in the set of computing power demand to be scheduled, call the longitude and latitude coordinates of each power grid node in the set of schedulable power supply, calculate the spatial distance between the first data center and each power grid node, and obtain the candidate node spatial distance set. S502: Based on the spatial distance set of candidate nodes, query the transmission loss coefficient table for each distance value to obtain the corresponding loss coefficient, and subtract the loss coefficient from the power abundance value of the corresponding node in the schedulable power supply set to obtain the effective supply capacity set of candidate nodes.

[0013] Preferably, the values ​​of the effective supply capacity set of the candidate nodes in S502 are compared, and the unique identifier code of the power grid node corresponding to the maximum value is selected as the matching object to generate the matching power grid node identifier code.

[0014] This application also provides a nonlocal mapping scheduling system for computing based on spatiotemporal vector decoupling. The system includes: a node data acquisition module, a node decoupling and filtering module, a feature orthogonal decoupling module, a supply and demand sorting generation module, and a spatial mapping matching module. The node data acquisition module is used to perform real-time acquisition of electrical data, unique identification codes and location data of each power grid node, and at the same time read the acquisition time and summarize them to form a node spatiotemporal dataset. The node decoupling and filtering module is used to perform operations based on the node spatiotemporal dataset to obtain the long-term average voltage and daily average voltage, calculate the difference between the historical average voltage and the current daily average voltage, and set a judgment threshold. When the difference is greater than the judgment threshold, the location information and time voltage data in the node spatiotemporal dataset are split to obtain the node spatial set and the node time electrical set. The orthogonal decoupling module is used to set a strong correlation threshold, collect computing power fluctuation data of each data center server, calculate the fluctuation amplitude within one hour to obtain the computing power feature sequence, extract the voltage value sequence of the corresponding time period of the node time electrical concentration, calculate the Pearson correlation coefficient of the two sequences, and when the absolute value exceeds the strong correlation threshold, perform residual orthogonal processing on the computing power feature and the power feature to obtain the orthogonal computing power feature set and the power feature set. The supply and demand sorting generation module is used to sort the computing power feature set in descending order according to the urgency of computing power supply demand, generate a set of computing power demand to be scheduled, and sort the power feature set in descending order according to the power abundance, generate a set of schedulable power supply. The spatial mapping and matching module is used to perform calculations based on the node spatial set, calculating the spatial distance between the first data center of the scheduling computing power demand set and each power grid node in the scheduling power supply set, then calculating the effective supply value, and sending the computing power demand to the power grid node with the largest effective supply value.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: In this invention, multi-dimensional spatiotemporal data of power grid nodes are collected in real time to decouple spatial and electrical temporal attributes, accurately select effective scheduling nodes, significantly reduce the data volume and computational load of scheduling operations, perform residual orthogonal processing on the strong correlation between computing power and power to eliminate collinearity interference, ensure scheduling accuracy, hierarchically sort computing power demand and power supply, and combine spatial distance to achieve accurate supply and demand matching, reduce scheduling losses, and improve the execution efficiency and configuration rationality of computing-electricity collaborative scheduling. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of a computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling provided in an embodiment of this application. Figure 2 This is a flowchart illustrating the steps of a computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling provided in an embodiment of this application. Figure 3 This is a detailed flowchart illustrating step S1 of a nonlocal mapping scheduling method for computing power based on spatiotemporal vector decoupling provided in an embodiment of this application. Figure 4 This is a detailed flowchart illustrating step S2 of a nonlocal mapping scheduling method for computing power based on spatiotemporal vector decoupling provided in this embodiment of the application. Figure 5 This is a detailed flowchart illustrating step S3 of a computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling provided in this embodiment of the application. Figure 6 This is a detailed flowchart illustrating step S4 of a computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling provided in an embodiment of this application. Figure 7 This is a detailed flowchart illustrating step S5 of a computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling provided in an embodiment of this application. Figure 8 This is a system module diagram of a computational nonlocal mapping scheduling system based on spatiotemporal vector decoupling provided in the embodiments of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0019] For easier understanding, please refer to Figure 1 and Figure 2 This application provides a nonlocal mapping scheduling method for computing based on spatiotemporal vector decoupling, comprising the following steps: S1: Real-time collection of electrical data, unique identifier codes and location data of each power grid node, while reading the collection time and summarizing them to form a node spatiotemporal dataset; S2: Based on the node spatiotemporal dataset, obtain the long-term average voltage and daily average voltage, calculate the difference between the historical average voltage and the current daily average voltage, and set a judgment threshold. When the difference is greater than the judgment threshold, split the location information and time voltage data in the node spatiotemporal dataset to obtain the node spatial set and the node time electrical set. S3: Set a strong correlation threshold, collect computing power fluctuation data of servers in each data center, calculate the fluctuation amplitude within one hour to obtain the computing power feature sequence, extract the voltage value sequence of the corresponding time period of the node time electrical concentration, calculate the Pearson correlation coefficient of the two sequences, and when the absolute value exceeds the strong correlation threshold, perform residual orthogonal processing on the computing power feature and power feature to obtain the orthogonal computing power feature set and power feature set. S4: Sort the computing power feature set in descending order according to the urgency of computing power supply demand to generate a set of computing power demand to be scheduled; sort the power feature set in descending order according to the power abundance to generate a set of power supply available for scheduling. S5: Based on the node space set, calculate the spatial distance between the first data center of the scheduling computing power demand set and each power grid node in the scheduling power supply set, then calculate the effective supply value, and send the computing power demand to the power grid node with the largest effective supply value.

[0020] Please see Figure 3 The specific steps of S1 are as follows: The specific steps of S1 are as follows: S101: Collect the analog current signal from the current transformer and the analog voltage signal from the voltage transformer at each power grid node, convert them into digital values ​​representing current and voltage values ​​via an analog-to-digital converter, and read the longitude coordinates, latitude coordinates, altitude, unique identifier code, and upstream and downstream node identifier code lists of the power grid node to form a node spatial attribution parameter set. For each power grid node, two analog electrical signals are simultaneously acquired by connecting to the secondary output terminals of current transformers and voltage transformers through corresponding acquisition channels. The acquired analog current and voltage signals are then sent to a signal conditioning circuit to perform range adaptation and low-pass filtering, removing high-frequency interference components outside the sampling range. The conditioned analog signal is then linearly mapped to the rated input range of the analog-to-digital converter (ADC). The ADC performs successive approximation operations to convert the analog signal to a digital signal, outputting the corresponding decimal numerical code. Finally, range conversion is used to obtain the corresponding digital signal. The digital quantities of current and voltage on the primary side, such as the 2.5A current signal collected on the secondary side, are converted to obtain the corresponding 300A current digital quantity on the primary side, and the 50V voltage signal collected on the secondary side is converted to obtain the corresponding 5000V voltage digital quantity on the primary side. Then, the node's longitude coordinates, latitude coordinates, altitude values, unique node identifier code, and upstream and downstream node identifier code list are read from the node's built-in storage chip through the communication bus. The above coordinates and code data are combined into a node spatial attribution parameter set according to the preset field order. S102: Based on the node space attribution parameter set, read the time value from the clock chip, and append the time value to the digital quantity representing the current value and the digital quantity representing the voltage value to obtain the electrical quantity with time stamp. Based on the node spatial attribution parameter set, a time reading command is sent to the high-precision temperature-compensated clock chip built into the node via the communication bus. The clock chip returns a time data frame containing fields for year, month, day, hour, minute, second, and microsecond. Cyclic redundancy check is performed on the received time data frame to remove invalid data frames with mismatches. The complete time value in the valid data frame is extracted, such as 14:30:00:125 on March 11, 2026. The extracted time value is converted into timestamp data with millisecond precision according to the general encoding rules. According to the sampling order of the sampling points, the converted timestamp data is sequentially appended to the corresponding positions of the data packets of current digital quantity and voltage digital quantity obtained at the corresponding sampling time. A one-to-one correspondence between the timestamp and the electrical digital quantity of a single sampling point is established. The timestamp appending and binding operation is completed for the electrical digital quantities of all sampling points to obtain the electrical quantities with time stamps.

[0021] S103: Call the information from the time-stamped electrical quantities and the node spatial attribution parameter set, and combine the two to form a node spatiotemporal dataset; The system retrieves all data of time-stamped electrical quantities and all field information from the node spatial attribution parameter set. First, it performs a consistency check on the node identifier field of both types of data to confirm that the unique identifier of the node associated with the time-stamped electrical quantity completely matches the unique identifier code in the node spatial attribution parameter set. After the check passes, it performs field-level combination and concatenation of all field contents of the node spatial attribution parameter set and all field contents of the time-stamped electrical quantity according to the preset dataset field structure specification. For example, it completes the combination of single sampling point data entries in a fixed order of the full field of node spatial attribution parameter and the full field of time-stamped electrical quantity. The complete data entries after the combination of all sampling points are sorted in ascending order according to the timestamp value from smallest to largest, and finally the node spatiotemporal dataset is formed.

[0022] Please see Figure 4 The specific steps of S2 are as follows: S201: Based on the node spatiotemporal dataset, the arithmetic mean of the voltage values ​​of the power grid nodes at one thousand consecutive time points is selected as the short-time average voltage value, and the arithmetic mean of the voltage values ​​of the power grid nodes at twenty-four consecutive hours is selected as the daily average voltage value. Based on the node spatiotemporal dataset, the structured fields within the dataset are decomposed to extract voltage values ​​with timestamps and the time series information of the corresponding sampling points. All sampling points are sorted in ascending order according to their timestamps. The continuity of the sampling point timestamps is verified, and invalid sampling data with interrupted timestamps are removed. From the sorted valid sampling sequence, one thousand consecutive, uninterrupted voltage values ​​corresponding to each time point are selected. These one thousand selected voltage values ​​are summed, and the sum is divided by one thousand to obtain the arithmetic mean. For example, if the sum of the voltage values ​​of the one thousand selected sampling points is 100... 20000V, divided by 1000, yields 10020V. This value is used as the short-time average voltage value. Then, from the full effective sampling sequence, all effective voltage values ​​within a continuous 24-hour interval are selected. The total number of effective sampling points within this interval is counted. All voltage values ​​within the interval are summed. The sum is divided by the total number of effective sampling points. For example, if the total number of effective sampling points in 24 hours is 86,400,000, the sum of voltage values ​​is 86,227,200,000V. Dividing this by the total number yields 9980V. This value is used as the daily average voltage value. S202: Calculate the absolute value of the difference between the short-time average voltage value and the daily average voltage value, set a judgment threshold, compare the absolute value with the set judgment threshold, if the absolute value of the difference is greater than the threshold, it is judged as exceeding the limit, otherwise it is not exceeding the limit, and a voltage fluctuation exceeding the limit mark is obtained. Based on the short-time average voltage value and the daily average voltage value, a difference calculation is performed on the two values. The absolute value of the difference result is taken. For example, if the short-time average voltage value is 10020V and the daily average voltage value is 9980V, the difference calculation result is 40V. After taking the absolute value, 40V is obtained. Then, a threshold for judging voltage fluctuation is set. The threshold setting refers to the national standards related to the rated voltage level and power quality of medium-voltage distribution networks. Taking the rated voltage of 10kV distribution network of 10000V as the benchmark, 2% of the rated voltage is taken as the judgment threshold, and the calculated threshold value is 200V. The absolute value of the calculated difference is compared with the set judgment threshold. If the absolute value of the difference is greater than the set judgment threshold, an over-limit judgment result is generated. If the absolute value of the difference is less than or equal to the set judgment threshold, an under-limit judgment result is generated. Based on the final judgment result, a corresponding voltage fluctuation over-limit mark is generated. Based on the voltage fluctuation over-limit marker in S202, when the marker is over-limit, the longitude coordinate value, latitude coordinate value, altitude value, unique identifier code and upstream and downstream node identifier code list are extracted from the node spatiotemporal dataset and incorporated into the node spatial set. The electrical measurement data with time stamps are incorporated into the node time electrical set, thus obtaining the node spatial set and the node time electrical set. Based on the voltage fluctuation exceeding limit marker, the judgment result corresponding to the voltage fluctuation exceeding limit marker is read. When the judgment result corresponding to the marker read is exceeding the limit, the pre-stored node longitude coordinate value, latitude coordinate value, altitude value, node unique identifier code, and upstream and downstream node identifier code list are extracted from the node spatiotemporal dataset. All extracted spatial and identifier data are completely included in the node spatial set according to the preset field arrangement order. At the same time, all time-stamped electrical measurement data within the corresponding exceeding time period are extracted from the node spatiotemporal dataset. The extracted electrical measurement data are arranged in ascending order by timestamp. The arranged full set of electrical measurement data is included in the node time electrical set. The classification and collection operation of the two types of datasets is completed, resulting in the node spatial set and the node time electrical set.

[0023] Please see Figure 5 The specific steps of S3 are as follows: S301: Set a strong correlation threshold, collect the percentage values ​​of CPU usage and memory usage of servers in each data center, calculate the difference between the maximum and minimum CPU usage percentage within a continuous 60 minutes as the computing power load fluctuation amplitude, extract the effective voltage value of the corresponding 60-minute period from the node time electrical collection, and obtain the computing power fluctuation amplitude sequence and voltage sequence. Referring to the linear correlation classification standard of Pearson correlation coefficient, a strong correlation threshold of 0.8 was set. Through the in-band management interface of the data center server, the CPU occupancy percentage and memory occupancy percentage of each data center server were collected at a fixed sampling interval of 1 minute. The timestamp corresponding to each sampled data was recorded synchronously. The collected full time series data was sorted in ascending order by timestamp. The continuity of sampling time was verified, and invalid data with interrupted timestamps were removed. Complete and valid sampling data for 60 consecutive minutes was selected. The maximum and minimum values ​​of CPU occupancy percentage data in this period were extracted. For example, if the maximum CPU occupancy was 92% and the minimum was 35% in 60 minutes, the difference between the two was calculated to be 57%, which was used as the computing power load fluctuation amplitude for this period. The amplitude was calculated for the whole period using a fixed sliding window to generate a computing power fluctuation amplitude sequence. From the node time electrical concentration, a time interval that was completely aligned with the above 60-minute period was matched, and the effective voltage values ​​of all sampling points in the interval were extracted and arranged in order of timestamp to obtain a voltage sequence that was synchronized with the computing power fluctuation amplitude sequence.

[0024] S302: Calculate the Pearson correlation coefficient between the computing power fluctuation amplitude sequence and the voltage sequence, compare the absolute value of the correlation coefficient with the strong correlation threshold, and obtain the correlation exceeding limit marker and the correlation coefficient value; Based on the power fluctuation amplitude sequence and voltage sequence, a timestamp consistency check is first performed on the two sets of sequences. The sampling time points of the two sets of sequences are matched one by one, and invalid data points with mismatched timestamps are removed to ensure that the number of effective samples and the sampling time points of the two sets of sequences are completely consistent. Then, Pearson correlation coefficient is calculated on the two sets of aligned effective sequence data. First, the arithmetic mean of each set of sequences is calculated. Then, the deviation of each sample point from the mean of the corresponding sequence is calculated. The sum of the products of the deviations of the two sets of sequences is counted. Then, the sum of the squares of the deviations of the two sets of sequences is calculated. The square root of the product of the two sums of squares of deviations is taken, and the sum of the products is divided by the square root to obtain the Pearson correlation coefficient. For example, the correlation coefficient calculated in this case is 0.87. The absolute value of the calculated correlation coefficient is taken and compared with the pre-set strong correlation threshold of 0.8. The corresponding judgment result is generated based on the comparison result, and a correlation exceeding the limit mark is generated simultaneously. The calculated correlation coefficient value is stored.

[0025] In S302, the correlation exceeding limit marker and correlation coefficient value are used. When the marker is exceeded, residual operation is performed on the computing power load fluctuation amplitude sequence and the voltage effective value numerical sequence. The voltage-related components are removed from the computing power load fluctuation amplitude sequence to obtain the orthogonalized computing power feature set. The computing power-related components are removed from the voltage effective value numerical sequence to obtain the orthogonalized power feature set. Thus, the orthogonalized computing power feature set and power feature set are obtained. Read the judgment result corresponding to the correlation exceeding the limit marker. When the judgment result corresponding to the read marker is exceeding the limit, retrieve the time-synchronized and aligned computing power load fluctuation amplitude sequence and voltage effective value numerical sequence. Perform residual operation on the two sets of sequences. First, with the voltage sequence as the independent variable and the computing power load fluctuation amplitude sequence as the dependent variable, fit a univariate linear regression equation and calculate the regression fitting value corresponding to the computing power load fluctuation amplitude sequence. Subtract the regression fitting value at the corresponding time point from the original computing power load fluctuation amplitude sequence to obtain the residual sequence. Use this residual sequence as the orthogonalized computing power feature set. Then, with the computing power load fluctuation amplitude sequence as the independent variable and the voltage effective value numerical sequence as the dependent variable, fit a univariate linear regression equation and calculate the regression fitting value corresponding to the voltage effective value numerical sequence. Subtract the regression fitting value at the corresponding time point from the original voltage effective value numerical sequence to obtain the corresponding residual sequence. Use this residual sequence as the orthogonalized power feature set to obtain the orthogonalized computing power feature set and power feature set.

[0026] Please see Figure 6 The specific steps of S4 are as follows: S401: Call the computing power feature set, extract the values ​​representing the urgency of computing power supply demand, sort them from largest to smallest, and obtain the computing power demand set to be scheduled; Based on the orthogonalized computing power feature set, the structured fields in the dataset are decomposed and their integrity is verified. The integrity of the node unique identifier, sampling timestamp, and computing power feature field bound to each data entry is checked one by one. Invalid data entries with missing fields or abnormal timestamps are removed. From the valid data entries that pass the verification, the corresponding quantitative values ​​representing the urgency of computing power supply and demand are extracted. These values ​​correspond to the fusion quantitative results of the data center computing power load gap rate, the proportion of pending computing tasks in the queue, and the proportion of computing power resource occupation exceeding the threshold. The extracted full quantitative values ​​are bound with the corresponding node identifier and timestamp information. The bound full data entries are sorted in descending order according to the extracted urgency quantitative values. For entries with equal values, a secondary sort is performed according to the order of sampling timestamps. After the sorting operation of the full data is completed, all sorted data entries are integrated to obtain the set of computing power demand to be scheduled. S402: Call the power feature set, extract the values ​​representing the availability of power, sort them from largest to smallest, and obtain the dispatchable power supply set; The structured fields within the dataset are decomposed and their integrity is verified. The integrity of the unique grid node identifier, sampling timestamp, and power characteristic fields bound to each data entry is checked one by one. Invalid data entries with missing fields or mismatches between timestamps and computing power sequences are removed. From the valid data entries that pass the verification, the corresponding quantitative values ​​representing the availability of electricity are extracted. These values ​​correspond to the fusion quantitative results of the proportion of renewable energy generation connected to the grid node, the remaining power allocation capacity, and the power supply margin of the distribution network node. The extracted full quantitative values ​​are bound to the corresponding node identifier and timestamp information. The bound full data entries are sorted in descending order according to the extracted availability quantitative values. For entries with equal values, a secondary sort is performed according to the order of sampling timestamps. After the sorting operation of the full data is completed, all sorted data entries are integrated to obtain the dispatchable power supply set.

[0027] Please see Figure 7 The specific steps of S5 are as follows: S501: Based on the node spatial set, call the longitude and latitude coordinates of the first data center in the set of computing power demand to be scheduled, call the longitude and latitude coordinates of each power grid node in the set of schedulable power supply, calculate the spatial distance between the first data center and each power grid node, and obtain the candidate node spatial distance set. Based on the node space set, the structured fields within the set are decomposed, and the integrity of the node unique identifier and coordinate fields of each data entry is verified. Invalid data entries with missing fields are removed. The sorted set of computing power demand to be scheduled is retrieved, the first data entry after sorting is located, and the node unique identifier code bound to the entry is extracted. The longitude and latitude coordinate values ​​corresponding to the unique identifier code are matched from the node space set to complete the validity verification of the coordinate values. Then, all valid data entries of the schedulable power supply set are retrieved, and the grid node unique identifier code bound to each entry is extracted one by one. The longitude and latitude coordinate values ​​corresponding to each unique identifier code are matched from the node space set to complete the collection and verification of all grid node coordinate data. Based on the spherical two-point distance calculation rule, the spatial distance value between the coordinates of the first data center and the coordinates of each grid node is calculated respectively. The calculated spatial distance value is bound to the unique identifier code of the corresponding grid node. All bound valid data entries are integrated to obtain the candidate node spatial distance set. S502: Based on the spatial distance set of candidate nodes, query the transmission loss coefficient table for each distance value to obtain the corresponding loss coefficient, and subtract the loss coefficient from the power abundance value of the corresponding node in the schedulable power supply set to obtain the effective supply capacity set of candidate nodes. Integrity checks are performed on all data entries within the candidate node spatial distance set. The integrity of the unique identifier code of the power grid node and the spatial distance value of each entry is verified, and invalid data entries with missing fields are removed. A pre-constructed transmission loss coefficient table is retrieved. This coefficient table is based on measured data of transmission loss per unit distance of distribution network lines, and is divided into corresponding fixed loss coefficients according to distance intervals. For each data entry that passes the verification, its corresponding spatial distance value is read, and the preset interval to which the distance value belongs is matched. The loss coefficient for the corresponding interval is retrieved from the transmission loss coefficient table, and the loss coefficient is bound to the unique identifier code of the corresponding power grid node. The power sufficiency value corresponding to the unique identifier code of the same node is matched from the dispatchable power supply set. The difference between this value and the corresponding loss coefficient is calculated, and the result is bound to the unique identifier code of the corresponding node. All bound result entries are integrated to obtain the effective supply capacity set of the candidate nodes.

[0028] The values ​​of the effective supply capacity set of candidate nodes in S502 are compared, and the unique identifier code of the power grid node corresponding to the maximum value is selected as the matching object to generate the matching power grid node identifier code. Completeness and validity checks are performed on all data entries within the matching power grid node identifier code. The completeness of the unique identifier code of the power grid node and the effective supply capacity value of each entry is verified. Invalid data entries with missing fields or values ​​exceeding reasonable ranges are removed. For all valid data entries that pass the verification, a full pairwise comparison of the effective supply capacity values ​​is performed. The maximum value that appears during the comparison process and the unique identifier code of the power grid node corresponding to the maximum value are recorded simultaneously. After the comparison operation of all values ​​is completed, the unique identifier code of the power grid node corresponding to the final maximum value is locked. This code is used as the optimal matching object, and a standardized matching power grid node identifier code is generated based on this identifier code.

[0029] Please see Figure 8 A nonlocal mapping scheduling system for computing based on spatiotemporal vector decoupling, comprising: a node data acquisition module, a node decoupling and filtering module, a feature orthogonal decoupling module, a supply and demand sorting generation module, and a spatial mapping matching module; The node data acquisition module is used to collect electrical data, unique identification codes, and location data of each power grid node in real time, and at the same time read the acquisition time and summarize them to form a node spatiotemporal dataset. The node decoupling and filtering module is used to perform operations based on the node spatiotemporal dataset to obtain the long-term average voltage and daily average voltage, calculate the difference between the historical average voltage and the current daily average voltage, and set a judgment threshold. When the difference is greater than the judgment threshold, the location information and time voltage data in the node spatiotemporal dataset are split to obtain the node spatial set and the node time electrical set. The orthogonal decoupling module is used to set a strong correlation threshold, collect computing power fluctuation data of servers in each data center, calculate the fluctuation amplitude of one hour to obtain the computing power feature sequence, extract the voltage value sequence of the corresponding time period of the node time electrical concentration, calculate the Pearson correlation coefficient of the two sequences, and when the absolute value exceeds the strong correlation threshold, perform residual orthogonal processing on the computing power feature and the power feature to obtain the orthogonal computing power feature set and the power feature set. The supply and demand ranking generation module is used to sort the computing power feature set in descending order according to the urgency of computing power supply demand, generate the computing power demand set to be scheduled, and sort the power feature set in descending order according to the power abundance, generate the power supply set available for scheduling. The spatial mapping and matching module is used to perform calculations based on the node spatial set, calculate the spatial distance between the first data center of the scheduling computing power demand set and each power grid node in the scheduling power supply set, calculate the effective supply value, and send the computing power demand to the power grid node with the largest effective supply value.

[0030] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for mapping and scheduling of non-locality based on spatiotemporal vector decoupling, characterized in that, Includes the following steps: S1: Real-time collection of electrical data, unique identifier codes and location data of each power grid node, while reading the collection time and summarizing them to form a node spatiotemporal dataset; S2: Based on the node spatiotemporal dataset, obtain the long-term average voltage and daily average voltage, calculate the difference between the historical average voltage and the current daily average voltage, and set a judgment threshold. When the difference is greater than the judgment threshold, split the location information and time voltage data in the node spatiotemporal dataset to obtain the node spatial set and the node time electrical set. S3: Set a strong correlation threshold, collect computing power fluctuation data of servers in each data center, calculate the fluctuation amplitude within one hour to obtain the computing power feature sequence, extract the voltage value sequence of the corresponding time period of the node time electrical concentration, calculate the Pearson correlation coefficient of the two sequences, and when the absolute value exceeds the strong correlation threshold, perform residual orthogonal processing on the computing power feature and power feature to obtain the orthogonal computing power feature set and power feature set. S4: Sort the computing power feature set in descending order according to the urgency of computing power supply demand to generate a set of computing power demand to be scheduled; sort the power feature set in descending order according to the power abundance to generate a set of power supply available for scheduling. S5: Based on the node space set, calculate the spatial distance between the first data center of the scheduling computing power demand set and each power grid node in the scheduling power supply set, then calculate the effective supply value, and send the computing power demand to the power grid node with the largest effective supply value.

2. The method of claim 1, wherein, The specific steps of S1 are as follows: S101: Collect the analog current signal from the current transformer and the analog voltage signal from the voltage transformer at each power grid node, convert them into digital values ​​representing current and voltage values ​​via an analog-to-digital converter, and read the longitude coordinates, latitude coordinates, altitude, unique identifier code, and upstream and downstream node identifier code lists of the power grid node to form a node spatial attribution parameter set. S102: Based on the node space attribution parameter set, read the time value from the clock chip, and append the time value to the digital quantity representing the current value and the digital quantity representing the voltage value to obtain the electrical quantity with time stamp. S103: Call the information from the time-stamped electrical quantities and the node spatial attribution parameter set, and combine the two to form a node spatiotemporal dataset.

3. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 2, characterized in that, The specific steps of S2 are as follows: S201: Based on the node spatiotemporal dataset, the arithmetic mean of the voltage values ​​of the power grid nodes at one thousand consecutive time points is selected as the short-time average voltage value, and the arithmetic mean of the voltage values ​​of the power grid nodes at twenty-four consecutive hours is selected as the daily average voltage value. S202: Calculate the absolute value of the difference between the short-time average voltage value and the daily average voltage value, set a judgment threshold, compare the absolute value with the set judgment threshold, if the absolute value of the difference is greater than the threshold, it is judged as exceeding the limit, otherwise it is not exceeding the limit, and a voltage fluctuation exceeding the limit mark is obtained.

4. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 3, characterized in that, Based on the voltage fluctuation exceeding limit marker in S202, when the marker is exceeding the limit, the longitude coordinate value, latitude coordinate value, altitude value, unique identifier code and upstream and downstream node identifier code list are extracted from the node spatiotemporal dataset and incorporated into the node spatial set. The electrical measurement data with time stamps are incorporated into the node time electrical set, thus obtaining the node spatial set and the node time electrical set.

5. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Set a strong correlation threshold, collect the percentage values ​​of CPU usage and memory usage of servers in each data center, calculate the difference between the maximum and minimum CPU usage percentage within a continuous 60 minutes as the computing power load fluctuation amplitude, extract the effective voltage value of the corresponding 60-minute period from the node time electrical collection, and obtain the computing power fluctuation amplitude sequence and voltage sequence. S302: Calculate the Pearson correlation coefficient between the computing power fluctuation amplitude sequence and the voltage sequence values, compare the absolute value of the correlation coefficient with the strong correlation threshold, and obtain the correlation exceeding limit marker and the correlation coefficient value.

6. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 5, characterized in that, In S302, the correlation exceeding limit marker and correlation coefficient value are used. When the marker is exceeding the limit, residual operation is performed on the computing power load fluctuation amplitude sequence and the voltage effective value numerical sequence. The voltage-related components are removed from the computing power load fluctuation amplitude sequence to obtain the orthogonalized computing power feature set. The computing power-related components are removed from the voltage effective value numerical sequence to obtain the orthogonalized power feature set. Thus, the orthogonalized computing power feature set and power feature set are obtained.

7. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 6, characterized in that, The specific steps of S4 are as follows: S401: Call the computing power feature set, extract the values ​​representing the urgency of computing power supply demand, sort them from largest to smallest, and obtain the computing power demand set to be scheduled; S402: Call the power feature set, extract the values ​​representing the availability of power, arrange them in descending order of value, and obtain the dispatchable power supply set.

8. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 7, characterized in that, The specific steps of S5 are as follows: S501: Based on the node spatial set, call the longitude and latitude coordinates of the first data center in the set of computing power demand to be scheduled, call the longitude and latitude coordinates of each power grid node in the set of schedulable power supply, calculate the spatial distance between the first data center and each power grid node, and obtain the candidate node spatial distance set. S502: Based on the spatial distance set of candidate nodes, query the transmission loss coefficient table for each distance value to obtain the corresponding loss coefficient, and subtract the loss coefficient from the power abundance value of the corresponding node in the schedulable power supply set to obtain the effective supply capacity set of candidate nodes.

9. The computational nonlocal mapping scheduling method based on spatiotemporal vector decoupling according to claim 8, characterized in that, The values ​​of the effective supply capacity set of candidate nodes in S502 are compared, and the unique identifier code of the power grid node corresponding to the maximum value is selected as the matching object to generate the matching power grid node identifier code.

10. A computational nonlocal mapping scheduling system based on spatiotemporal vector decoupling, characterized in that, The system is used to implement the nonlocal mapping scheduling method for computing based on spatiotemporal vector decoupling as described in any one of claims 1-9. The system includes: a node data acquisition module, a node decoupling and filtering module, a feature orthogonal decoupling module, a supply and demand sorting generation module, and a spatial mapping matching module. The node data acquisition module is used to perform real-time acquisition of electrical data, unique identification codes and location data of each power grid node, and at the same time read the acquisition time and summarize them to form a node spatiotemporal dataset. The node decoupling and filtering module is used to perform operations based on the node spatiotemporal dataset to obtain the long-term average voltage and daily average voltage, calculate the difference between the historical average voltage and the current daily average voltage, and set a judgment threshold. When the difference is greater than the judgment threshold, the location information and time voltage data in the node spatiotemporal dataset are split to obtain the node spatial set and the node time electrical set. The orthogonal decoupling module is used to set a strong correlation threshold, collect computing power fluctuation data of each data center server, calculate the fluctuation amplitude within one hour to obtain the computing power feature sequence, extract the voltage value sequence of the corresponding time period of the node time electrical concentration, calculate the Pearson correlation coefficient of the two sequences, and when the absolute value exceeds the strong correlation threshold, perform residual orthogonal processing on the computing power feature and the power feature to obtain the orthogonal computing power feature set and the power feature set. The supply and demand sorting generation module is used to sort the computing power feature set in descending order according to the urgency of computing power supply demand, generate a set of computing power demand to be scheduled, and sort the power feature set in descending order according to the power abundance, generate a set of schedulable power supply. The spatial mapping and matching module is used to perform calculations based on the node spatial set, calculating the spatial distance between the first data center of the scheduling computing power demand set and each power grid node in the scheduling power supply set, then calculating the effective supply value, and sending the computing power demand to the power grid node with the largest effective supply value.

Images