Intelligent matching method for wind and light resources and data center load in extreme weather situation

Abstract

The invention provides a wind and light resource and data center load intelligent matching method in an extreme weather situation, and the method comprises the steps: carrying out the time-space alignment of the obtained meteorological sudden change data, multi-physical field equipment state parameters, power grid operation parameters and load demands, and obtaining a four-dimensional time series data set; inputting the four-dimensional time series data set into the dynamic prediction model to generate a cascade failure conduction factor matrix; wherein the dynamic prediction model is used for representing a conduction relation between a meteorological parameter sudden change threshold value and equipment cascade failure; taking the cascade failure conduction factor matrix as a constraint condition, performing multi-objective optimization solution under space-time constraint, and generating an elastic matching instruction set; wherein the elastic matching instruction set is used for indicating the energy system to execute intelligent resource scheduling. By adopting the method, the limitation of traditional single-factor static correlation analysis can be broken through, and collaborative optimization across weather, energy and load centers is realized, so that dynamic balance between wind and light resource supply and data center demands is maintained in extreme weather, and the comprehensive energy efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent technology of wind and solar resources, and in particular relates to an intelligent matching method for wind and solar resources and data center load under extreme weather conditions. Background Technology

[0002] With the development of collaborative scheduling technology between renewable energy and high-energy-consuming data centers, integrated wind-solar-storage systems are gradually becoming a key means to improve green computing power. Traditional technologies often use single-factor static mapping models (such as simplifying wind speed and wind turbine output into fixed power curves) or offline fault tree analysis (based on historical data to statistically analyze the failure probability of a single device) to establish the correlation between meteorological conditions and energy supply.

[0003] However, in actual operation, under extreme weather event chains (such as typhoons accompanied by heavy rainfall, and sandstorms combined with high temperatures), the sudden change thresholds of various meteorological parameters (such as short-term wind speed surges and sudden increases in particulate matter concentration) can trigger cascading failures of multiple physical field equipment, such as wind turbine tower resonance, photovoltaic panel dust accumulation and efficiency collapse, and transmission line galloping. Traditional methods ignore the nonlinear correlation between meteorological sudden change thresholds and equipment failure mechanisms, and sever the multi-level transmission path of "meteorology-equipment-grid-load". This leads to problems such as supply and demand mismatch and local optimization but global failure in resource allocation strategies under extreme weather conditions, which seriously restricts the system's environmental adaptability and comprehensive energy efficiency. Summary of the Invention

[0004] Based on this, it is necessary to provide an intelligent matching method for wind and solar resources and data center load under extreme weather conditions to address the above-mentioned technical problems. This method can overcome the limitations of traditional single-factor static correlation analysis by dynamically modeling the transmission chain of meteorological sudden events and equipment failures, and achieve collaborative optimization across meteorology, energy and load centers. This will maintain a dynamic balance between the supply of wind and solar resources and the demand of data centers under extreme weather conditions, thereby improving overall energy efficiency.

[0005] Firstly, this application provides a method for intelligently matching wind and solar resources with data center load under extreme weather conditions, including:

[0006] The acquired meteorological change data, multi-physics field equipment status parameters, power grid operation parameters and load demand are spatiotemporally aligned to obtain a four-dimensional time series dataset;

[0007] The four-dimensional time series dataset is input into the dynamic prediction model to generate the cascade failure transmission factor matrix; the dynamic prediction model is used to characterize the relationship between the meteorological parameter mutation threshold and the equipment cascade failure transmission.

[0008] Using the cascade failure propagation factor matrix as a constraint, a multi-objective optimization solution is performed under spatiotemporal constraints to generate an elastic matching instruction set; the elastic matching instruction set is used to instruct the energy system to perform intelligent resource scheduling.

[0009] In one embodiment, a four-dimensional time-series dataset is input into a dynamic prediction model to generate a cascading failure propagation factor matrix, including:

[0010] A four-dimensional time-series dataset is input into a physical information neural network, and meteorological parameter mutation thresholds are predicted by embedding device physical equations.

[0011] Based on the spatial propagation path of meteorological abrupt change data, a cascaded failure hypernetwork model is constructed. The cascaded failure hypernetwork model is used to quantify the transmission effect of meteorological abrupt change data on the state of multiple physical devices, power grid power deficit and load demand.

[0012] Generate the cascade failure transmission factor matrix based on the cascade failure hypernetwork model.

[0013] In one embodiment, using the cascade failure propagation factor matrix as a constraint, multi-objective optimization is performed under spatiotemporal constraints to generate an elastic matching instruction set, including:

[0014] The spatiotemporal constraint parameters are extracted from the cascade failure transmission factor matrix. These parameters include the influence radius of the meteorological front and the transmission delay time window.

[0015] A multi-objective optimization algorithm is used to solve the multi-objective optimization function to generate a Pareto optimal solution set. The objectives of the multi-objective optimization function include maximizing the utilization rate of wind and solar resources, ensuring the satisfaction of load priority, and minimizing energy storage loss.

[0016] Feasibility verification of Pareto optimal solution set is performed based on digital twin system, and flexible matching instruction set is generated.

[0017] In one embodiment, a multi-objective optimization algorithm is used to solve the multi-objective optimization function to generate a Pareto optimal solution set, including:

[0018] Based on the spatiotemporal constraint parameters, the decision variables are spatially divided to obtain the search subspace, which includes the wind turbine output adjustment quantum space, the energy storage charging and discharging plan subspace, and the load migration ratio subspace.

[0019] Based on the search subspace, the differential evolution algorithm is used to generate the initial population;

[0020] Based on the dynamic adjustment of constraint violation weights using the conduction delay time window, the population evolution direction is optimized through the initial population to generate an intermediate population ranked by fitness score.

[0021] Non-dominated sorting and crowding calculation are performed on the intermediate population to screen feasible solution sets that satisfy the constraints of the multi-objective optimization function;

[0022] Merge feasible solution sets to generate Pareto optimal solution sets.

[0023] In one embodiment, the method further includes:

[0024] The elastic matching instruction set is analyzed and decomposed into multi-timescale control instructions based on the conduction delay time window. These multi-timescale control instructions include second-level energy storage charging and discharging instructions, minute-level load migration instructions, and hour-level equipment maintenance instructions.

[0025] Multi-timescale control commands are sent to each execution terminal of the energy system, and actual execution effect data is obtained;

[0026] Calculate the prediction error between the actual implementation effect data and the meteorological parameter mutation threshold;

[0027] Based on the prediction error, the parameters of the physical information neural network are updated through an online learning mechanism.

[0028] Secondly, this application also provides an intelligent matching device for wind and solar resources and data center load under extreme weather conditions, comprising:

[0029] The data alignment module is used to perform spatiotemporal alignment of the acquired meteorological change data, multi-physics field equipment status parameters, power grid operation parameters and load demand to obtain a four-dimensional time series dataset.

[0030] The dynamic prediction module is used to input a four-dimensional time series dataset into the dynamic prediction model to generate a cascade failure propagation factor matrix; the dynamic prediction model is used to characterize the relationship between meteorological parameter mutation thresholds and equipment cascade failure propagation.

[0031] The intelligent matching module is used to perform multi-objective optimization under spatiotemporal constraints with the cascaded failure propagation factor matrix as the constraint condition, and generate an elastic matching instruction set; the elastic matching instruction set is used to instruct the energy system to perform intelligent resource scheduling.

[0032] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-mentioned intelligent matching method for wind and solar resources and data center load under extreme weather conditions.

[0033] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for intelligent matching of wind and solar resources with data center load under extreme weather conditions.

[0034] The aforementioned intelligent matching method for wind and solar resources and data center loads under extreme weather scenarios constructs a four-dimensional time-series dataset (rather than traditional single-factor data fragmentation analysis) by spatiotemporally aligning meteorological mutation data, multi-physics equipment status parameters, power grid operating parameters, and load demand. This dataset overcomes the limitations of traditional single-factor data fragmentation analysis. The four-dimensional time-series dataset is input into a dynamic prediction model to generate a cascade failure propagation factor matrix, representing the relationship between meteorological parameter mutation thresholds and equipment cascade failure propagation. This dynamically quantifies the nonlinear impact path of extreme weather event chains on energy supply and load demand. Based on this, using the cascade failure propagation factor matrix as spatiotemporal constraints, a flexible matching instruction set is generated through multi-objective optimization. This coordinates the output of wind and solar resources, energy storage charging and discharging, and load priority allocation, thereby achieving a dynamic balance between wind and solar resource supply and data center demand under extreme weather conditions. Furthermore, it improves overall energy efficiency through a cross-system collaborative optimization mechanism. This solution addresses the technical shortcomings of traditional static correlation models, which struggle to adapt to meteorological mutations and multi-equipment cascade failures, through multi-dimensional dynamic modeling. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating a method for intelligent matching of wind and solar resources with data center load under extreme weather conditions, provided by an embodiment of the present invention.

[0037] Figure 2 A flowchart illustrating a multi-objective optimization intelligent matching method provided in an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of a device for intelligent matching of wind and solar resources with data center load under extreme weather conditions, provided in an embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0040] First, a brief introduction to the terms used in the embodiments of this application will be given.

[0041] Wind and solar resources refer to the collective term for wind energy and solar energy, which fall under the category of renewable energy. Specifically, they refer to environmental resources that can be captured and converted into electrical energy within a specific geographical area through natural meteorological conditions (such as wind speed and sunlight intensity). Wind energy relies on the kinetic energy generated by air movement to drive wind turbines to generate electricity, while solar energy converts solar radiation energy into electricity through the photovoltaic effect or photothermal conversion technology. Together, they constitute the core energy supply unit of a clean energy system.

[0042] Physics-Informed Neural Networks (PINNs) are machine learning models that combine deep learning with knowledge of physics. Their main principle is to use physical principles (usually expressed mathematically as partial differential equations) as prior knowledge, penalizing the residuals of these equations. During training, PINNs not only adjust based on the data but are also constrained by physical laws, ensuring that the learned results not only fit the data but also conform to physical laws.

[0043] The Hypernetwork Model is a network structure used to describe complex interactions between multi-level and multi-type nodes. Its core feature is to build cross-domain coupling transmission paths by integrating heterogeneous sub-networks (such as physical device networks, meteorological parameter networks, power grid topology networks, etc.).

[0044] Pareto optimality refers to the optimal equilibrium state in a multi-objective optimization problem when the system reaches a resource allocation state where it is impossible to further optimize a certain objective (such as improving the utilization rate of wind and solar resources) without harming other objectives (such as increasing energy storage losses or reducing load priority satisfaction) by adjusting any one objective variable. All feasible solutions at this point constitute the Pareto optimal solution set (Pareto Front). Essentially, it represents a trade-off between multiple conflicting objectives, ensuring that under the complex constraints of extreme weather, the wind and solar resource scheduling strategy can achieve a globally optimal balance of comprehensive energy efficiency, reliability, and economy, rather than a local optimum of a single objective.

[0045] Based on the above definitions, the implementation environment of the intelligent matching method for wind and solar resources and data center load under extreme weather conditions provided in this application embodiment will be described. Indicatively, this implementation environment includes: multimodal sensors, an execution terminal, and a processor. The processor, multimodal sensors, and execution terminal are connected via network signals. Multimodal sensors include, but are not limited to, vibration sensors, infrared thermal imagers, broadband voltage / frequency sensors, meteorological sensors, and smart meters. The processor can be a central processing unit, a multi-core processor, or an artificial intelligence chip, etc., and is not limited here.

[0046] Based on the above definitions and implementation environment, the application scenarios of the embodiments of this application are described. The intelligent matching method for wind and solar resources and data center load under extreme weather conditions provided in the embodiments of this application can be applied to scenarios including but not limited to the following:

[0047] In coastal areas prone to typhoons, strong winds, torrential rains, and sudden changes in air pressure can cause drastic fluctuations in wind turbine output. Traditional static scheduling models struggle to predict the resonance risk to turbine towers caused by sudden wind speed increases and the efficiency degradation caused by rainwater accumulation on photovoltaic panels. This solution uses meteorological sensors to capture real-time changes in wind speed gradients and rainfall intensity, combined with equipment status sensors to monitor mechanical stress in wind turbines and surface humidity on photovoltaic panels, dynamically constructing a transmission model between meteorological mutation thresholds and equipment health. Based on the cascading failure paths predicted by the hypernetic network model, the processor preemptively adjusts the charging and discharging strategies of the energy storage system, prioritizing continuous power supply to the core computing load of the data center while mitigating the risk of wind turbine overload shutdowns, achieving a stable match between wind and solar resources and high-priority loads during typhoons.

[0048] In arid regions prone to sandstorms, a sudden increase in the concentration of suspended particulate matter in the air can lead to rapid dust accumulation on the surface of photovoltaic panels, significantly reducing photoelectric conversion efficiency. Simultaneously, sand particles carried by strong winds can cause wear on wind turbine blades. This solution utilizes particulate matter concentration sensors to monitor the sandstorm diffusion trend in real time, dynamically tracks changes in the surface contamination of photovoltaic panels using infrared thermal imagers, and combines physical information neural networks to predict the power output attenuation curve caused by dust accumulation. Based on a cascaded failure conduction factor matrix, the processor dynamically prioritizes the cleaning of photovoltaic power plants, coordinates the scheduling of energy storage systems to compensate for power output shortfalls, and transfers non-urgent computing tasks to data centers in other areas via load migration commands, ensuring a dynamic balance between wind and solar resources and load demand during sandstorms.

[0049] In mountainous areas with frequent lightning activity, lightning strikes can instantly trigger protective shutdowns of wind and solar power plants. Simultaneously, sudden fluctuations in wind and solar power output caused by severe convective weather pose a threat to grid frequency stability. This solution uses electric field strength sensors and a lightning strike location system to capture the movement trajectory of thunderstorm fronts in real time. Combined with broadband grid sensors monitoring frequency deviations, a hypernetwork model is used to predict cascading outage paths caused by lightning strikes. The processor issues energy storage frequency regulation commands with a second-level latency to quickly smooth grid frequency fluctuations. Furthermore, it triggers millisecond-level task migration of data center loads through edge computing nodes, prioritizing the continuous operation of time-sensitive loads such as financial transactions and real-time AI inference, achieving elastic resource matching response during thunderstorms.

[0050] As an illustration, the intelligent matching method for wind and solar resources and data center load under extreme weather conditions provided in this application embodiment can also be applied to other application scenarios. This is only an example and does not limit the specific application scenarios.

[0051] In one exemplary embodiment, such as Figure 1 As shown, a method for intelligent matching of wind and solar resources with data center load under extreme weather conditions is provided. This embodiment illustrates the application of this method to an execution terminal in the aforementioned implementation environment. It is understood that this method can also be applied to servers, and can also be applied to systems including terminals and servers, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to 103:

[0052] Step 101: Spatiotemporally align the acquired meteorological change data, multiphysics equipment status parameters, power grid operation parameters, and load demand to obtain a four-dimensional time series dataset.

[0053] Specifically, meteorological abrupt change data (such as wind speed gradient change rate and particulate matter concentration abrupt change rate), multi-physics field equipment status parameters (such as wind turbine tower mechanical stress and photovoltaic panel surface contamination), power grid operation parameters (such as node voltage deviation and frequency fluctuation), and load demand data (such as computing power priority distribution and cooling power consumption) can be processed by timestamp alignment and spatial coordinate mapping to generate a four-dimensional time-series dataset of meteorology-equipment-power grid-load. For example, meteorological abrupt change data is collected by meteorological sensors deployed at wind and solar power plants; equipment status parameters are acquired through embedded strain gauges and thermal imagers; power grid parameters are read in real time through the power grid dispatch system interface; and load demand data is extracted through the data center energy management system. Timestamp alignment uses a GPS timing protocol to unify the sampling clock, and spatial mapping is based on a Geographic Information System (GIS) to associate the meteorological front movement trajectory with the equipment location coordinates. By eliminating the temporal asynchrony and spatial coverage deviation of multi-source data, a unified spatiotemporal benchmark dataset is constructed, providing highly consistent input for subsequent models and avoiding prediction errors caused by data misalignment.

[0054] Step 102: Input the four-dimensional time series dataset into the dynamic prediction model to generate the cascade failure transmission factor matrix; wherein, the dynamic prediction model is used to characterize the relationship between the meteorological parameter mutation threshold and the equipment cascade failure transmission.

[0055] Specifically, a four-dimensional time-series dataset is input into a pre-trained dynamic prediction model. For example, this model can employ a Long Short-Term Memory (LSTM) network architecture, trained using historical meteorological abrupt events and corresponding equipment failure records. It learns the intrinsic transmission relationship between meteorological parameter abruptness thresholds (such as a sudden increase in wind speed exceeding 20 m / s or a sharp drop in temperature exceeding 10 °C / h) triggering cascading equipment failures (such as wind turbine blade icing leading to an abnormal decrease in rotational speed, subsequently causing gearbox overload, generator power fluctuations, and a series of chain reactions). The cascading failure transmission factor matrix output by the model intuitively quantifies the mapping weights between meteorological abruptness feature vectors and equipment failure modes. For instance, when meteorological data indicates an impending extreme low temperature, the factors in the matrix related to the capacity decay of the battery energy storage system will significantly increase, providing an early warning of the impact of energy storage equipment performance degradation on system stability.

[0056] Step 103: Using the cascade failure propagation factor matrix as a constraint, perform multi-objective optimization under spatiotemporal constraints to generate a flexible matching instruction set; wherein, the flexible matching instruction set is used to instruct the energy system to perform intelligent resource scheduling.

[0057] Specifically, using the cascaded failure propagation factor matrix as a key constraint, multiple variables such as wind and solar power supply, grid transmission capacity, data center load demand, and energy storage equipment status are incorporated into the objective function to construct a multi-objective optimization model that includes temporal continuity constraints (ensuring smooth system state transition) and spatial distribution constraints (considering the coordination of energy equipment in different geographical locations). The model is solved using a multi-objective evolutionary algorithm to obtain a set of Pareto optimal solutions. The optimal solution combination is further selected and transformed into a specific elastic matching instruction set. This instruction set can cover specific operational commands such as adjusting the wind turbine pitch angle, setting the output value of the photovoltaic inverter, charging and discharging power commands for energy storage batteries, and adjusting the operating frequency of data center chiller units. These commands are sent to each execution unit in real time through the energy management system, driving the energy system to implement intelligent resource scheduling based on the dynamically changing supply and demand relationship under extreme weather conditions.

[0058] The aforementioned intelligent matching method for wind and solar resources and data center loads under extreme weather scenarios constructs a four-dimensional time-series dataset (rather than traditional single-factor data fragmentation analysis) by spatiotemporally aligning meteorological mutation data, multi-physics equipment status parameters, power grid operating parameters, and load demand. This dataset overcomes the limitations of traditional single-factor data fragmentation analysis. The four-dimensional time-series dataset is input into a dynamic prediction model to generate a cascade failure propagation factor matrix, representing the relationship between meteorological parameter mutation thresholds and equipment cascade failure propagation. This dynamically quantifies the nonlinear impact path of extreme weather event chains on energy supply and load demand. Based on this, using the cascade failure propagation factor matrix as spatiotemporal constraints, a flexible matching instruction set is generated through multi-objective optimization. This coordinates the output of wind and solar resources, energy storage charging and discharging, and load priority allocation, thereby achieving a dynamic balance between wind and solar resource supply and data center demand under extreme weather conditions. Furthermore, it improves overall energy efficiency through a cross-system collaborative optimization mechanism. This solution addresses the technical shortcomings of traditional static correlation models, which struggle to adapt to meteorological mutations and multi-equipment cascade failures, through multi-dimensional dynamic modeling.

[0059] In one embodiment, a four-dimensional time-series dataset is input into a dynamic prediction model to generate a cascading failure propagation factor matrix, including:

[0060] The four-dimensional time-series dataset is input into the physical information neural network, and the threshold for meteorological parameter mutations is predicted by embedding the device physical equation.

[0061] Specifically, by embedding equipment physical equations such as wind turbine aerodynamic equations and photovoltaic panel thermodynamic equations during neural network training, the threshold for meteorological parameter mutations leading to cascading equipment failure can be predicted. For example, the neural network output can be constrained by the differential equation of wind turbine tower vibration to predict the critical failure threshold of wind turbine mechanical stress under different wind speed gradients; or the threshold for a sudden drop in photovoltaic output caused by a sudden change in particulate matter concentration can be predicted based on the correlation equation between photovoltaic panel surface contamination and photoelectric conversion efficiency. Furthermore, a transfer learning strategy can be adopted to pre-train the network using historical normal weather data and fine-tune it with small sample extreme weather data to improve the accuracy of mutation threshold prediction. By fusing data-driven approaches and equipment mechanisms through physical information neural networks, the limitations of traditional models in predicting extreme weather data sparse scenarios can be overcome, accurately identifying meteorological mutation critical points and avoiding equipment overload or power outages due to threshold misjudgments.

[0062] Based on the spatial propagation path of meteorological abrupt change data, a cascaded failure hypernetwork model is constructed. The cascaded failure hypernetwork model is used to quantify the transmission effect of meteorological abrupt change data on the state of multiple physical devices, power grid power deficit, and load demand.

[0063] Specifically, the spatial propagation path of meteorological abrupt changes can be analyzed using Geographic Information Systems (GIS) to determine the movement trajectory of meteorological events (such as blizzards and sandstorms) across terrain and their sequential impact on energy equipment (such as wind turbines and photovoltaic power plants) distributed along the path. The cascaded failure hypernetwork model treats meteorological abrupt changes, the states of multiple physical devices, grid power deficits, and load demands as nodes, and their interactions as edges, thereby quantifying the transmission effect of meteorological abrupt changes on each element. For example, the model can represent the complete transmission chain from a sudden drop in temperature leading to capacity decay in battery energy storage systems, which in turn causes grid frequency fluctuations, ultimately affecting the power supply quality of data center loads. This model can be constructed using a graph neural network (GNN) architecture, defining node features and edge weights to characterize complex interactions.

[0064] Generate the cascade failure transmission factor matrix based on the cascade failure hypernetwork model.

[0065] Specifically, a cascaded failure transmission factor matrix is ​​generated based on the constructed cascaded failure hypernetwork model. The elements of this matrix represent the quantified mapping weights between meteorological abrupt change feature vectors and equipment failure modes, power grid deficits, and load fluctuations. For example, an element in the matrix might represent the sensitivity coefficient of a sudden drop in temperature to the power output reduction of a certain type of wind turbine. Exemplarily, methods for generating this matrix include performing a graph traversal algorithm on the hypernetwork model to calculate the cumulative path weights between nodes, or using matrix factorization techniques to extract core transmission factors. Through the above technical solutions, this embodiment achieves accurate modeling of complex cascading failure processes caused by meteorological abrupt events. The introduction of physical information neural networks enables the model to integrate the physical characteristics of equipment with the advantages of data-driven approaches, which not only improves the accuracy of meteorological parameter abrupt change threshold prediction but also reduces the dependence on massive amounts of labeled data. The cascading failure hypernetwork model breaks through the limitations of traditional single-device analysis, quantifying the comprehensive transmission effect of meteorological abrupt changes on multi-physics field equipment clusters, power grid stability, and load demand at the system level, providing a global perspective for the preventive control of energy systems. The generated cascading failure transmission factor matrix has high interpretability, providing clear quantitative basis for subsequent optimized scheduling, enabling energy systems to adjust resource allocation strategies in a targeted manner before meteorological disasters occur, such as adjusting wind turbine pitch angles in advance and optimizing energy storage charging and discharging plans, thereby effectively maintaining the dynamic balance of energy supply and demand under extreme weather conditions and improving the resilience and overall energy efficiency of the system.

[0066] like Figure 2 As shown, in one embodiment, using the cascade failure propagation factor matrix as a constraint, multi-objective optimization is performed under spatiotemporal constraints to generate an elastic matching instruction set, including:

[0067] Step 201: Extract the spatiotemporal constraint parameters from the cascade failure transmission factor matrix. The spatiotemporal constraint parameters include the influence radius of the meteorological front and the transmission delay time window.

[0068] Specifically, the influence radius of the meteorological front and the transmission delay time window are extracted as spatiotemporal constraint parameters from the cascaded failure transmission factor matrix. The influence radius of the meteorological front is determined by mapping the spatial relationship between the meteorological front's movement trajectory and the equipment location using a geographic information system, such as the set of wind turbine coordinates within the typhoon eyewall's influence area. The transmission delay time window is determined by analyzing the distribution of time-series delay parameters in the matrix, such as the average response time from a sudden change in wind speed gradient to wind turbine stress exceeding limits. This method can accurately extract the spatiotemporal constraint boundaries of meteorological changes on equipment and the power grid, providing quantifiable inputs for multi-objective optimization and avoiding strategy failures due to constraint omissions.

[0069] Step 202: The multi-objective optimization function is solved using a multi-objective optimization algorithm to generate a Pareto optimal solution set; wherein, the objectives of the multi-objective optimization function include maximizing the utilization rate of wind and solar resources, ensuring the satisfaction of load priority, and minimizing energy storage loss.

[0070] Specifically, a multi-objective function can be constructed with the optimization objectives of maximizing wind and solar resource utilization, ensuring load priority satisfaction, and minimizing energy storage losses. This function can then be solved using an improved non-dominated sorting genetic algorithm (NSGA-II). The search space of decision variables is decomposed according to spatiotemporal constraint parameters. For example, wind turbine output adjustment, energy storage charging and discharging plans, and load migration ratios are divided into independent subspaces. Furthermore, a dynamic constraint weight mechanism can be combined to prioritize satisfying high-urgency transmission delay constraints. Through spatiotemporal constraint decomposition and dynamic weight adjustment, the Pareto optimal solution set globally balances multi-objective conflicts, achieving a balance between resource utilization efficiency and energy supply reliability under extreme weather conditions, outperforming traditional single-objective optimization methods.

[0071] Step 203: Verify the feasibility of the Pareto optimal solution set based on the digital twin system, and generate a flexible matching instruction set.

[0072] Specifically, the Pareto optimal solution set is input into the digital twin system for dynamic simulation verification, simulating the execution effects of different scheduling strategies under real-time weather front movement and equipment status changes. For example, it verifies the feasibility of energy storage charging and discharging commands under grid frequency mutation scenarios, or tests the executability of load migration ratios under cooling system power limitations. Solutions that violate physical constraints or exceed the execution terminal's capabilities are iteratively corrected and eliminated to generate the final elastic matching command set. Through a closed-loop verification mechanism involving virtual and real interaction, the executability and stability of the elastic command set under complex extreme weather scenarios are ensured, providing a highly reliable solution for intelligent matching of wind and solar resources and loads under extreme weather conditions.

[0073] In one embodiment, a multi-objective optimization algorithm is used to solve the multi-objective optimization function to generate a Pareto optimal solution set, including:

[0074] The decision variables are spatially partitioned based on the spatiotemporal constraint parameters to obtain the search subspace, which includes the wind turbine output adjustment quantum space, the energy storage charging and discharging plan subspace, and the load migration ratio subspace.

[0075] Specifically, the decision variable space refers to the set of key controllable parameters that need to be adjusted during multi-objective optimization. The search subspace is obtained by spatially partitioning the decision variables according to spatiotemporal constraint parameters. The meteorological frontal influence radius and conduction delay time window in the spatiotemporal constraint parameters can be used to determine the reasonable range of the decision variables. For example, the meteorological frontal influence radius can limit the magnitude of wind turbine output adjustment, and the conduction delay time window can limit the time range of energy storage charging and discharging plans. Based on these parameters, the decision variable space is divided into a wind turbine output adjustment quantum space, an energy storage charging and discharging plan subspace, and a load migration ratio subspace. The wind turbine output adjustment quantum space determines the adjustment range based on the wind turbine's rated power and meteorological conditions; the energy storage charging and discharging plan subspace formulates charging and discharging plans based on the capacity and current state of charge of the energy storage equipment; and the load migration ratio subspace determines the migration range based on the transferability characteristics of the data center load and business priorities. Methods for implementing this partitioning include, but are not limited to, using clustering algorithms to classify the decision variables, or directly dividing intervals based on equipment operating constraints.

[0076] Based on the search subspace, the differential evolution algorithm is used to generate the initial population.

[0077] Specifically, based on the predefined search subspace, a differential evolution algorithm is used to generate the initial population. The differential evolution algorithm utilizes the boundary conditions of the search subspace to randomly initialize individuals, with each individual representing a combination of decision variable values. For example, multiple wind turbine output adjustment values ​​are randomly generated in the wind turbine output adjustment quantum space, energy storage charging and discharging power sequences are generated in the energy storage charging and discharging plan subspace, and load allocation ratios are generated in the load migration ratio subspace. In this way, the initial population can cover all regions of the search space, providing a diverse foundation for subsequent optimization. The initial population can also be generated using uniform sampling techniques such as Latin hypercube sampling to improve the initial distribution uniformity of the population.

[0078] By dynamically adjusting the constraint violation weights based on the conduction delay time window, the evolutionary direction of the population is optimized through the initial population, and an intermediate population ranked by fitness score is generated.

[0079] Specifically, the constraint violation weights are dynamically adjusted based on the conduction delay time window, and the evolutionary direction of the initial population is optimized to generate an intermediate population ranked by fitness scores. During the optimization process, the importance of constraints varies at different stages. For example, in the early stages of a meteorological anomaly, the constraint of grid power deficit may be more critical, while in the later stages, the constraint of energy storage loss may be more important. Therefore, the constraint violation weights are dynamically adjusted according to the conduction delay time window, enabling the algorithm to better balance constraints and the objective function at different stages. Individuals in the initial population are ranked according to their fitness scores, i.e., the degree to which they satisfy the objective function (maximizing wind and solar resource utilization, ensuring load priority satisfaction, and minimizing energy storage loss). Individuals with higher fitness scores are more likely to become part of the intermediate population. Methods for implementing fitness scoring include, but are not limited to, calculating the objective function value of an individual and comprehensively considering the degree of constraint violation.

[0080] Non-dominated sorting and crowding calculation are performed on the intermediate population to screen feasible solution sets that satisfy the constraints of the multi-objective optimization function.

[0081] Specifically, non-dominated sorting and crowding calculation are performed on the intermediate population to screen feasible solutions that satisfy the constraints of the multi-objective optimization function. Non-dominated sorting stratifies individuals in the intermediate population according to their relative merits within the objective function; individuals in the same stratum have the same non-dominated level. For example, individual A is better than individual B in terms of wind and solar resource utilization and load priority satisfaction, but slightly worse in terms of energy storage loss; therefore, individuals A and B belong to the same non-dominated level. Crowding calculation measures the distribution density of individuals in the objective function space, ensuring solution diversity. Through non-dominated sorting and crowding calculation, feasible solutions that satisfy the constraints are screened. Furthermore, the non-dominated sorting method can employ a fast non-dominated sorting algorithm, and the crowding calculation can use a neighborhood distance-based method.

[0082] Merge feasible solution sets to generate Pareto optimal solution sets.

[0083] Specifically, the selected feasible solution set is merged, and duplicate or similar solutions are removed to obtain a set of solutions that are mutually exclusive on the objective function, i.e., the Pareto optimal solution set. These solutions represent the optimal trade-offs in multi-objective optimization problems, providing multiple options for energy system scheduling. Merging methods include, but are not limited to, calculating the similarity between solutions and removing solutions with excessive similarity, or selecting representative solutions based on the distribution of objective function values. This embodiment, by dividing the search subspace according to spatiotemporal constraint parameters, can accurately limit the reasonable range of decision variables, improve search efficiency, and avoid wasting computational resources in invalid or unreasonable decision variable spaces. The differential evolution algorithm is used to generate the initial population, which can fully utilize its global search capability and ease of implementation to quickly cover the search space, providing a foundation for finding high-quality solutions. Dynamically adjusting the constraint violation weights based on the propagation delay time window allows the optimization process to flexibly respond to different stages of extreme weather event development, improving the adaptability and practicality of the solution.

[0084] In one embodiment, the intelligent matching method for wind and solar resources and data center load under extreme weather conditions provided in this application further includes the following steps:

[0085] The elastic matching instruction set is analyzed and decomposed into multi-timescale control instructions based on the conduction delay time window. These multi-timescale control instructions include second-level energy storage charging and discharging instructions, minute-level load migration instructions, and hour-level equipment maintenance instructions.

[0086] Specifically, based on the conduction delay time window in the cascaded failure conduction factor matrix, the elastic matching instruction set is decomposed into second-level energy storage charging and discharging instructions, minute-level load migration instructions, and hour-level equipment maintenance instructions according to response timeliness. Furthermore, the batch execution time periods of load migration instructions can be dynamically divided based on a sliding window algorithm to avoid grid impact caused by concentrated operations. This multi-timescale decomposition mechanism balances the timeliness and executability of resource scheduling under extreme weather conditions, improving the dynamic matching accuracy between wind and solar power output fluctuations and load demand fluctuations.

[0087] Multi-timescale control commands are sent to each execution terminal of the energy system, and actual execution effect data is obtained.

[0088] Specifically, multi-timescale control commands can be sent to various execution terminals within the energy system, and actual execution effect data can be obtained. Execution terminals can include energy storage inverters, data center load balancers, and control systems for wind turbines and photovoltaic power plants. For example, after a second-level energy storage charge / discharge command is sent to the energy storage inverter, the execution effect is evaluated by collecting data such as the inverter's output current, voltage, and the state of charge of the energy storage battery; after a minute-level load migration command is sent to the data center load balancer, the validity of the command is judged based on indicators such as changes in server power consumption and service response time; after an hour-level equipment maintenance command is executed, the maintenance operation is verified to have achieved the expected results through information such as equipment operation logs and fault alarm records. Data acquisition methods can include using an Internet of Things (IoT) sensor network to collect equipment operation data in real time, or using the equipment's own monitoring interface to periodically upload operating status information.

[0089] Calculate the prediction error between the actual implementation effect data and the meteorological parameter mutation threshold.

[0090] Specifically, the prediction error is calculated between the actual performance data and the meteorological parameter mutation threshold. The predicted data for the meteorological parameter mutation threshold comes from the output of the physical information neural network, while the actual performance data reflects the actual performance of the energy system in responding to meteorological mutations. For example, if the prediction is that the wind speed in a certain area will exceed 15 m / s within the next 10 minutes, triggering the threshold for wind turbine power reduction, but the actual performance data shows that the wind turbine power only decreased by 80% of the predicted value, then the difference between the two is the prediction error. Methods for calculating the prediction error include using statistical indicators such as mean squared error (MSE) and mean absolute error (MAE) to quantify the deviation between the predicted and actual values.

[0091] Based on the prediction error, the parameters of the physical information neural network are updated through an online learning mechanism.

[0092] Specifically, based on the prediction error, the parameters of the physical information neural network are updated through an online learning mechanism. This online learning mechanism allows the network to adjust its parameters in real time to adapt to new data inputs without stopping operation. For example, when a large prediction error is calculated, the gradient of the loss function with respect to the network parameters is calculated using the backpropagation algorithm, and the weights and biases are updated according to a preset learning rate. Online learning can be implemented using an incremental learning strategy, updating parameters sample by sample, or using mini-batch online learning, periodically collecting a batch of new data for parameter adjustment. Furthermore, to maintain the physical consistency of the physical information neural network, the constraints of the embedded device physical equations must still be satisfied during the update process. This can be achieved by adding physical constraint terms to the loss function or using optimization methods such as projective gradient descent. The above technical solution matches the responsiveness of the execution terminal through a multi-timescale decomposition mechanism, avoiding instruction backlog or resource conflicts. Combined with the online learning mechanism, the prediction model dynamically tracks changes in weather and equipment status, maintaining the long-term reliability of the scheduling strategy and ensuring the continuous optimization capability of the intelligent resource matching method under extreme weather conditions.

[0093] In summary, the intelligent matching method for wind and solar resources and data center load under extreme weather conditions provided in this application collects meteorological mutation parameters, multi-physics equipment status parameters, power grid operation parameters, and load demand data in real time using multi-modal sensors, and constructs a four-dimensional time-series dataset through spatiotemporal alignment. This dataset is then input into a physical information neural network and a cascaded failure hypernetwork model to dynamically predict meteorological parameter mutation thresholds and quantify their transmission effects on equipment status, power grid power deficit, and load demand, generating a transmission factor matrix characterizing cross-system cascaded failure paths. Using this matrix as spatiotemporal constraints, a multi-objective optimization algorithm is employed to solve for a Pareto optimal solution set that considers wind and solar resource utilization, load priority, and energy storage loss. After digital twin verification, a flexible matching instruction set is generated. By decomposing the instructions into multi-timescale control strategies and feeding back actual execution data, the physical information neural network is driven to update model parameters online. This technical method overcomes the limitations of traditional single-factor static correlation analysis by dynamically modeling the transmission chain of meteorological mutation events and equipment failures, maintaining a dynamic balance between wind and solar resource supply and data center demand under extreme weather conditions, while simultaneously improving overall energy efficiency and the stability of supply-demand balance.

[0094] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0095] Based on the same inventive concept, this application also provides an intelligent matching device 10 for wind and solar resources and data center load under extreme weather conditions, used to implement the intelligent matching method for wind and solar resources and data center load under extreme weather conditions described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the intelligent matching device 10 for wind and solar resources and data center load under extreme weather conditions provided below can be found in the limitations of the intelligent matching method for wind and solar resources and data center load under extreme weather conditions described above, and will not be repeated here.

[0096] In one exemplary embodiment, such as Figure 3 As shown, a smart matching device 10 for wind and solar resources and data center load under extreme weather conditions is provided, comprising:

[0097] The data alignment module 11 is used to perform spatiotemporal alignment of the acquired meteorological change data, multi-physics field equipment status parameters, power grid operation parameters and load demand to obtain a four-dimensional time series dataset.

[0098] The dynamic prediction module 12 is used to input the four-dimensional time series dataset into the dynamic prediction model to generate the cascade failure transmission factor matrix; wherein, the dynamic prediction model is used to characterize the relationship between the meteorological parameter mutation threshold and the equipment cascade failure transmission.

[0099] The intelligent matching module 13 is used to perform multi-objective optimization under spatiotemporal constraints with the cascaded failure propagation factor matrix as a constraint, and generate an elastic matching instruction set; wherein, the elastic matching instruction set is used to instruct the energy system to perform intelligent resource scheduling.

[0100] In one embodiment, the dynamic prediction module 12 includes:

[0101] The physical information neural network unit is used to input a four-dimensional time-series dataset into the physical information neural network and predict the threshold for meteorological parameter mutations by embedding the device's physical equations.

[0102] The cascaded failure supernetwork building unit is used to construct a cascaded failure supernetwork model based on the spatial propagation path of meteorological mutation data. The cascaded failure supernetwork model is used to quantify the transmission effect of meteorological mutation data on the state of multiple physical devices, power grid power deficit, and load demand.

[0103] The transmission factor matrix generation unit is used to generate the cascade failure transmission factor matrix based on the cascade failure hypernetwork model.

[0104] In one embodiment, the intelligent matching module 13 includes:

[0105] The constraint parameter extraction unit is used to extract the spatiotemporal constraint parameters in the cascade failure transmission factor matrix. The spatiotemporal constraint parameters include the influence radius of the meteorological front and the transmission delay time window.

[0106] The multi-objective optimization unit is used to solve the multi-objective optimization function using a multi-objective optimization algorithm to generate a Pareto optimal solution set. The objectives of the multi-objective optimization function include maximizing the utilization rate of wind and solar resources, ensuring the satisfaction of load priority, and minimizing energy storage loss.

[0107] The digital twin verification unit is used to verify the feasibility of Pareto optimal solution sets based on the digital twin system and generate a flexible matching instruction set.

[0108] In one embodiment, the multi-objective optimization unit includes:

[0109] The spatial sub-unit is used to divide the decision variables into search subspaces based on spatiotemporal constraint parameters. The search subspaces include the wind turbine output adjustment quantum space, the energy storage charging and discharging plan subspace, and the load migration ratio subspace.

[0110] Differential evolution subunits are used to generate an initial population based on a search subspace using a differential evolution algorithm.

[0111] The dynamic weight adjustment subunit is used to dynamically adjust the constraint violation weights based on the propagation delay time window, optimize the population evolution direction through the initial population, and generate an intermediate population ranked by fitness score.

[0112] The non-dominated sorting subunit is used to perform non-dominated sorting and crowding calculation on the intermediate population, and to screen the feasible solution set that satisfies the constraints of the multi-objective optimization function.

[0113] The solution set is merged into sub-units to combine feasible solution sets and generate Pareto optimal solution sets.

[0114] In one embodiment, the intelligent matching device 10 for wind and solar resources and data center load under extreme weather conditions further includes:

[0115] The instruction decomposition unit is used to parse the elastic matching instruction set. Based on the conduction delay time window, the elastic matching instruction set is decomposed into multi-timescale control instructions, which include second-level energy storage charging and discharging instructions, minute-level load migration instructions, and hour-level equipment maintenance instructions.

[0116] The instruction issuing unit is used to send multi-timescale control instructions to various execution terminals of the energy system and obtain actual execution effect data.

[0117] The error calculation unit is used to calculate the prediction error between the actual execution effect data and the meteorological parameter mutation threshold.

[0118] An online update unit is used to update the parameters of the physical information neural network based on the prediction error through an online learning mechanism.

[0119] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the intelligent matching method for wind and solar resources and data center load under extreme weather conditions as described above.

[0120] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0121] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units 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 disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0122] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for intelligently matching wind and solar resources with data center load under extreme weather conditions, characterized in that, The method includes: The acquired meteorological change data, multi-physics field equipment status parameters, power grid operation parameters and load demand are spatiotemporally aligned to obtain a four-dimensional time series dataset; The four-dimensional time-series dataset is input into the dynamic prediction model to generate a cascade failure propagation factor matrix; wherein, the dynamic prediction model is used to characterize the relationship between meteorological parameter mutation threshold and equipment cascade failure propagation. Using the cascaded failure propagation factor matrix as a constraint, a multi-objective optimization solution under spatiotemporal constraints is performed to generate an elastic matching instruction set; wherein, the elastic matching instruction set is used to instruct the energy system to perform intelligent resource scheduling.

2. The method according to claim 1, characterized in that, The step of inputting the four-dimensional time-series dataset into the dynamic prediction model to generate the cascading failure propagation factor matrix includes: The four-dimensional time-series dataset is input into a physical information neural network, and the meteorological parameter mutation threshold is predicted by embedding the device physical equation. Based on the spatial propagation path of the meteorological abrupt change data, a cascaded failure supernetwork model is constructed. The cascaded failure supernetwork model is used to quantify the transmission effect of meteorological abrupt change data on the state of multiple physical devices, power grid power deficit, and load demand. The cascade failure transmission factor matrix is ​​generated based on the cascade failure hypernetwork model.

3. The method according to claim 2, characterized in that, The step of performing multi-objective optimization under spatiotemporal constraints, using the cascaded failure propagation factor matrix as a constraint, to generate an elastic matching instruction set includes: Extract the spatiotemporal constraint parameters from the cascaded failure transmission factor matrix. The spatiotemporal constraint parameters include the influence radius of the meteorological front and the transmission delay time window. A multi-objective optimization algorithm is used to solve the multi-objective optimization function to generate a Pareto optimal solution set; wherein, the objectives of the multi-objective optimization function include maximizing the utilization rate of wind and solar resources, ensuring the satisfaction of load priority, and minimizing energy storage loss; The feasibility of the Pareto optimal solution set is verified based on the digital twin system, and the elastic matching instruction set is generated.

4. The method according to claim 3, characterized in that, The process of using a multi-objective optimization algorithm to solve the multi-objective optimization function and generate a Pareto optimal solution set includes: Based on the spatiotemporal constraint parameters, the decision variables are spatially divided to obtain a search subspace, which includes a wind turbine output adjustment quantum space, an energy storage charging and discharging plan subspace, and a load migration ratio subspace. Based on the search subspace, an initial population is generated using a differential evolution algorithm; Based on the conduction delay time window, the constraint violation weight is dynamically adjusted, and the population evolution direction is optimized through the initial population to generate an intermediate population ranked by fitness score. The intermediate population is subjected to non-dominated sorting and crowding calculation to screen feasible solution sets that satisfy the constraints of the multi-objective optimization function; The feasible solution sets are merged to generate the Pareto optimal solution set.

5. The method according to claim 3, characterized in that, The method further includes: The elastic matching instruction set is analyzed, and according to the conduction delay time window, the elastic matching instruction set is decomposed into multi-time scale control instructions, wherein the multi-time scale control instructions include second-level energy storage charging and discharging instructions, minute-level load migration instructions, and hour-level equipment maintenance instructions. The multi-timescale control commands are sent to each execution terminal of the energy system, and the actual execution effect data is obtained. Calculate the prediction error between the actual execution effect data and the meteorological parameter mutation threshold; Based on the prediction error, the parameters of the physical information neural network are updated through an online learning mechanism.

6. A device for intelligent matching of wind and solar resources with data center load under extreme weather conditions, characterized in that, The device includes: The data alignment module is used to perform spatiotemporal alignment of the acquired meteorological change data, multi-physics field equipment status parameters, power grid operation parameters and load demand to obtain a four-dimensional time series dataset. The dynamic prediction module is used to input the four-dimensional time series dataset into the dynamic prediction model to generate a cascade failure transmission factor matrix; wherein, the dynamic prediction model is used to characterize the relationship between meteorological parameter mutation threshold and equipment cascade failure transmission. The intelligent matching module is used to perform multi-objective optimization under spatiotemporal constraints using the cascaded failure transmission factor matrix as a constraint, and generate an elastic matching instruction set; wherein, the elastic matching instruction set is used to instruct the energy system to perform intelligent resource scheduling.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images