A power grid congestion dredging peak shaving method and device, electronic equipment and storage medium

Abstract

The application discloses a power grid congestion dredging peak regulation method and device, electronic equipment and storage medium, belongs to the technical field of power system operation control, the method comprises the following steps: obtaining the task queue state, network state parameter, original congestion power flow direction and dynamic regulation parameter of each data center of the power grid; based on the bandwidth and the calculation intensity characteristic generation time translation load and space transfer load, construct the time and space schedulable feasible region, and map to the effective virtual power transmission capacity; the decision model is established and solved with the maximum total net income as the target, and the cross-region computing power migration strategy is obtained; combined with the actual migration rate, the packet loss rate and the time delay jitter, the congestion factor and the response deviation are calculated, the upper limit of the bandwidth is dynamically corrected, the congestion dredging and the peak regulation control are realized. Through the implementation of the application, the problem that the control error accumulates due to the lack of closed-loop feedback mechanism in the cross-region computing power scheduling of the prior art, and the virtual power transmission capacity is difficult to respond accurately on the power grid side can be solved.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to a method, device, electronic equipment, and storage medium for grid congestion mitigation and peak shaving. Background Technology

[0002] With the rapid development of new power systems, the integration of high-proportion renewable energy sources and the dramatic fluctuations in load have exacerbated congestion and peak-shaving pressure on local power grid transmission lines. Therefore, implementing efficient grid congestion mitigation and peak-shaving control is crucial for ensuring the safe and stable operation of the power system. Data centers, as carriers of massive computing power, possess both high energy capacity and high load elasticity. By scheduling the spatial transfer of computing tasks between different geographical nodes, they can achieve coordination between computing load and power grid flow.

[0003] However, current grid congestion mitigation methods based on cross-regional computing power scheduling generally lack a closed-loop feedback mechanism, making it difficult to accurately translate computing load into reliable grid regulation capacity. The main reason is that existing computing power transfer strategies often lack dynamic correction mechanisms for the actual transmission process after execution. Due to the inevitable bandwidth fluctuations, packet loss, and latency jitter in wide area network communication links, the cross-regional migration tasks predicted by the system based on theoretical calculations are prone to rate deviations in the actual physical network. This operating mode, which relies solely on unidirectional open-loop command issuance and cannot adaptively correct based on the actual operating conditions and transmission rates of the underlying network, makes it difficult for the "virtual transmission capacity" based on computing power transfer mapping to achieve an accurate equivalent response on the grid side. Ultimately, this leads to the continuous accumulation of control errors in grid congestion mitigation, affecting overall peak-shaving accuracy. Summary of the Invention

[0004] This invention provides a grid congestion mitigation and peak shaving method, device, electronic device, and storage medium, which can solve the problems in the prior art where cross-regional computing power scheduling suffers from the lack of a closed-loop feedback mechanism, leading to the accumulation of control errors and the difficulty in achieving high-precision response of virtual transmission capacity on the grid side.

[0005] An embodiment of the present invention provides a method for relieving grid congestion and shaving peak loads, comprising: The system acquires the computing task queue status of each data center in the power grid, the network status parameters between the data centers, the original congestion flow direction of the target transmission line, and the dynamic control parameters of the power grid. The network status parameters include the real-time bandwidth of the wide area network, the predicted migration rate, the real-time packet loss rate, and the round-trip time jitter. From the computation task queue status, based on the real-time bandwidth of the wide area network and the preset task computation density characteristics, time-shifted loads and spatial transfer loads are generated; based on the time-shifted loads and spatial transfer loads, a spatiotemporal schedulable feasible domain matrix is ​​generated; based on the preset power transmission distribution factor and the original congested power flow direction, the spatial transfer loads are mapped to effective virtual transmission capacity. Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and power grid dynamic control parameters, a net revenue decision model is constructed with the goal of maximizing the total net revenue of data centers. The net revenue decision model is solved to generate an optimal computing power flow strategy that includes cross-regional migration power allocation instructions. According to the cross-regional migration power allocation instructions, cross-regional computing power migration is performed between data centers in the power grid, and the actual migration rate between data centers is monitored. The network congestion factor is calculated based on the real-time packet loss rate and round-trip time jitter; the virtual power transmission response deviation is calculated based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate; the bandwidth limit for controlling computing power migration between data centers is adjusted based on the virtual power transmission response deviation to achieve grid congestion mitigation and peak shaving control.

[0006] Furthermore, based on the real-time bandwidth of the wide area network and the preset task computation density characteristics, time-shifted load and spatial transfer load are generated from the computing task queue status, including: Analyze the status of the computing task queue and extract the maximum allowed latency period, total number of logical operations, and total amount of data transmission for each computing task; Computational tasks with a maximum allowed delay period exceeding a preset time threshold are grouped into a time-shifted task set; a time-shifted workload is generated based on the time-shifted task set. Calculate the ratio of the total logical operations to the total data transfers of the remaining computational tasks not included in the time-shifted task set; Based on the ratio and the preset server energy efficiency ratio, calculate the dynamic conversion coefficient of the remaining computing tasks; The computational tasks whose dynamic conversion coefficients are greater than a preset coefficient threshold and whose corresponding total data transmission volume satisfies the transmission constraints generated based on the real-time bandwidth of the wide area network are combined into a spatial transfer task set; a spatial transfer load is generated based on the spatial transfer task set.

[0007] Furthermore, based on the time-shifted load and the spatially transferred load, a spatiotemporally schedulable feasible region matrix is ​​generated, including: Obtain the instantaneous power limit, maximum operating power capacity, and rigid base load of each data center in the power grid; Calculate the energy requirements of each computation task in the time-shifted task set to generate the total energy requirements; Based on the instantaneous power limit, the total energy demand, and the maximum allowable delay period, construct a time-domain constraint vector set; Based on the real-time bandwidth of the wide area network, the real-time packet loss rate, and the dynamic conversion coefficient, construct a spatial domain constraint vector set; Based on the maximum operating power capacity and the rigid foundation load, construct a set of physical coupling constraint vectors; The time-domain constraint vector set, the spatial-domain constraint vector set, and the physical coupling constraint vector set are combined into a spatiotemporal schedulable feasible domain matrix.

[0008] Furthermore, the spatial load transfer characterizes the computing power migration between multiple sets of source data center nodes and target data center nodes, and the preset power transmission distribution factor includes the power transmission distribution factor of each data center node of the power grid to the target transmission line. Based on a preset power transmission distribution factor and the original congested power flow direction, spatially transferred loads are mapped to effective virtual transmission capacity, including: For each pair of source data center nodes and target data center nodes, extract the first power transmission distribution factor of the current source data center node to the target transmission line and the second power transmission distribution factor of the current target data center node to the target transmission line from the preset power transmission distribution factors. Based on the first power transmission distribution factor and the second power transmission distribution factor, determine the power flow change components generated on the target transmission line by the current computing power migration between the source data center node and the target data center node. After determining the power flow change components corresponding to each group of source data center nodes and target data center nodes, the total power flow change generated by the spatial transfer load on the target transmission line is generated based on the power flow change components corresponding to each group of source data center nodes and target data center nodes. Compare the direction of the total power flow change with the original blocked power flow direction; When the direction of the total power flow change is opposite to the direction of the original blocked power flow, the effective virtual transmission capacity is determined based on the total power flow change. When the direction of the total power flow change is the same as the direction of the original blocked power flow, the effective virtual transmission capacity is determined to be zero.

[0009] Furthermore, the power grid dynamic control parameters include the current power grid time-of-use tariff, peak-shaving compensation price, and virtual transmission incentive price; Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and dynamic grid control parameters, a net revenue decision model is constructed with the objective of maximizing the total net revenue of the data center, including: Based on the preset computing power service revenue rate, determine the total revenue of computing power services for data centers to perform computing tasks; The revenue from peak shaving ancillary services is determined based on the peak shaving compensation price and the time-shifted load. The virtual transmission diversion revenue is determined based on the virtual transmission incentive price and the effective virtual transmission capacity. The total operating cost of the system is determined based on the time-of-use electricity price of the power grid, the preset wide area network bandwidth rate, the time-shifted load, and the spatially transferred load. A net revenue objective function is generated based on the total revenue from computing power services, the revenue from peak shaving auxiliary services, the revenue from virtual power transmission diversion, and the total operating cost of the system. Using the spatiotemporal schedulable feasible region matrix as a constraint boundary, a net profit decision model is constructed with the objective function of maximizing net profit as the goal.

[0010] Furthermore, the net revenue decision model is solved to generate an optimal computing power flow strategy that includes cross-regional migration power allocation instructions, including: A preset optimization algorithm is used to solve the net income decision model, generating the optimal set of decision variables that maximize the net income objective function under the constraints. For each pair of source data center nodes and target data center nodes, the migration power allocation values ​​corresponding to the current source data center nodes and target data center nodes are extracted from the optimal decision variable set. After extracting the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes, a cross-region migration power allocation instruction for controlling cross-region computing power scheduling is generated based on the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes. The cross-regional migration power allocation command is combined with the remaining scheduling parameters in the optimal decision variable set to generate the optimal computing power flow strategy.

[0011] Furthermore, the network congestion factor is calculated based on the real-time packet loss rate and round-trip time jitter; the virtual transmission response deviation is calculated based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate; and the bandwidth limit for controlling computing power migration between data centers is corrected based on the virtual transmission response deviation to achieve grid congestion mitigation and peak shaving control, including: By combining the real-time packet loss rate and the round-trip time jitter, the network congestion factor characterizing the current link transmission quality is determined; Based on the network congestion factor, the difference between the actual migration rate and the predicted migration rate is quantified to generate a virtual power transmission response deviation; Based on the virtual power transmission response deviation, determine the bandwidth compensation adjustment amount for the computing power migration channel between each data center; The bandwidth compensation adjustment amount is superimposed on the current bandwidth limit between each data center to generate a corrected bandwidth limit, and the corrected bandwidth limit is updated in the spatiotemporal schedulable feasible domain matrix of the next control cycle.

[0012] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0013] One embodiment of the present invention provides a power grid congestion mitigation and peak shaving device, comprising: a parameter data acquisition module, a spatiotemporal evolution mapping module, an optimization decision execution module, and a closed-loop feedback correction module; The parameter data acquisition module is used to acquire the computing task queue status of each data center of the power grid, the network status parameters between each data center of the power grid, the original congestion flow direction of the target transmission line, and the dynamic control parameters of the power grid; wherein, the network status parameters include the real-time bandwidth of the wide area network, the predicted migration rate, the real-time packet loss rate, and the round-trip delay jitter. The spatiotemporal evolution mapping module is used to generate time-shifted loads and spatial transfer loads from the computing task queue state, based on the real-time bandwidth of the wide area network and the preset task computing density characteristics; generate a spatiotemporal schedulable feasible domain matrix based on the time-shifted loads and spatial transfer loads; and map the spatial transfer loads to effective virtual transmission capacity based on the preset power transmission distribution factor and the original congestion power flow direction. The optimization decision execution module is used to construct a net revenue decision model based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and power grid dynamic control parameters, with the goal of maximizing the total net revenue of the data center; solve the net revenue decision model to generate an optimal computing power flow strategy that includes cross-regional migration power allocation instructions; execute cross-regional computing power migration between data centers in the power grid according to the cross-regional migration power allocation instructions, and monitor the actual migration rate between data centers. The closed-loop feedback correction module is used to calculate the network congestion factor based on the real-time packet loss rate and round-trip delay jitter; calculate the virtual power transmission response deviation based on the network congestion factor, the actual migration rate between data centers and the predicted migration rate; and correct the bandwidth limit for controlling computing power migration between data centers based on the virtual power transmission response deviation, so as to realize grid congestion relief and peak shaving control.

[0014] Based on the above method embodiments, the present invention provides corresponding electronic device embodiments.

[0015] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements any of the power grid congestion mitigation and peak shaving methods described in the above-described method embodiments.

[0016] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments.

[0017] One embodiment of the present invention provides a storage medium storing a computer program thereon, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the above-described methods for grid congestion mitigation and peak shaving.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method, apparatus, electronic device, and storage medium for grid congestion mitigation and peak shaving. The method acquires the task queue status, network status parameters, original congestion flow direction, and dynamic control parameters of each data center in the power grid; generates time-shifted loads and spatially transferred loads based on bandwidth and computational density characteristics, constructs a spatiotemporally schedulable feasible region, and maps the spatially transferred loads to effective virtual transmission capacity; establishes and solves a decision model with the goal of maximizing the total net revenue of the data centers to obtain cross-regional computing power migration strategies; and calculates congestion factors and response deviations by combining actual migration rates, packet loss rates, and time delay jitter, dynamically correcting the bandwidth upper limit to achieve grid congestion mitigation and peak shaving optimization control.

[0019] After executing the optimal computing power transfer strategy, this invention monitors the actual migration rate between data centers and combines the network congestion factor calculated from real-time packet loss rate and round-trip delay jitter with the predicted migration rate to quantify the virtual power transmission response deviation. Based on this deviation, this invention dynamically adjusts the bandwidth limit used to control computing power migration between data centers, thereby constructing a closed-loop feedback control mechanism that adapts to the actual operating conditions of the underlying network. This effectively suppresses execution deviations caused by communication link fluctuations and ensures high-precision response and reliable adjustment of virtual power transmission capacity on the grid side. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a power grid congestion mitigation and peak shaving method according to an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of a power grid congestion mitigation and peak shaving device provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, to address the problems of accumulated control errors and difficulty in achieving high-precision response of virtual transmission capacity on the grid side due to the lack of a closed-loop feedback mechanism in existing cross-regional computing power scheduling, an embodiment of the present invention provides a grid congestion mitigation and peak shaving method, which includes at least the following steps: It should be noted that the method provided in this embodiment can be executed by a power grid congestion mitigation and peak shaving control device. Specifically, this device can be a server, control terminal, or cloud computing platform deployed in the power system dispatch center or computing network hub node, aiming to serve as a global decision-making center to coordinate the collaborative dispatch of computing power and electricity.

[0024] Step S1: Obtain the computing task queue status of each data center in the power grid, the network status parameters between each data center in the power grid, the original congestion flow direction of the target transmission line, and the dynamic control parameters of the power grid; wherein, the network status parameters include the real-time bandwidth of the wide area network, the predicted migration rate, the real-time packet loss rate, and the round-trip time jitter. Specifically, this involves acquiring the computing task queue status of each data center in the power grid, network status parameters between these data centers, the original congestion flow direction of the target transmission line, and dynamic control parameters of the power grid. Specifically, it involves collecting data on the backlog of unprocessed computing tasks and resource consumption requirements within data center nodes distributed across different geographical locations to obtain the computing task queue status. The computing task queue status encompasses the arrival time information of various business loads, the total logical operation requirements, and the total data transmission requirements. Comprehensive awareness of the computing task queue status allows for accurate understanding of the real-time computing power pressure on each data center node, thus providing underlying data support for subsequently differentiating loads with varying time sensitivity and spatial transfer value.

[0025] Obtaining network status parameters includes determining the WAN real-time bandwidth, predicted migration rate, real-time packet loss rate, and round-trip time jitter. WAN real-time bandwidth characterizes the available network transmission channel capacity for data scheduling between the source and target data center nodes. Real-time packet loss rate reflects the proportion of data packets lost during data packet transmission in the cross-regional network link. Round-trip time jitter quantifies the fluctuation range of the cross-regional network link in round-trip data transmission time. The predicted migration rate is determined by combining the aforementioned multi-dimensional parameters to accurately reflect the actual transmission capacity of the WAN link. The predicted migration rate is equal to the WAN real-time bandwidth multiplied by the first effective transmission factor, and divided by the jitter penalty factor. The first effective transmission factor is equal to the difference between a natural number and the real-time packet loss rate. The jitter penalty factor is equal to the sum of a natural number and the round-trip time jitter. The relationship for the predicted migration rate is expressed as: In the formula, Indicates the predicted migration rate. Indicates the real-time bandwidth of the wide area network. Indicates the real-time packet loss rate. This represents round-trip time jitter. Before performing specific algebraic calculations, both the real-time packet loss rate and round-trip time jitter have been pre-normalized and converted into dimensionless per-unit values ​​to ensure that the dimensions of the natural number 1 in the denominator are consistent with those of the round-trip time jitter during the summation, thereby ensuring the accuracy of the physical meaning of the predicted migration rate calculation results. Obtaining the above network state parameters aims to gain a deeper understanding of the actual physical communication conditions at the bottom layer of the wide area network, ensuring that cross-regional computing power scheduling commands do not fail due to congestion in the underlying communication links after issuance.

[0026] Identify the specific transmission lines in the current power grid that exceed their transmission capacity limits, and record the initial power flow direction vector that caused the transmission line to exceed its limits, in order to obtain the original congested power flow direction of the target transmission line. The original congested power flow direction is used in subsequent control processes to determine whether the changes in local power demand caused by the spatial transfer of computing power load can produce a positive mitigation effect in reducing the over-limit power flow.

[0027] The synchronously acquired power grid dynamic control parameters cover the current power grid time-of-use pricing, peak-shaving compensation pricing, and virtual transmission incentive pricing. Time-of-use pricing is introduced to accurately calculate the electricity costs of computing load. Peak-shaving compensation pricing is integrated to quantify the ancillary service revenue generated by computing power time-shifting behavior. The virtual transmission incentive pricing is combined to assess the economic value of mitigating transmission line congestion through computing power spatial transfer behavior.

[0028] By acquiring and integrating task status at the computing node level, network conditions at the communication link level, power flow parameters at the physical power grid level, and control indicators at the economic level, the deep fusion of underlying state information of the computing power network and the power network is achieved, thus constructing a solid data foundation for multi-dimensional constraint modeling and cross-regional collaborative scheduling of computing power.

[0029] Step S2: From the computing task queue status, generate time-shifted loads and spatial transfer loads based on the real-time bandwidth of the wide area network and the preset task computing density characteristics; generate a spatiotemporal schedulable feasible domain matrix based on the time-shifted loads and spatial transfer loads; map the spatial transfer loads to effective virtual transmission capacity based on the preset power transmission distribution factor and the original congestion power flow direction. In a preferred embodiment, time-shifted load and spatial transfer load are generated from the computing task queue state based on the real-time bandwidth of the wide area network and preset task computing density characteristics, including: Analyze the status of the computing task queue and extract the maximum allowed latency period, total number of logical operations, and total amount of data transmission for each computing task; Computational tasks with a maximum allowed delay period exceeding a preset time threshold are grouped into a time-shifted task set; a time-shifted workload is generated based on the time-shifted task set. Calculate the ratio of the total logical operations to the total data transfers of the remaining computational tasks not included in the time-shifted task set; Based on the ratio and the preset server energy efficiency ratio, calculate the dynamic conversion coefficient of the remaining computing tasks; The computational tasks whose dynamic conversion coefficients are greater than a preset coefficient threshold and whose corresponding total data transmission volume satisfies the transmission constraints generated based on the real-time bandwidth of the wide area network are combined into a spatial transfer task set; a spatial transfer load is generated based on the spatial transfer task set.

[0030] In a preferred embodiment, a spatiotemporally schedulable feasible region matrix is ​​generated based on time-shifted loads and spatially transferred loads, including: Obtain the instantaneous power limit, maximum operating power capacity, and rigid base load of each data center in the power grid; Calculate the energy requirements of each computation task in the time-shifted task set to generate the total energy requirements; Based on the instantaneous power limit, the total energy demand, and the maximum allowable delay period, construct a time-domain constraint vector set; Based on the real-time bandwidth of the wide area network, the real-time packet loss rate, and the dynamic conversion coefficient, construct a spatial domain constraint vector set; Based on the maximum operating power capacity and the rigid foundation load, construct a set of physical coupling constraint vectors; The time-domain constraint vector set, the spatial-domain constraint vector set, and the physical coupling constraint vector set are combined into a spatiotemporal schedulable feasible domain matrix.

[0031] In a preferred embodiment, the spatial transfer load represents the computing power migration between multiple sets of source data center nodes and target data center nodes, and the preset power transmission distribution factor includes the power transmission distribution factor of each data center node of the power grid to the target transmission line. Based on a preset power transmission distribution factor and the original congested power flow direction, spatially transferred loads are mapped to effective virtual transmission capacity, including: For each pair of source data center nodes and target data center nodes, extract the first power transmission distribution factor of the current source data center node to the target transmission line and the second power transmission distribution factor of the current target data center node to the target transmission line from the preset power transmission distribution factors. Based on the first power transmission distribution factor and the second power transmission distribution factor, determine the power flow change components generated on the target transmission line by the current computing power migration between the source data center node and the target data center node. After determining the power flow change components corresponding to each group of source data center nodes and target data center nodes, the total power flow change generated by the spatial transfer load on the target transmission line is generated based on the power flow change components corresponding to each group of source data center nodes and target data center nodes. Compare the direction of the total power flow change with the original blocked power flow direction; When the direction of the total power flow change is opposite to the direction of the original blocked power flow, the effective virtual transmission capacity is determined based on the total power flow change. When the direction of the total power flow change is the same as the direction of the original blocked power flow, the effective virtual transmission capacity is determined to be zero.

[0032] Specifically, based on the real-time bandwidth of the wide area network and the preset task computation density characteristics, time-shifted loads and spatially transferred loads are generated from the computation task queue status. Specifically, the computation task queue status is analyzed to comprehensively extract the maximum allowable delay period, total logical operations, and total data transmission volume for each computation task. Computation tasks with a maximum allowable delay period exceeding a preset time threshold are grouped into a time-shifted task set. The time threshold reflects the minimum standard of grid peak-shaving demand tolerance for task delays. Based on the time-shifted task set, time-shifted loads are generated.

[0033] For the remaining computational tasks not included in the time-shift task set, the feasibility of cross-region scheduling needs further evaluation. The ratio of the total logical operations to the total data transmission for the remaining computational tasks is calculated to obtain the task computational intensity characteristic. Based on the task computational intensity characteristic and a preset server energy efficiency ratio, a dynamic conversion coefficient for the remaining computational tasks is calculated. The value of the dynamic conversion coefficient is equal to the task computational intensity characteristic multiplied by the preset server energy efficiency ratio. The relationship of the dynamic conversion coefficient is expressed as: In the formula, Indicates the dynamic conversion coefficient. Represents the total amount of logical operations. Indicates the total amount of data transmitted. This represents the preset server energy efficiency ratio. Based on actual operating scenarios, the aforementioned preset time threshold, preset server energy efficiency ratio, and preset coefficient threshold are pre-calculated and calibrated by operations and maintenance personnel based on historical data center operation logs, statistical patterns of latency sensitivity of various business loads, and the factory technical specifications of the hardware equipment. These parameters are then persistently stored in the device's memory to ensure the objectivity and reliability of the underlying control parameters.

[0034] Computational tasks whose dynamic conversion coefficients exceed preset thresholds and whose corresponding total data transmission volume satisfies the transmission constraints generated based on the real-time bandwidth of the wide area network are precisely combined into a spatial transfer task set. Based on this spatial transfer task set, a spatial transfer load is generated.

[0035] Based on time-shifted load and spatially transferred load, a spatiotemporally schedulable feasible domain matrix is ​​generated. The instantaneous power ceiling, maximum operating power capacity, and rigid base load of each data center in the power grid are obtained. The energy requirement of each computing task in the time-shifted task set is calculated to generate the total energy requirement. A time-domain constraint vector set is constructed based on the instantaneous power ceiling, total energy requirement, and maximum allowable latency period. A spatial-domain constraint vector set is constructed based on the WAN real-time bandwidth, real-time packet loss rate, and dynamic conversion coefficients. A physical coupling constraint vector set is constructed based on the maximum operating power capacity and rigid base load. The time-domain constraint vector set, spatial-domain constraint vector set, and physical coupling constraint vector set are combined into the spatiotemporally schedulable feasible domain matrix. This matrix integrates the comprehensive constraint boundaries of computing load in both the time latency dimension and the spatial cross-regional transfer dimension. The array of elements inside the spatiotemporal schedulable feasible domain matrix corresponds to the available adjustment margin and scheduling state boundary values ​​of different geographic data center nodes in different control periods. The structured alignment of discrete power constraints and computing power constraints is achieved through the row and column dimensions of the matrix, thereby transforming multidimensional heterogeneous constraints into a standard mathematical matrix form that is suitable for optimization algorithms to directly read and iteratively calculate.

[0036] Spatial load transfer characterizes the migration of computing power between multiple sets of source data center nodes and target data center nodes. The preset power transfer distribution factor includes the power transfer distribution factor of each data center node in the power grid to the target transmission line. Based on the preset power transfer distribution factor and the original congested power flow direction, the spatial load transfer is mapped to effective virtual transmission capacity. For each set of source and target data center nodes, the first power transfer distribution factor of the current source data center node to the target transmission line and the second power transfer distribution factor of the current target data center node to the target transmission line are extracted from the preset power transfer distribution factors.

[0037] Based on the first and second power transfer distribution factors, the power flow change components generated on the target transmission line by the current power migration between the source and target data center nodes are determined. The value of each power flow change component is equal to the migration power between the source and target data center nodes multiplied by the difference between the first and second power transfer distribution factors. The relationship between the power flow change components is expressed as: In the formula, Indicates the component of tidal change. This indicates the current migration power between the source data center node and the target data center node. Represents the first power transfer distribution factor. This represents the second power transmission distribution factor.

[0038] After determining the power flow change components corresponding to each group of source data center nodes and target data center nodes, the total power flow change generated by the spatially transferred load on the target transmission line is generated based on these components. The determination of the total power flow change involves globally aggregating and accumulating the power flow change components corresponding to all groups of source data center nodes and target data center nodes.

[0039] The direction of the total power flow change is compared with the original congested power flow direction. When the direction of the total power flow change is opposite to the original congested power flow direction, it is determined that the underlying physical power flow change caused by the computing power migration has a positive mitigating effect on the congested line, and the effective virtual transmission capacity is determined based on the total power flow change. When the direction of the total power flow change is the same as the original congested power flow direction, it is determined that the power flow change caused by the computing power migration has exacerbated the burden on the congested line, directly triggering the defense mechanism and setting the effective virtual transmission capacity to zero.

[0040] By executing the above process, not only is the fine-grained classification and constraint modeling of massive computing loads in the time and space dimensions realized, but the cross-regional computing power flow behavior is also rigorously transformed into virtual transmission capacity that can be directly mapped to the physical power grid level. This effectively eliminates ineffective or even deteriorating scheduling actions that affect the power grid's operating status, laying a reliable physical equivalent foundation for the precise routing of transmission lines at the bottom of the power grid.

[0041] Step S3: Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and power grid dynamic control parameters, construct a net revenue decision model with the goal of maximizing the total net revenue of the data center; solve the net revenue decision model to generate the optimal computing power flow strategy that includes cross-regional migration power allocation instructions; execute cross-regional computing power migration between data centers in the power grid according to the cross-regional migration power allocation instructions, and monitor the actual migration rate between data centers. In a preferred embodiment, the power grid dynamic control parameters include the current power grid time-of-use tariff, peak-shaving compensation price, and virtual transmission incentive price; Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and dynamic grid control parameters, a net revenue decision model is constructed with the objective of maximizing the total net revenue of the data center, including: Based on the preset computing power service revenue rate, determine the total revenue of computing power services for data centers to perform computing tasks; The revenue from peak shaving ancillary services is determined based on the peak shaving compensation price and the time-shifted load. The virtual transmission diversion revenue is determined based on the virtual transmission incentive price and the effective virtual transmission capacity. The total operating cost of the system is determined based on the time-of-use electricity price of the power grid, the preset wide area network bandwidth rate, the time-shifted load, and the spatially transferred load. A net revenue objective function is generated based on the total revenue from computing power services, the revenue from peak shaving auxiliary services, the revenue from virtual power transmission diversion, and the total operating cost of the system. Using the spatiotemporal schedulable feasible region matrix as a constraint boundary, a net profit decision model is constructed with the objective function of maximizing net profit as the goal.

[0042] In a preferred embodiment, the net revenue decision model is solved to generate an optimal computing power transfer strategy that includes cross-regional migration power allocation instructions, including: A preset optimization algorithm is used to solve the net income decision model, generating the optimal set of decision variables that maximize the net income objective function under the constraints. For each pair of source data center nodes and target data center nodes, the migration power allocation values ​​corresponding to the current source data center nodes and target data center nodes are extracted from the optimal decision variable set. After extracting the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes, a cross-region migration power allocation instruction for controlling cross-region computing power scheduling is generated based on the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes. The cross-regional migration power allocation command is combined with the remaining scheduling parameters in the optimal decision variable set to generate the optimal computing power flow strategy.

[0043] Specifically, based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and grid dynamic control parameters, a net revenue decision model is constructed with the objective of maximizing the total net revenue of the data center. Specifically, the grid dynamic control parameters encompass the current grid time-of-use pricing, peak-shaving compensation price, and virtual transmission incentive price. These aforementioned economic indicators constitute the underlying numerical benchmark for calculating the overall multidimensional revenue and operating costs.

[0044] The process of constructing a net revenue decision model requires detailed quantification of various revenue and expenditure dimensions. Combined with a pre-defined computing power service rate of return, the total computing power service revenue for the data center executing computing tasks is determined. Total computing power service revenue represents the basic service return generated by completing information processing tasks. The value of total computing power service revenue equals the pre-defined computing power service rate of return multiplied by the total load of computing tasks executed by the data center. The relationship for total computing power service revenue is expressed as: In the formula, This represents the total revenue from computing power services. This indicates the preset rate of return on computing power services. This indicates the total workload of the data center performing computing tasks.

[0045] The revenue from peak-shaving ancillary services is determined based on the peak-shaving compensation price and the time-shifted load. Peak-shaving ancillary service revenue quantifies the economic return obtained by proactively delaying non-time-sensitive tasks in response to the power grid's call for peak shaving and valley filling. The value of peak-shaving ancillary service revenue equals the peak-shaving compensation price multiplied by the time-shifted load. The relationship between peak-shaving ancillary service revenue and revenue is expressed as: In the formula, This indicates revenue from peak-shaving ancillary services. This indicates the peak-shaving compensation price. This indicates a time-shifted load.

[0046] The virtual transmission diversion revenue is determined based on the virtual transmission incentive price and the effective virtual transmission capacity. This revenue reflects the special funding incentive received for alleviating localized line congestion in the power network by converting cross-regional computing power dispatch into equivalent physical transmission capacity. The virtual transmission diversion revenue is equal to the virtual transmission incentive price multiplied by the effective virtual transmission capacity. The formula for virtual transmission diversion revenue is expressed as: In the formula, This indicates the revenue from virtual power transmission diversion. Indicates the virtual transmission incentive price. This represents the effective virtual transmission capacity.

[0047] The total system operating cost is determined based on the grid time-of-use tariff, the preset wide area network (WAN) bandwidth rate, time-shifted load, and spatially transferred load. The total system operating cost includes the energy consumption for maintaining daily equipment operation and the network communication costs generated by cross-regional data transfer. The total system operating cost is equal to the grid time-of-use tariff multiplied by the sum of the time-shifted load and the spatially transferred load, plus the product of the preset WAN bandwidth rate and the spatially transferred load. The formula for the total system operating cost is expressed as: In the formula, This represents the total operating cost of the system. This indicates the current time-of-use electricity price on the power grid. This indicates the preset WAN bandwidth rate. This indicates a spatial transfer of load.

[0048] A net revenue objective function is generated based on the total revenue from computing power services, peak shaving ancillary services, virtual power transmission diversion, and the total system operating cost. The net revenue objective function is the difference between the sum of the three assets (total revenue from computing power services, peak shaving ancillary services, and virtual power transmission diversion) and the total system operating cost. The relationship of the net revenue objective function is expressed as follows: In the formula, This represents the objective function for net income.

[0049] Using the spatiotemporal schedulable feasible region matrix as the constraint boundary, and the physical security and communication network capacity limits defined by the aforementioned spatiotemporal schedulable feasible region matrix as the mandatory constraint conditions for model solving, a net profit decision model is constructed with the objective function of maximizing net profit as the goal.

[0050] Solving the net profit decision model generates an optimal computing power flow strategy that includes cross-regional migration power allocation instructions. A preset optimization algorithm is used to iteratively solve the net profit decision model, generating a set of optimal decision variables that maximizes the net profit objective function while satisfying the aforementioned constraints. Specifically, the preset optimization algorithm includes business mathematical programming solver algorithms based on interior point methods and branch-and-bound methods, or heuristic intelligent optimization algorithms such as particle swarm optimization and genetic algorithms. By selecting an appropriate algorithm, the multidimensional variables and complex constraints in the net profit decision model can be effectively handled, ensuring the solution of a globally or locally optimal control strategy. The optimal decision variable set contains the complete set of optimal control parameter settings corresponding to achieving the extreme value of economic benefits.

[0051] For each pair of source and target data center nodes, the migration power allocation values ​​corresponding to the current source and target data center nodes are accurately extracted from the optimal decision variable set. After extracting the migration power allocation values ​​for each pair of source and target data center nodes, cross-region migration power allocation instructions for controlling cross-region computing power scheduling are generated based on these values. The cross-region migration power allocation instructions are then integrated with the remaining scheduling parameters in the optimal decision variable set to generate the optimal computing power flow strategy. This optimal computing power flow strategy not only includes spatial power allocation schemes but also incorporates time-dimensional task start / stop plans.

[0052] Based on the cross-regional migration power allocation instructions, cross-regional computing power migration operations are actually executed between various data centers in the power grid, and the actual migration rate between data centers is continuously monitored. The actual migration rate reflects the real throughput performance of the underlying wide area network links when carrying cross-regional computing power scheduling data flows.

[0053] By executing the complete modeling, solving, and command issuance process described above, the abstract grid congestion relief needs and complex computing power scheduling tasks are transformed into a mathematical optimization problem that pursues the extreme value of economic benefits. Under strict physical boundary constraints, the globally optimal planning configuration is obtained, realizing a two-way deep synergy between improving the net revenue of autonomous data center operation and meeting the peak shaving and relief needs of the underlying power system.

[0054] Step S4: Calculate the network congestion factor based on the real-time packet loss rate and round-trip time jitter; calculate the virtual power transmission response deviation based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate; adjust the bandwidth limit for controlling computing power migration between data centers based on the virtual power transmission response deviation, so as to achieve grid congestion relief and peak shaving control.

[0055] In a preferred embodiment, a network congestion factor is calculated based on the real-time packet loss rate and round-trip time jitter; a virtual transmission response deviation is calculated based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate; and the bandwidth limit for controlling computing power migration between data centers is corrected based on the virtual transmission response deviation to achieve grid congestion mitigation and peak shaving control, including: By combining the real-time packet loss rate and the round-trip time jitter, the network congestion factor characterizing the current link transmission quality is determined; Based on the network congestion factor, the difference between the actual migration rate and the predicted migration rate is quantified to generate a virtual power transmission response deviation; Based on the virtual power transmission response deviation, determine the bandwidth compensation adjustment amount for the computing power migration channel between each data center; The bandwidth compensation adjustment amount is superimposed on the current bandwidth limit between each data center to generate a corrected bandwidth limit, and the corrected bandwidth limit is updated in the spatiotemporal schedulable feasible domain matrix of the next control cycle.

[0056] Specifically, the network congestion factor is calculated based on the real-time packet loss rate and round-trip delay jitter. The network congestion factor is used to quantitatively assess the physical congestion level of the underlying WAN communication links when carrying cross-regional computing power scheduling data streams. By fusing the real-time packet loss rate and round-trip delay jitter, the network congestion factor, characterizing the current link transmission quality, is determined. The specific calculation logic is to multiply the real-time packet loss rate by a preset packet loss penalty weight, and multiply the round-trip delay jitter by a preset delay penalty weight, then sum the two products to obtain the network congestion factor. The relationship of the network congestion factor is expressed as: In the formula, Indicates the network congestion factor. This indicates the preset packet loss penalty weight. This indicates the preset latency penalty weight.

[0057] Based on the network congestion factor, the difference between the actual migration rate and the predicted migration rate between data centers is quantified to generate a virtual power transmission response deviation. The virtual power transmission response deviation reflects the latency in computing power flow caused by link conditions when theoretically allocated commands are executed in the underlying physical network. The determination process involves subtracting the actual migration rate from the predicted migration rate to obtain the rate difference. Subsequently, the network congestion factor is used to weight and amplify the rate difference to generate the virtual power transmission response deviation. The relationship of the virtual power transmission response deviation is expressed as: In the formula, Indicates the virtual transmission response deviation. This indicates the actual migration rate.

[0058] Based on the virtual power transmission response deviation, the bandwidth compensation adjustment amount for the computing power migration channels between data centers is determined. The bandwidth compensation adjustment amount represents the reduction in channel limits that need to be imposed at the top-level control layer to eliminate physical transmission response lag. The calculation steps involve multiplying the virtual power transmission response deviation by a preset feedback adjustment gain and taking the negative number to determine the corresponding bandwidth compensation adjustment amount. Taking the negative number aims to proactively reduce the subsequent bandwidth upper limit setting boundary when a positive response deviation exists in the network. The relationship for the bandwidth compensation adjustment amount is expressed as: In the formula, This indicates the bandwidth compensation adjustment amount. This indicates the preset feedback adjustment gain.

[0059] The bandwidth compensation adjustment is added to the current bandwidth limit between each data center to generate the corrected bandwidth limit. The current bandwidth limit indicates the maximum throughput physical threshold allowed for cross-regional data transmission between data centers within this adjustment period. This addition process makes the corrected bandwidth limit more closely reflect the actual physical link capacity. The formula for generating the corrected bandwidth limit is expressed as: In the formula, This indicates the revised bandwidth limit. This indicates the current bandwidth limit.

[0060] The revised bandwidth limit is updated in the spatiotemporal schedulable feasible domain matrix for the next control cycle. The updated communication network capacity limit is used as a mandatory constraint for solving the decision model in the next control cycle, ensuring that the cross-regional migration power allocation instructions generated in the next round are fully adapted to the dynamically changing underlying network environment.

[0061] By executing the above-mentioned closed-loop monitoring and parameter correction process based on the actual transmission status of the underlying link, a dynamic feedback correction mechanism adapted to the fluctuations of the physical operating conditions of the wide area network was constructed. This effectively eliminated the execution error of computing power flow caused by communication link congestion, ensured the high-precision response and reliable adjustment of virtual power transmission capacity on the grid side, and ultimately achieved highly robust grid congestion mitigation and peak-shaving control.

[0062] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0063] like Figure 2 As shown, an embodiment of the present invention provides a power grid congestion mitigation and peak shaving device, including: a parameter data acquisition module, a spatiotemporal evolution mapping module, an optimization decision execution module, and a closed-loop feedback correction module; The parameter data acquisition module is used to acquire the computing task queue status of each data center of the power grid, the network status parameters between each data center of the power grid, the original congestion flow direction of the target transmission line, and the dynamic control parameters of the power grid; wherein, the network status parameters include the real-time bandwidth of the wide area network, the predicted migration rate, the real-time packet loss rate, and the round-trip delay jitter. The spatiotemporal evolution mapping module is used to generate time-shifted loads and spatial transfer loads from the computing task queue state, based on the real-time bandwidth of the wide area network and the preset task computing density characteristics; generate a spatiotemporal schedulable feasible domain matrix based on the time-shifted loads and spatial transfer loads; and map the spatial transfer loads to effective virtual transmission capacity based on the preset power transmission distribution factor and the original congestion power flow direction. The optimization decision execution module is used to construct a net revenue decision model based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and power grid dynamic control parameters, with the goal of maximizing the total net revenue of the data center; solve the net revenue decision model to generate an optimal computing power flow strategy that includes cross-regional migration power allocation instructions; execute cross-regional computing power migration between data centers in the power grid according to the cross-regional migration power allocation instructions, and monitor the actual migration rate between data centers. The closed-loop feedback correction module is used to calculate the network congestion factor based on the real-time packet loss rate and round-trip delay jitter; calculate the virtual power transmission response deviation based on the network congestion factor, the actual migration rate between data centers and the predicted migration rate; and correct the bandwidth limit for controlling computing power migration between data centers based on the virtual power transmission response deviation, so as to realize grid congestion relief and peak shaving control.

[0064] In a preferred embodiment, the spatiotemporal evolution mapping module generates time-shifted loads and spatial transfer loads from the computing task queue state, based on the wide area network real-time bandwidth and preset task computing density characteristics, including: Analyze the status of the computing task queue and extract the maximum allowed latency period, total number of logical operations, and total amount of data transmission for each computing task; Computational tasks with a maximum allowed delay period exceeding a preset time threshold are grouped into a time-shifted task set; a time-shifted workload is generated based on the time-shifted task set. Calculate the ratio of the total logical operations to the total data transfers of the remaining computational tasks not included in the time-shifted task set; Based on the ratio and the preset server energy efficiency ratio, calculate the dynamic conversion coefficient of the remaining computing tasks; The computational tasks whose dynamic conversion coefficients are greater than a preset coefficient threshold and whose corresponding total data transmission volume satisfies the transmission constraints generated based on the real-time bandwidth of the wide area network are combined into a spatial transfer task set; a spatial transfer load is generated based on the spatial transfer task set.

[0065] In a preferred embodiment, the spatiotemporal evolution mapping module generates a spatiotemporally schedulable feasible region matrix based on time translation load and spatial transfer load, including: Obtain the instantaneous power limit, maximum operating power capacity, and rigid base load of each data center in the power grid; Calculate the energy requirements of each computation task in the time-shifted task set to generate the total energy requirements; Based on the instantaneous power limit, the total energy demand, and the maximum allowable delay period, construct a time-domain constraint vector set; Based on the real-time bandwidth of the wide area network, the real-time packet loss rate, and the dynamic conversion coefficient, construct a spatial domain constraint vector set; Based on the maximum operating power capacity and the rigid foundation load, construct a set of physical coupling constraint vectors; The time-domain constraint vector set, the spatial-domain constraint vector set, and the physical coupling constraint vector set are combined into a spatiotemporal schedulable feasible domain matrix.

[0066] In a preferred embodiment, the spatiotemporal evolution mapping module, wherein the spatial transfer load characterizes the computing power migration between multiple sets of source data center nodes and target data center nodes, and the preset power transmission distribution factor includes the power transmission distribution factor of each data center node in the power grid to the target transmission line; combined with the actual operation scenario, the preset power transmission distribution factor is obtained in advance offline based on the topology parameters and line impedance parameters of the power system transmission network, and is persistently stored in the memory of the terminal device to objectively reflect the physical influence weight of the power injected by each data center node on the power flow of the target transmission line.

[0067] Based on a preset power transmission distribution factor and the original congested power flow direction, spatially transferred loads are mapped to effective virtual transmission capacity, including: For each pair of source data center nodes and target data center nodes, extract the first power transmission distribution factor of the current source data center node to the target transmission line and the second power transmission distribution factor of the current target data center node to the target transmission line from the preset power transmission distribution factors. Based on the first power transmission distribution factor and the second power transmission distribution factor, determine the power flow change components generated on the target transmission line by the current computing power migration between the source data center node and the target data center node. After determining the power flow change components corresponding to each group of source data center nodes and target data center nodes, the total power flow change generated by the spatial transfer load on the target transmission line is generated based on the power flow change components corresponding to each group of source data center nodes and target data center nodes. Compare the direction of the total power flow change with the original blocked power flow direction; When the direction of the total power flow change is opposite to the direction of the original blocked power flow, the effective virtual transmission capacity is determined based on the total power flow change. When the direction of the total power flow change is the same as the direction of the original blocked power flow, the effective virtual transmission capacity is determined to be zero.

[0068] In a preferred embodiment, the optimization decision execution module includes the current grid time-of-use electricity price, peak-shaving compensation price, and virtual transmission incentive price as the grid dynamic control parameters. Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and dynamic grid control parameters, a net revenue decision model is constructed with the objective of maximizing the total net revenue of the data center, including: Based on the preset computing power service revenue rate, determine the total revenue of computing power services for data centers to perform computing tasks; The revenue from peak shaving ancillary services is determined based on the peak shaving compensation price and the time-shifted load. The virtual transmission diversion revenue is determined based on the virtual transmission incentive price and the effective virtual transmission capacity. The total operating cost of the system is determined based on the time-of-use electricity price of the power grid, the preset wide area network bandwidth rate, the time-shifted load, and the spatially transferred load. A net revenue objective function is generated based on the total revenue from computing power services, the revenue from peak shaving auxiliary services, the revenue from virtual power transmission diversion, and the total operating cost of the system. Using the spatiotemporal schedulable feasible region matrix as a constraint boundary, a net profit decision model is constructed with the objective function of maximizing net profit as the goal.

[0069] In a preferred embodiment, the optimization decision execution module solves the net profit decision model and generates an optimal computing power transfer strategy that includes cross-regional migration power allocation instructions, including: A preset optimization algorithm is used to solve the net income decision model, generating the optimal set of decision variables that maximize the net income objective function under the constraints. For each pair of source data center nodes and target data center nodes, the migration power allocation values ​​corresponding to the current source data center nodes and target data center nodes are extracted from the optimal decision variable set. After extracting the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes, a cross-region migration power allocation instruction for controlling cross-region computing power scheduling is generated based on the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes. The cross-regional migration power allocation command is combined with the remaining scheduling parameters in the optimal decision variable set to generate the optimal computing power flow strategy.

[0070] In a preferred embodiment, the closed-loop feedback correction module calculates the network congestion factor based on the real-time packet loss rate and round-trip time jitter; calculates the virtual transmission response deviation based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate; and corrects the bandwidth limit used to control computing power migration between data centers based on the virtual transmission response deviation, so as to achieve grid congestion mitigation and peak shaving control, including: By combining the real-time packet loss rate and the round-trip time jitter, the network congestion factor characterizing the current link transmission quality is determined; Based on the network congestion factor, the difference between the actual migration rate and the predicted migration rate is quantified to generate a virtual power transmission response deviation; Based on the virtual power transmission response deviation, determine the bandwidth compensation adjustment amount for the computing power migration channel between each data center; The bandwidth compensation adjustment amount is superimposed on the current bandwidth limit between each data center to generate a corrected bandwidth limit, and the corrected bandwidth limit is updated in the spatiotemporal schedulable feasible domain matrix of the next control cycle.

[0071] It should be noted that the embodiments of the device described above correspond to the embodiments of the present invention described above, and can realize the power grid congestion mitigation and peak shaving method described in any one of the above embodiments of the present invention. Furthermore, the embodiments of the device described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without creative effort.

[0072] Based on the above-described method embodiments of the present invention, a corresponding embodiment of an electronic device is provided.

[0073] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power grid congestion mitigation and peak shaving method according to any one of the present invention, or, when the processor executes the computer program, it implements the functions of each module in the above-described device embodiments.

[0074] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0075] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0076] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0077] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0078] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments; Another embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the above-described power grid congestion mitigation and peak shaving methods of the present invention.

[0079] The aforementioned storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0080] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0081] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for relieving grid congestion and shaving peak loads, characterized in that, include: The system acquires the computing task queue status of each data center in the power grid, the network status parameters between the data centers, the original congestion flow direction of the target transmission line, and the dynamic control parameters of the power grid. The network status parameters include the real-time bandwidth of the wide area network, the predicted migration rate, the real-time packet loss rate, and the round-trip time jitter. From the computation task queue status, based on the real-time bandwidth of the wide area network and the preset task computation density characteristics, time-shifted loads and spatial transfer loads are generated; based on the time-shifted loads and spatial transfer loads, a spatiotemporal schedulable feasible domain matrix is ​​generated; based on the preset power transmission distribution factor and the original congested power flow direction, the spatial transfer loads are mapped to effective virtual transmission capacity. Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and power grid dynamic control parameters, a net revenue decision model is constructed with the goal of maximizing the total net revenue of data centers. The net revenue decision model is solved to generate an optimal computing power flow strategy that includes cross-regional migration power allocation instructions. According to the cross-regional migration power allocation instructions, cross-regional computing power migration is performed between data centers in the power grid, and the actual migration rate between data centers is monitored. The network congestion factor is calculated based on the real-time packet loss rate and round-trip time jitter; the virtual power transmission response deviation is calculated based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate; the bandwidth limit for controlling computing power migration between data centers is adjusted based on the virtual power transmission response deviation to achieve grid congestion mitigation and peak shaving control.

2. The power grid congestion mitigation and peak shaving method as described in claim 1, characterized in that, Based on the WAN real-time bandwidth and preset task computation density characteristics, time-shifted load and spatial transfer load are generated from the computation task queue status, including: Analyze the status of the computing task queue and extract the maximum allowed latency period, total number of logical operations, and total amount of data transmission for each computing task; Computational tasks with a maximum allowed delay period exceeding a preset time threshold are grouped into a time-shifted task set; a time-shifted workload is generated based on the time-shifted task set. Calculate the ratio of the total logical operations to the total data transmission of the remaining computational tasks not included in the time-shifted task set; Based on the ratio and the preset server energy efficiency ratio, calculate the dynamic conversion coefficient of the remaining computing tasks; The computational tasks whose dynamic conversion coefficients are greater than a preset coefficient threshold and whose corresponding total data transmission volume satisfies the transmission constraints generated based on the real-time bandwidth of the wide area network are combined into a spatial transfer task set; a spatial transfer load is generated based on the spatial transfer task set.

3. The power grid congestion mitigation and peak shaving method as described in claim 2, characterized in that, Based on time-shifted loads and spatially transferred loads, a spatiotemporally schedulable feasible region matrix is ​​generated, including: Obtain the instantaneous power limit, maximum operating power capacity, and rigid base load of each data center in the power grid; Calculate the energy requirements of each computation task in the time-shifted task set to generate the total energy requirements; Based on the instantaneous power limit, the total energy demand, and the maximum allowable delay period, construct a time-domain constraint vector set; Based on the real-time bandwidth of the wide area network, the real-time packet loss rate, and the dynamic conversion coefficient, construct a spatial domain constraint vector set; Based on the maximum operating power capacity and the rigid foundation load, construct a set of physical coupling constraint vectors; The time-domain constraint vector set, the spatial-domain constraint vector set, and the physical coupling constraint vector set are combined into a spatiotemporal schedulable feasible domain matrix.

4. The power grid congestion mitigation and peak shaving method as described in claim 3, characterized in that, The spatial load transfer characterizes the computing power migration between multiple source data center nodes and target data center nodes, and the preset power transmission distribution factor includes the power transmission distribution factor of each data center node in the power grid to the target transmission line. Based on a preset power transmission distribution factor and the original congested power flow direction, spatially transferred loads are mapped to effective virtual transmission capacity, including: For each pair of source data center nodes and target data center nodes, extract the first power transmission distribution factor of the current source data center node to the target transmission line and the second power transmission distribution factor of the current target data center node to the target transmission line from the preset power transmission distribution factors. Based on the first power transmission distribution factor and the second power transmission distribution factor, determine the power flow change components generated on the target transmission line by the current computing power migration between the source data center node and the target data center node. After determining the power flow change components corresponding to each group of source data center nodes and target data center nodes, the total power flow change generated by the spatial transfer load on the target transmission line is generated based on the power flow change components corresponding to each group of source data center nodes and target data center nodes. Compare the direction of the total power flow change with the original blocked power flow direction; When the direction of the total power flow change is opposite to the direction of the original blocked power flow, the effective virtual transmission capacity is determined based on the total power flow change. When the direction of the total power flow change is the same as the direction of the original blocked power flow, the effective virtual transmission capacity is determined to be zero.

5. The power grid congestion mitigation and peak shaving method as described in claim 4, characterized in that, The power grid dynamic control parameters include the current power grid time-of-use tariff, peak-shaving compensation price, and virtual transmission incentive price; Based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and dynamic grid control parameters, a net revenue decision model is constructed with the objective of maximizing the total net revenue of the data center, including: Based on the preset computing power service revenue rate, determine the total revenue of computing power services for data centers to perform computing tasks; The revenue from peak shaving ancillary services is determined based on the peak shaving compensation price and the time-shifted load. The virtual transmission diversion revenue is determined based on the virtual transmission incentive price and the effective virtual transmission capacity. The total operating cost of the system is determined based on the time-of-use electricity price of the power grid, the preset wide area network bandwidth rate, the time-shifted load, and the spatially transferred load. A net revenue objective function is generated based on the total revenue from computing power services, the revenue from peak shaving auxiliary services, the revenue from virtual power transmission diversion, and the total operating cost of the system. Using the spatiotemporal schedulable feasible region matrix as a constraint boundary, a net profit decision model is constructed with the objective function of maximizing net profit as the goal.

6. The power grid congestion mitigation and peak shaving method as described in claim 5, characterized in that, Solve the net revenue decision model to generate the optimal computing power transfer strategy that includes cross-regional migration power allocation instructions, including: A preset optimization algorithm is used to solve the net income decision model, generating the optimal set of decision variables that maximize the net income objective function under the constraints. For each pair of source data center nodes and target data center nodes, the migration power allocation values ​​corresponding to the current source data center nodes and target data center nodes are extracted from the optimal decision variable set. After extracting the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes, a cross-region migration power allocation instruction for controlling cross-region computing power scheduling is generated based on the migration power allocation values ​​corresponding to each group of source data center nodes and target data center nodes. The cross-regional migration power allocation command is combined with the remaining scheduling parameters in the optimal decision variable set to generate the optimal computing power flow strategy.

7. The power grid congestion mitigation and peak shaving method as described in claim 6, characterized in that, The network congestion factor is calculated based on the real-time packet loss rate and round-trip time jitter; the virtual power transmission response deviation is calculated based on the network congestion factor, the actual migration rate between data centers, and the predicted migration rate. Based on the virtual power transmission response deviation, the bandwidth limit used to control computing power migration between data centers is adjusted to achieve grid congestion mitigation and peak shaving control, including: By combining the real-time packet loss rate and the round-trip time jitter, the network congestion factor characterizing the current link transmission quality is determined; Based on the network congestion factor, the difference between the actual migration rate and the predicted migration rate is quantified to generate a virtual power transmission response deviation; Based on the virtual power transmission response deviation, determine the bandwidth compensation adjustment amount for the computing power migration channel between each data center; The bandwidth compensation adjustment amount is superimposed on the current bandwidth limit between each data center to generate a corrected bandwidth limit, and the corrected bandwidth limit is updated in the spatiotemporal schedulable feasible domain matrix of the next control cycle.

8. A power grid congestion mitigation and peak-shaving device, characterized in that, include: The module includes a parameter data acquisition module, a spatiotemporal evolution mapping module, an optimization decision execution module, and a closed-loop feedback correction module. The parameter data acquisition module is used to acquire the computing task queue status of each data center of the power grid, the network status parameters between each data center of the power grid, the original congestion flow direction of the target transmission line, and the dynamic control parameters of the power grid; wherein, the network status parameters include the real-time bandwidth of the wide area network, the predicted migration rate, the real-time packet loss rate, and the round-trip delay jitter. The spatiotemporal evolution mapping module is used to generate time-shifted loads and spatial transfer loads from the computing task queue state, based on the real-time bandwidth of the wide area network and the preset task computing density characteristics; generate a spatiotemporal schedulable feasible domain matrix based on the time-shifted loads and spatial transfer loads; and map the spatial transfer loads to effective virtual transmission capacity based on the preset power transmission distribution factor and the original congestion power flow direction. The optimization decision execution module is used to construct a net revenue decision model based on the spatiotemporal schedulable feasible domain matrix, effective virtual transmission capacity, and power grid dynamic control parameters, with the goal of maximizing the total net revenue of the data center; solve the net revenue decision model to generate an optimal computing power flow strategy that includes cross-regional migration power allocation instructions; execute cross-regional computing power migration between data centers in the power grid according to the cross-regional migration power allocation instructions, and monitor the actual migration rate between data centers. The closed-loop feedback correction module is used to calculate the network congestion factor based on the real-time packet loss rate and round-trip delay jitter; calculate the virtual power transmission response deviation based on the network congestion factor, the actual migration rate between data centers and the predicted migration rate; and correct the bandwidth limit for controlling computing power migration between data centers based on the virtual power transmission response deviation, so as to realize grid congestion relief and peak shaving control.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the power grid congestion mitigation and peak shaving method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to perform the power grid congestion mitigation and peak shaving method as described in any one of claims 1 to 7.

Images