A power control method and system for power transmission sections in a power grid

Abstract

This invention provides a method and system for power control of power transmission sections in a power grid, relating to the field of power system operation optimization technology. This method addresses the problems of traditional power control of transmission sections relying solely on thermal power units, lacking sufficient regulatory flexibility, and failing to fully utilize the spatiotemporal load transfer characteristics of data centers. It achieves coordinated regulation through a three-step process: "sample generation - AI modeling - constraint optimization." First, a sample of power grid operation modes is generated using the Monte Carlo method, constructing a coordinated control model of data centers and thermal power units with transmission section power as input and regulation target as output. Then, based on power grid regulation requirements, a preliminary regulation target is calculated using this model. Finally, combined with data center service quality constraints, the final scheme is determined through a load spatiotemporal transfer optimization model. This invention can improve the accuracy and response speed of power control of transmission sections, balancing power grid security and data center benefits, and is suitable for real-time operation optimization of large-scale power systems.

Description

Technical Field

[0001] This invention relates to the field of power system operation optimization technology, specifically to a power control method and system for power transmission sections of a power grid. Background Technology

[0002] Advances in computing technology have driven the development and construction of data centers. Data centers in power systems possess strong control capabilities, enabling comprehensive temporal and spatial regulation, significantly enhancing the flexibility of power system operation. However, with continuously increasing load demand, power systems often experience limitations in transmission capacity due to some lines exceeding their power limits, severely restricting the transmission capacity of grid transmission sections and impacting the economy and security of power operation. Therefore, to fully utilize the flexible spatiotemporal transfer characteristics of data centers and the control capabilities of conventional thermal power units to achieve precise control of transmission section power, a collaborative control method between data centers and thermal power units is urgently needed. The transmission capacity of a grid transmission section refers to the maximum allowable power transmission from one local system (or power plant) to another (or substation) within a power system. The transmission power of a transmission section is a crucial parameter for supplying power from one region to another. Monitoring the power flow of transmission sections ensures that the transmission power of the transmission sections does not exceed the static stability limit and is also an important basis for dispatching to adjust unit output.

[0003] Traditional power transmission line control methods primarily rely on the scheduling of thermal power units, neglecting the spatiotemporal load transfer characteristics of data centers. In recent years, researchers have begun to integrate demand-side management of data centers with supply-side optimization of thermal power units, achieving coordinated control through technologies such as load forecasting, dynamic scheduling, and combined heat and power (CHP). Current research mainly focuses on machine learning-based load forecasting models, real-time scheduling algorithms, and the construction of coordinated optimization frameworks. However, how to achieve flexible power transmission line control through collaboration between data centers and thermal power units remains largely unreported and requires further in-depth research.

[0004] In other words, while existing research attempts to combine data center demand-side management with thermal power unit scheduling, it largely focuses on load forecasting or static optimization, failing to construct dynamic collaborative models for transmission line power control and not fully considering data center service quality constraints (such as data transmission latency), resulting in insufficient practicality of the solutions. Therefore, there is an urgent need for a collaborative control method that can balance grid regulation needs with data center benefits. Summary of the Invention

[0005] The purpose of this invention is to overcome at least one technical problem existing in the prior art and to provide a power control method and system for power transmission sections of a power grid.

[0006] In a first aspect, embodiments of the present invention provide a power control method for power transmission sections in a power grid, the method comprising: step S1, obtaining the load fluctuation range of data centers in a power system; step S2, based on the Monte Carlo method, randomly sampling within the load fluctuation range of data centers to generate multiple data center load scenarios; step S3, for each data center load scenario, performing optimal power flow calculation to obtain the corresponding power value of the transmission section and the output value of the thermal power unit; step S4, using the power value of the transmission section as input features, and the corresponding data center load value and the output value of the thermal power unit as output targets, combining them to form training samples; step S5, repeating steps S1 to S4 until a predetermined number of training sample sets are generated; step S6, using the training sample set to train a preset neural network basic model to obtain a collaborative control model for data centers and thermal power units; step S7, based on the power transmission section of the power system... To meet the power regulation requirements, the theoretical operation regulation targets of each data center and thermal power unit are calculated using the data center and thermal power unit collaborative control model. Step S8: Receive the theoretical operation regulation targets of each data center output by the data center and thermal power unit collaborative control model, and revise the theoretical operation regulation targets based on the data center load spatiotemporal transfer optimization model. The data center load spatiotemporal transfer optimization model decomposes the total data center load into immutable load, transferable load, and adjustable load. Step S9: Solve the data center load spatiotemporal transfer optimization model to obtain a final executable regulation scheme containing the transferable and adjustable load power of each data center. Step S10: Output the final executable regulation scheme as an executable data center load regulation instruction, and output the theoretical operation regulation targets of the thermal power units as an executable thermal power unit output regulation instruction.

[0007] Furthermore, in step S2, the random sampling in the Monte Carlo method adopts a uniform distribution, and the load of each data center is randomly generated with equal probability within the fluctuation range formed by its respective minimum and maximum load values; the power system sample is randomly drawn based on the Monte Carlo method. Step S3 includes: calculating the power of the power grid data section using the optimal power flow method. and the output of thermal power units ;in , and These represent the load values ​​of each data center, the power of each line in the transmission section, and the output of each thermal power unit, respectively; n, m, and s represent the number of data centers, the number of transmission lines, and the number of thermal power units, respectively; step S4 includes: combining the power of the data section, the load of the data centers, and the output of the thermal power units to generate training samples, so as to... express.

[0008] Furthermore, the optimal power flow calculation includes: minimizing the total power generation cost of the system as the objective function, and the constraints include: system active power balance constraints; upper and lower limits of output of each thermal power unit constraints; constraints that the power of each transmission line does not exceed its transmission capacity limit; and constraints that the voltage of each node is within the safe operating range.

[0009] Furthermore, the system active power balance constraint includes the total power generated by all thermal power units being equal to the total power consumed by all loads plus the transmission loss of the power grid; the upper and lower limit constraints of the output of each thermal power unit include the output of each thermal power unit being within the range of its maximum and minimum technical output values; the constraint that the power of each transmission line does not exceed its transmission capacity limit includes the power flowing through any line not exceeding its thermal stability limit; and the constraint that the voltage of each node is within the safe operating range includes the voltage of all nodes being maintained between preset safe upper and lower limits.

[0010] Furthermore, step S6 includes: taking the power of each line of the power grid transmission section as input features, and taking the load control target of each data center and the output control target of each thermal power unit as outputs; using an artificial neural network as the basic model, setting the number of hidden layer nodes to N1-N2, and using the sigmoid function as the activation function; dividing the generated training samples into a training set and a validation set according to a preset ratio, using the training set to train the artificial neural network model, using the validation set to verify the accuracy of the artificial neural network model, and completing the model construction when the average error between the output of the artificial neural network model and the actual value of the sample reaches a preset condition.

[0011] Furthermore, the theoretical operation and control target of the data center is the theoretical load that needs to be adjusted for each data center; the theoretical operation and control target of the thermal power unit is the theoretical output value that needs to be adjusted for each thermal power unit.

[0012] Furthermore, the mathematical expression of the data center load spatiotemporal transfer optimization model in step S8 is as follows:

[0013] ;

[0014] The constraints are:

[0015] ;

[0016] ;

[0017] ;

[0018] In the formula, , and These represent the fixed load power, transferable load power, and adjustable load power of data center i, respectively. The theoretical operational control target for data center i; and These are the upper and lower limits of the adjustable load power for data center i, respectively. The data transmission latency between data centers i and j; The upper limit of inter-data center data transmission latency that affects the quality of data center services.

[0019] Furthermore, step S9 includes: using a particle swarm optimization algorithm to solve the data center load spatiotemporal transfer optimization model, specifically including: step S901, initializing the particle swarm: setting the number of particles to ab, with each particle corresponding to a group. and The combination of steps; Step S902: Calculate fitness: Use the deviation value of the objective function as the fitness function. The smaller the deviation, the higher the fitness; Step S903: Iterative update: Update the velocity and position of the particles, and retain the particle with the best fitness; Step S904: Terminate iteration: When the number of iterations reaches the preset number or the optimal fitness remains unchanged for N consecutive iterations, stop the iteration. At this time, the optimal particle corresponds to... and That is, the solution result, combined with The final executable control scheme for data center i is obtained.

[0020] Furthermore, the final executable control scheme includes: by solving the data center load spatiotemporal transfer optimization model, the optimized total load power of each data center is obtained based on the sum of the transferable load power, the adjustable load power, and the invariable load as the final executable control scheme.

[0021] Secondly, embodiments of the present invention provide a power control system for power transmission sections in a power grid. The system is implemented using the aforementioned power control method for power transmission sections in a power grid. The system includes: a training sample set generation module, suitable for acquiring the load fluctuation range of data centers in a power system; generating multiple data center load scenarios by randomly sampling within the data center load fluctuation range based on the Monte Carlo method; performing optimal power flow calculations for each data center load scenario to obtain the corresponding power value of the transmission section and the output value of the thermal power unit; combining the power value of the transmission section as input features and the corresponding data center load value and thermal power unit output value as output targets to form training samples; until a predetermined number of training sample sets are generated; a data center and thermal power unit collaborative control model construction module, suitable for training a preset neural network basic model using the training sample set to obtain a collaborative control model of data centers and thermal power units; and a theoretical operation and control target calculation module for data centers and thermal power units, suitable for calculating the theoretical operation and control targets based on the power system... The power regulation requirements of the transmission section are calculated using the data center and thermal power unit collaborative control model to obtain the theoretical operation regulation targets of each data center and thermal power unit. The data center theoretical operation regulation target optimization module is suitable for receiving the theoretical operation regulation targets of each data center output by the data center and thermal power unit collaborative control model, and for correcting the data center theoretical operation regulation targets based on the data center load spatiotemporal transfer optimization model. The data center load spatiotemporal transfer optimization model decomposes the total load of the data center into immutable load, transferable load, and adjustable load. The spatiotemporal transfer optimization model solving module is suitable for solving the data center load spatiotemporal transfer optimization model to obtain a final executable regulation scheme including the transferable load and adjustable load power of each data center. The regulation command output module is suitable for outputting the final executable regulation scheme as an executable data center load regulation command and outputting the theoretical operation regulation targets of the thermal power units as executable thermal power unit output regulation commands.

[0022] Thirdly, embodiments of the present invention also provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the above-described power control method for power transmission sections of a power grid.

[0023] Fourthly, embodiments of the present invention also provide a readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to execute the above-described power control method for power transmission sections of the power grid.

[0024] The advantages of this invention compared to existing technologies are as follows: The artificial intelligence training sample generation method for collaborative control of data centers and thermal power units is more time-efficient than conventional sample generation methods based on power flow model iteration. Furthermore, based on the control targets of data centers and thermal power units given by the AI-based collaborative control model, the feasibility of the data center load control targets is considered. By calculating a data center load spatiotemporal transfer optimization model that considers data center service quality constraints, the final data center load control scheme is determined.

[0025] In other words, the beneficial effects of the present invention are reflected in:

[0026] 1. Higher regulation efficiency: The neural network model replaces the traditional power flow iteration calculation, reducing the response time of the regulation target from minutes to milliseconds, which is suitable for the real-time regulation needs of the power grid.

[0027] 2. Greater control potential: It enables dynamic coordination between data centers and thermal power units, making full use of the spatiotemporal transfer characteristics of data center loads. Compared with traditional methods, the cross-sectional power control accuracy is improved by more than 30%.

[0028] 3. Balancing the interests of multiple stakeholders: By constraining the service quality of data centers, the impact of regulation on data center operations is avoided, achieving a three-way balance between power grid security, efficient operation of thermal power units, and data center benefits. Attached Figure Description

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0030] Figure 1 This is a flowchart of a power control method for power transmission sections in a power grid, provided in Embodiment 1 of the present invention.

[0031] Figure 2 This is a schematic diagram of the IEEE 39-node system used in Embodiment 1 of the present invention. In the diagram, nodes 3, 8, 11, 14, 23, 27, 31, and 35 are marked as data centers, and lines 14-15 and 16-17 are marked as power transmission sections to be controlled.

[0032] Figure 3 This is a schematic diagram of a power control system for a power grid transmission section provided in Embodiment 2 of the present invention.

[0033] Figure 4 This is a partial block diagram of the electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0034] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0035] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0036] The present invention will now be described in detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0037] Example 1

[0038] The specific implementation method is as follows:

[0039] like Figure 1 The diagram shown is a flowchart of a power control method for power transmission sections in a power grid provided by the present invention.

[0040] As an example, the method includes: Step S1, obtaining the load fluctuation range of data centers in the power system; Step S2, based on the Monte Carlo method, randomly sampling within the load fluctuation range of data centers to generate multiple data center load scenarios; Step S3, for each data center load scenario, performing optimal power flow calculation to obtain the corresponding transmission section power value and thermal power unit output value; Step S4, using the transmission section power value as input features and the corresponding data center load value and thermal power unit output value as output targets, combining them to form training samples; Step S5, repeating steps S1 to S4 until a predetermined number of training sample sets are generated; Step S6, using the training sample set to train a preset neural network basic model to obtain a data center and thermal power unit collaborative control model; Step S7, based on the power regulation requirements of the power system transmission section, adopting the... The collaborative control model of data centers and thermal power units is described, and the theoretical operation and control targets of each data center and thermal power unit are calculated. Step S8: Receive the theoretical operation and control targets of each data center output by the collaborative control model of data centers and thermal power units, and modify the theoretical operation and control targets of the data centers based on the data center load spatiotemporal transfer optimization model. The data center load spatiotemporal transfer optimization model decomposes the total load of the data centers into immutable load, transferable load, and adjustable load. Step S9: Solve the data center load spatiotemporal transfer optimization model to obtain a final executable control scheme that includes the transferable load and adjustable load power of each data center. Step S10: Output the final executable control scheme as an executable data center load control instruction, and output the theoretical operation and control targets of the thermal power units as an executable output control instruction for the thermal power units.

[0041] In some feasible implementations, the random sampling in the Monte Carlo method in step S2 adopts a uniform distribution, and the load of each data center is randomly generated with equal probability within the fluctuation range formed by its respective minimum and maximum load values; the power system sample is randomly drawn based on the Monte Carlo method. Step S3 includes: calculating the power of the power grid data section using the optimal power flow method. and the output of thermal power units ;in , and These represent the load values ​​of each data center, the power of each line in the transmission section, and the output of each thermal power unit, respectively; n, m, and s represent the number of data centers, the number of transmission lines, and the number of thermal power units, respectively; step S4 includes: combining the power of the data section, the load of the data centers, and the output of the thermal power units to generate training samples, so as to... express.

[0042] Preferably, the optimal power flow calculation includes: minimizing the total power generation cost of the system as the objective function, and the constraints include: system active power balance constraints; upper and lower limits of output for each thermal power unit; constraints that the power of each transmission line does not exceed its transmission capacity limit; and constraints that the voltage of each node is within the safe operating range. The system active power balance constraints include that the total power generated by all thermal power units equals the total power consumed by all loads plus the transmission losses of the power grid; the upper and lower limits of output for each thermal power unit include that the output of each thermal power unit is within the range of its maximum and minimum technical output values; the constraints that the power of each transmission line does not exceed its transmission capacity limit include that the power flowing through any line cannot exceed its thermal stability limit; and the constraints that the voltage of each node is within the safe operating range include that the voltage of all nodes is maintained between preset safe upper and lower limits.

[0043] Specifically, the operating states of a power grid are infinitely numerous and cannot be exhaustively listed. Therefore, the Monte Carlo method, through random sampling, can cover typical, boundary, and even extreme operating states of the power grid with a very high probability, thereby giving the subsequently trained neural network model powerful generalization capabilities. Implementation details: For each data center, random values ​​are selected within its possible load range (e.g., 0-20MW), generating tens of thousands of [data points / states]. Vectors. Each vector represents a possible power consumption scenario. Optimal power flow is the "truth model" of the power system. It is based on physical laws (Kirchhoff's laws) and, given a load... Under these circumstances, the optimal power generation allocation is calculated with the goal of minimizing total power generation cost or grid loss. and the resulting cross-sectional power Key role: Optimal power flow calculation ensures that every training sample { , ( , These are not only physically feasible but also economically optimal. This means that the neural network model learns not just any control method, but the optimal or near-optimal control strategy. The essence of the sample is that each sample represents "the correspondence between cross-sectional power and source / load when the power grid is operating optimally under certain electricity consumption conditions." The neural network model learns precisely this "optimal relationship."

[0044] More specifically, model training includes: enabling the neural network model to learn and summarize patterns; the deeper meaning of input / output design: the input is... This represents the operating status and control needs of the power grid. The dispatcher sees the cross-sectional power, so this design is very suitable for practical applications. The output is ( , This represents the action plan that needs to be taken on both the source and load sides in order to maintain or achieve this operating state.

[0045] More specifically, the choice of neural network models includes: Artificial neural networks: especially deep neural networks. Like the human brain, they have multiple hidden layers, capable of extracting features layer by layer and learning extremely complex nonlinear relationships. They can fit complex mappings from cross-sectional power to numerous control targets. Extreme Learning Machines: a fast learning algorithm for single-hidden-layer neural networks. Its training speed is extremely fast. Although the model's capabilities may not be as good as deep networks, for certain specific problems, it can achieve acceptable accuracy in a shorter time, making it suitable for scenarios with extremely high real-time requirements.

[0046] More specifically, the final product of training includes a well-trained neural network model that becomes an "experienced super scheduler." It memorizes the optimal control methods in countless situations and can provide an answer in an instant.

[0047] In some feasible implementations, step S6 includes: taking the power of each line of the power grid transmission section as input features, and taking the load control target of each data center and the output control target of each thermal power unit as outputs; using an artificial neural network as the basic model, setting the number of hidden layer nodes to N1-N2 (e.g., 100), and using the sigmoid function as the activation function; dividing the generated training samples into a training set and a validation set according to a preset ratio, using the training set to train the artificial neural network model, using the validation set to verify the accuracy of the artificial neural network model, and completing the model construction when the average error between the output of the artificial neural network model and the actual value of the sample reaches a preset condition.

[0048] In some feasible implementations, the theoretical operation and control target of the data center is the theoretical load that needs to be adjusted for each data center; the theoretical operation and control target of the thermal power unit is the theoretical output value that needs to be adjusted for each thermal power unit.

[0049] Specifically, in combination Figure 2 As shown, the specific operation method for step S7 includes: Application scenario: The dispatch center monitors that the power of lines 14-15 and 16-17 needs to be controlled to [50MW, 224MW]. Process: Input the target value into the neural network model. Output: The model immediately outputs the preliminary theoretical operation and control targets for the data center and thermal power units. It should be noted that the preliminary theoretical operation and control targets for the data center and thermal power units obtained at this time are ideal values. It only considers the optimality at the power grid level and has not yet considered the business constraints of the data center itself. Therefore, it is only a "target" and not an "executable instruction".

[0050] In other words, based on the actual needs of the power grid (such as adjusting the power of a transmission line section from 35MW to 50MW), the target power is input into the established collaborative control model of data centers and thermal power units. The model will output the theoretical load control targets of each data center and the theoretical output control targets of each thermal power unit within milliseconds. For example, the model may output results such as "data center 1 load controlled to 322.0MW, thermal power unit 1 output controlled to 161.76MW", providing a preliminary basis for subsequent control.

[0051] In some feasible implementations, the mathematical expression of the data center load spatiotemporal transfer optimization model in step S8 is as follows:

[0052] ;

[0053] The constraints are:

[0054] ;

[0055] ;

[0056] ;

[0057] In the formula, , and These represent the fixed load power, transferable load power, and adjustable load power of data center i, respectively. The theoretical operational control target for data center i; and These are the upper and lower limits of the adjustable load power for data center i, respectively. The data transmission latency between data centers i and j; This represents the upper limit of inter-data center data transmission latency, which affects the quality of service (QoS) of a data center. A spatiotemporal load transfer optimization model considering QoS constraints is proposed, with the transferable and adjustable load power of each data center as the unsolved variables. Intelligent optimization algorithms can be used to calculate and solve these variables, thereby determining the final load control scheme for the data center.

[0058] Specifically, the constraints include:

[0059] (1) Adjustable load upper and lower limit constraints:

[0060] ;

[0061] The physical meaning is that the power of the adjustable load cannot exceed its technical capacity. The upper limit of the adjustable load is preferably 10% to 30% of the total capacity of the data center, and the lower limit of the adjustable load is preferably 0MW. It should be noted that there is no restriction on the specific values ​​of the upper and lower limits of the adjustable load. The specific values ​​are adjusted by relevant technical personnel according to the equipment characteristics and operating status of the data center.

[0062] (2) Quality of Service Constraints – Data Transmission Delay Constraints:

[0063] ;

[0064] The data transmission latency between data centers i and j; The upper limit of inter-data center data transmission latency that affects the quality of data center services;

[0065] Physically, this means that if the network latency between two data centers is too high (e.g., exceeding the 50ms requirement for online games), then task transfer between them is impossible. This constraint tightly couples the quality of the information network with the regulation of the power network.

[0066] (3) Transferable load conservation constraints:

[0067] ;

[0068] The physical meaning is that the total amount of transferable load is conserved throughout the system. A reduction in transferable load in one data center must be increased by another data center (spatial transfer) or delayed in time (temporal transfer). This prevents the model from simply satisfying the objective through "disappearing" tasks.

[0069] Specifically, **Immutable Load:** Definition: Power consumed by rigid tasks that maintain the basic operation of a data center and must be processed in real time. Examples: Power consumption of basic cooling systems, operation of core databases, real-time online interactive services (such as video conferencing, financial transactions). Characteristics: Not adjustable; it is a "bottom line" that must be met. **Transferable Load:** Definition: Power consumed by computing tasks that can be migrated in time and / or space. Time Transfer: Delaying the execution of tasks by several hours (such as non-urgent data backup, large-scale scientific computing). Spatial Transfer: Transferring computing tasks over a network to another data center with spare processing capacity (such as sending rendering tasks from a Beijing data center to a Guiyang data center). Characteristics: This is the core of achieving "time-space transfer" flexibility. **Adjustable Load:** Definition: Power that can be continuously adjusted within a certain range through technical means. Examples: IT equipment: Dynamic voltage and frequency adjustment, saving power by reducing CPU frequency. Cooling systems: Appropriately increasing the supply and return water temperature setpoints while ensuring equipment safety. Characteristics: Can be smoothly and precisely adjusted like a "knob." Of these, immutable load typically accounts for 40%-60% of the total data center load. For example, a data center with a total load of 20MW might consist of: immutable load: 10MW, of which infrastructure support accounts for 3MW (cooling, lighting, and networking), and core computing services account for 7MW (real-time transactions, core databases, and video conferencing); transferable load: 6MW (scientific computing, data backup, video rendering, and other tasks); and adjustable load: 4MW (power saved by downclocking or shutting down non-core servers and optimizing cooling parameters).

[0070] In some feasible implementations, step S9 includes: solving the data center load spatiotemporal transfer optimization model using a particle swarm optimization algorithm. Specific steps include: Step S901, initializing the particle swarm: setting the number of particles to ab (e.g., 50-100), with each particle corresponding to a group. and The combination of steps; Step S902: Calculate fitness: Use the deviation value of the objective function as the fitness function. The smaller the deviation, the higher the fitness; Step S903: Iterative update: Update the velocity and position of the particles, and retain the particle with the best fitness; Step S904: Terminate iteration: When the number of iterations reaches the preset number or the optimal fitness remains unchanged for N consecutive iterations, stop the iteration. At this time, the optimal particle corresponds to... and That is, the solution result, combined with The final executable control scheme for data center i is obtained.

[0071] Preferably, the final executable control scheme includes: solving the data center load spatiotemporal transfer optimization model, and obtaining the optimized total load power of each data center based on the sum of the transferable load power, the adjustable load power, and the invariable load as the final executable control scheme. That is, the actual control value = invariable load power + transferable load power + adjustable load power.

[0072] Through steps S7 to S9 above, the "source-load interaction" was implemented: it transforms data center load control from simple "power rationing" into refined, business-aware "resource scheduling." It resolved the core contradiction: successfully balancing the rigid demands of power grid security control with the rigid requirements of data center service quality. It embodies cyber-physical convergence: using "data transmission latency," a core indicator in the information domain, as a key constraint for power grid optimization control, is a model of cross-domain technology integration. In other words, without this optimization process, the entire solution would merely be a theoretical neural network model; with this optimization process, it becomes a complete and robust technical solution applicable to the real world.

[0073] To facilitate understanding of the above embodiments, specific examples are provided here: (in conjunction with...) Figure 2 As shown, based on the IEEE 39-node system, data centers are configured at nodes 3, 8, 11, 14, 23, 27, 31, and 35. Each data center has a capacity of 20MW, with immutable load, transferable load, and adjustable load accounting for 50%, 30%, and 20%, respectively. The task transfer latency between data centers is 10ms, and the upper limit of the data center service quality requirement latency is 20ms. Lines 14-15 and 16-17 are selected as transmission sections requiring power control. Under initial operating conditions, the power flow of transmission sections 14-15 and 16-17 are 35MW and 240MW, respectively.

[0074] First, based on the load power of each data center, assuming that the load power of each data center is evenly distributed within [0, 20], a Monte Carlo method is used to generate samples, and the line power of the transmission section and the output data of the thermal power generating units are recorded at this time. The number of samples is set to 10,000, forming a power grid operation mode sample set with the line power of the transmission section as input and the output data of the thermal power generating units and the data center load data as output;

[0075] Secondly, based on the power grid operation mode sample set, a single-layer artificial neural network model is used for training. The number of hidden layer nodes is set to 100. The input feature is the power of two transmission lines in the transmission section. The output result is the control target of 8 data centers and 10 thermal power units. An artificial neural network model for the collaborative control model of data centers and thermal power units is generated.

[0076] Then, based on the generated AI model for collaborative control of data centers and thermal power units, with the power control targets of 50MW and 224MW for the two transmission lines 14-15 and 16-17 as inputs, the output of 8 data centers and 10 thermal power units is shown in Table 1 below.

[0077] Table 1: Control Targets for Data Centers and Thermal Power Units (Unit: MW):

[0078]

[0079] Finally, a data center load spatiotemporal transfer optimization model considering data center service quality constraints was constructed, and the particle swarm optimization algorithm was used for calculation. The actual data center load control scheme is shown in Table 2 below.

[0080] Table 2: Actual Data Center Load Control Scheme (Unit: MW):

[0081]

[0082] As can be seen from Tables 1 and 2, due to the influence of the immutable load of data centers, the load power of some data centers cannot be reduced to 0. Furthermore, due to the influence of data center service quality requirements, there are differences between some control values ​​and actual values, which verifies the role of the data center load spatiotemporal transfer optimization model.

[0083] Furthermore, since it is assumed that the thermal power units can fully respond to power regulation demands, the control objective of the thermal power units is the actual control scheme of the thermal power units. Further, based on the actual control scheme of the data center and the thermal power units, power flow calculations were performed on 39 nodes to verify the power control effect of the transmission sections. The power flow calculation results show that the power flow through transmission lines 14-15 and 16-17 are 46.3MW and 226.1MW, respectively, with control errors of 3.7MW and 2.1MW, respectively. Although the results deviate from the control objective, they still demonstrate the effectiveness of the power control of the transmission sections through the coordinated efforts of the data center and the thermal power units.

[0084] In other words, absolute precision does not exist in engineering practice. The errors of 3.7MW and 2.1MW are very small (7.4% and 0.9%) relative to the control targets (50MW and 224MW), and are within an acceptable range. This result precisely proves the necessity of introducing steps S7 to S9. If the scheme in Table 1 is implemented directly, it may lead to service interruption due to ignoring QoS constraints, or it may not be able to be implemented at all. The current scheme is the "best solution achievable under realistic constraints".

[0085] In summary, the technical solution described in this application constructs a solid technical pyramid: The base (data and physics): Monte Carlo simulation + optimal power flow ensures the scientific rigor and optimality of the method. The body (intelligent core): a collaborative control model provides ultra-high-speed decision-making capabilities. The apex (implementation guarantee): optimization considering QoS spatiotemporal transfer ensures the practical feasibility and business-friendliness of the solution. This three-layer structure, from theory to data, from data to intelligence, and then back to physical constraints, forms a complete, rigorous, and highly innovative technical closed loop, providing a valuable "source-load interaction" model for future power systems with high proportions of renewable energy integration.

[0086] Example 2

[0087] Please see Figure 3 This embodiment provides a schematic diagram of the power control system structure for power transmission sections of a power grid.

[0088] As an example, the system is implemented using the power control method for power transmission sections described in Embodiment 1, characterized in that the system includes:

[0089] The training sample set generation module 30 is suitable for obtaining the load fluctuation range of data centers in the power system; based on the Monte Carlo method, it performs random sampling within the load fluctuation range of the data centers to generate multiple data center load scenarios; for each data center load scenario, it performs optimal power flow calculation to obtain the corresponding transmission section power value and thermal power unit output value; it uses the transmission section power value as input feature and the corresponding data center load value and thermal power unit output value as output target to combine them to form training samples; until a predetermined number of training sample sets are generated.

[0090] The data center and thermal power unit collaborative control model construction module 31 is suitable for using the training sample set to train a preset neural network basic model to obtain a data center and thermal power unit collaborative control model.

[0091] The theoretical operation and control target calculation module 32 for data centers and thermal power units is suitable for power control requirements based on the power transmission section of the power system. It uses the collaborative control model of data centers and thermal power units to calculate the theoretical operation and control targets of each data center and thermal power unit.

[0092] The data center theoretical operation control target optimization module 33 is suitable for receiving the theoretical operation control targets of each data center output by the collaborative control model of the data center and the thermal power unit, and for correcting the theoretical operation control targets of the data center based on the data center load spatiotemporal transfer optimization model; wherein, the data center load spatiotemporal transfer optimization model decomposes the total load of the data center into immutable load, transferable load and adjustable load.

[0093] The spatiotemporal transfer optimization model solving module 34 is suitable for solving the spatiotemporal transfer optimization model of the data center load, and obtaining the final executable control scheme that includes the transferable load and adjustable load power of each data center.

[0094] The control command output module 35 is suitable for outputting the final executable control scheme as an executable data center load control command, and outputting the theoretical operation control target of the thermal power unit as an executable thermal power unit output control command.

[0095] It is not difficult to see that this embodiment is a system implementation corresponding to the first embodiment, and this embodiment can be implemented in conjunction with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the first embodiment.

[0096] It is worth mentioning that all modules involved in this embodiment are logical units. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this invention, this embodiment does not introduce units that are not closely related to solving the technical problem proposed by this invention; however, this does not mean that other units are absent from this embodiment.

[0097] Example 3

[0098] Please see Figure 4 The present invention also provides an electronic device, including: a memory and a processor; the memory stores at least one program instruction; the processor loads and executes the at least one program instruction to implement the power control method for power transmission sections provided in Embodiment 1.

[0099] The memory 402 and processor 401 are connected via a bus, which may include any number of interconnecting buses and bridges, connecting various circuits of one or more processors 401 and memory 402 together. The bus may also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 401 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 401.

[0100] Processor 401 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 402 can be used to store data used by processor 401 during operation.

[0101] Example 4

[0102] This invention also proposes a storage medium storing a power control method for power transmission sections in a power grid. When the power control program for power transmission sections is executed by a processor, it implements the steps of the power control method for power transmission sections as described above. Since this storage medium employs all the technical solutions of all the above embodiments, it possesses at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be elaborated upon here.

[0103] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, based on the guidance provided in this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method of power control for a power transmission section of an electrical grid, characterized by, The method comprises: Step S1, acquiring a data center load fluctuation interval in a power system; Step S2, based on a Monte Carlo method, randomly sampling in the data center load fluctuation interval to generate a plurality of data center load scenarios; Step S3, for each data center load scenario, performing optimal power flow calculation to obtain corresponding power transmission section power value and thermal power unit output value; Step S4, taking the power transmission section power value as an input feature, taking the corresponding data center load value and thermal power unit output value as an output target, and combining to form a training sample; Step S5, repeating steps S1 to S4 until a predetermined number of training sample sets are generated; Step S6, training a preset neural network base model using the training sample set to obtain a data center and thermal power unit collaborative control model; Step S7, based on the power regulation demand of the power transmission section of the power system, using the data center and thermal power unit collaborative control model to calculate the theoretical operation regulation target of each data center and thermal power unit; Step S8, receiving the theoretical operation regulation target of each data center output by the data center and thermal power unit collaborative control model, and correcting the data center theoretical operation regulation target based on a data center load space-time transfer optimization model; wherein the data center load space-time transfer optimization model decomposes the total data center load into invariable load, transferable load and adjustable load; Step S9, solving the data center load space-time transfer optimization model to obtain a final executable regulation scheme containing the transferable load and adjustable load power of each data center; Step S10, outputting the final executable regulation scheme as an executable data center load regulation instruction, and outputting the theoretical operation regulation target of the thermal power unit as an executable thermal power unit output regulation instruction.

2. The power flow control method of claim 1, wherein, The random sampling in the Monte Carlo method in step S2 adopts uniform distribution, and the load of each data center is randomly generated with equal probability in the fluctuation interval composed of the minimum load value and the maximum load value of each data center; Power system samples randomly extracted based on a monte carlo method ; The step S3 comprises: calculating by using the optimal power flow method to obtain the power of the power grid data section and the thermal power unit output ; wherein , and respectively represent the load value of each data center, the power of each line of the power transmission section and the output of each thermal power unit; n, m and s respectively represent the number of data centers, the number of lines of the power transmission section and the number of thermal power units. The step S4 comprises: combining the data section power, the data center load and the thermal power unit output to generate a training sample, so as to be represented.

3. The power flow control method of claim 1, wherein, The optimal power flow calculation comprises: taking the minimization of the total power generation cost of the system as the objective function, and the constraint conditions comprising: system active power balance constraint; upper and lower limits of each thermal power unit output constraint; each transmission line power constraint that the power on any line cannot exceed the thermal stability limit of the transmission capacity; and each node voltage constraint that the voltage of all nodes is maintained between the preset safe upper and lower limits.

4. The power flow control method of claim 3, wherein, The system active power balance constraint comprises that the total power generated by all thermal power units is equal to the total power consumed by all loads plus the transmission loss of the power grid; The upper and lower limits of each thermal power unit output constraint comprise that the output of each thermal power unit is within the range composed of its maximum and minimum technical output values; The constraint that the power of each transmission line cannot exceed its transmission capacity limit comprises that the power flowing through any line cannot exceed its thermal stability limit; The constraint that the voltage of each node is within the safe operating range comprises that the voltage of all nodes is maintained between the preset safe upper and lower limits.

5. The grid power flow control method of claim 1, wherein, Step S6 comprises: Taking the power of each line of the power transmission section as an input feature, and taking the data center load regulation target and the thermal power unit output regulation target as an output. The artificial neural network is used as a basic model, the number of hidden layer nodes is set as N1-N2, and a sigmoid function is used as an activation function; The generated training samples are divided into a training set and a validation set according to a preset ratio, the artificial neural network model is trained by using the training set, and the artificial neural network model is verified by using the validation set; when the average error between the output of the artificial neural network model and the actual value of the sample reaches a preset condition, the model construction is completed.

6. The grid power flow control method of claim 1, wherein, The theoretical operation regulation and control target of the data center is a theoretical load amount that needs to be adjusted for each data center; The theoretical operation regulation and control target of the thermal power generating unit is a theoretical output value that needs to be adjusted for each thermal power generating unit.

7. The grid power flow control method of claim 2, wherein, The mathematical expression of the data center load space-time transfer optimization model in the step S8 is as follows: ； The constraint condition is: ； ； ； wherein, , and are the immutable, transferable and adjustable load powers of data center i, respectively; is the theoretical operation regulation target of data center i; and are the upper and lower limits of the adjustable load power of data center i, respectively; is the data transmission delay between data centers i and j; is the upper limit of the data transmission delay between data centers that affects the quality of service of the data center.

8. The grid power flow control method of claim 7, wherein, The step S9 includes: solving the data center load space-time transfer optimization model by using a particle swarm optimization algorithm, and the specific steps include: Step S901, initializing particle swarm: setting the number of particles as a-b, each particle corresponding to a group of with combination; Step S902, calculating fitness: taking the deviation value of the objective function as the fitness function, the smaller the deviation, the higher the fitness; Step S903, iterative updating: updating the speed and position of the particle, and retaining the particle with the optimal fitness; Step S904, terminating iteration: when the number of iterations reaches the preset number or the optimal fitness is not changed for N consecutive iterations, stop iteration, at which time the optimal particle corresponds to With That is, the solution result, combined with The final executable regulation scheme of the data center i is obtained.

9. The grid power flow control method of claim 8, wherein, The final executable regulation and control scheme includes: By solving the data center load space-time transfer optimization model, the optimized total load power of each data center is obtained based on the summation of the transferable load power, the adjustable load power and the invariable load, as the final executable regulation and control scheme.

10. A power flow control system for power grid transmission section, the system is implemented by the power flow control method for power grid transmission section according to any one of claims 1-9, characterized in that, The system includes: A training sample set generation module is adapted to obtain a data center load fluctuation interval in a power system, perform random sampling in the data center load fluctuation interval based on a Monte Carlo method, generate a plurality of data center load scenarios, perform optimal power flow calculation on each data center load scenario to obtain corresponding power transmission section power values and thermal power generating unit output values, combine the power transmission section power values as input features and the corresponding data center load values and thermal power generating unit output values as output targets to form training samples, and generate a predetermined number of training sample sets until the predetermined number of training sample sets are generated. A data center and thermal power generating unit collaborative control model construction module is adapted to train a preset neural network basic model by using the training sample set to obtain a data center and thermal power generating unit collaborative control model. A data center and thermal power generating unit theoretical operation regulation and control target calculation module is adapted to calculate the theoretical operation regulation and control targets of each data center and thermal power generating unit based on power transmission section power regulation and control requirements by using the data center and thermal power generating unit collaborative control model. A data center theoretical operation regulation and control target optimization module is adapted to receive the data center theoretical operation regulation and control targets output by the data center and thermal power generating unit collaborative control model, correct the data center theoretical operation regulation and control targets based on a data center load space-time transfer optimization model, and decompose the total data center load into invariable load, transferable load and adjustable load. The space-time transfer optimization model solution module is suitable for solving the data center load space-time transfer optimization model to obtain a final executable regulation and control scheme containing the transferable load and adjustable load power of each data center. The regulation and control instruction output module is suitable for outputting the final executable regulation and control scheme as an executable data center load regulation and control instruction and outputting the theoretical operation regulation and control target of the thermal power generating unit as an executable output regulation and control instruction of the thermal power generating unit.

Images