A clean energy driven computing power and power collaborative regulation method

Abstract

The application relates to the technical field of algorithm and energy coordination, and discloses a clean energy driven algorithm and electric power coordination regulation method, which comprises the following steps: collecting clean energy power generation data, algorithm demand data and electric power system load data; performing multi-dimensional clean energy, algorithm and electric power coordination analysis and processing based on the clean energy power generation data, the algorithm demand data and the electric power system load data, to generate clean energy-algorithm-electric power coordination characteristic data; when clean energy driven algorithm and electric power coordination regulation is performed, clean energy power generation data, algorithm demand data and electric power system load data are collected by using a unified standard, so that the standardization and comparability of data from different sources are ensured; the coupling relationship between clean energy volatility and algorithm elastic demand is analyzed in real time based on a multi-dimensional coordination analysis model, the forward-looking and accuracy of clean energy driven algorithm and electric power resource coordination regulation are ensured, and the overall energy efficiency of the system is improved.

Description

Technical Field

[0001] This invention relates to the field of computing power coordination technology, specifically a method for coordinating and regulating computing power and electricity driven by clean energy. Background Technology

[0002] As climate change intensifies and fossil fuels are depleted, the global energy structure is accelerating its transformation towards cleaner and lower-carbon energy. Clean energy, including solar, wind, hydro, and geothermal energy, has become a key tool for achieving carbon reduction goals due to its zero carbon emissions and renewable resources.

[0003] Currently, in the process of coordinating the regulation of computing power and electricity driven by clean energy, due to the intermittent and fluctuating nature of clean energy power generation and the sudden and elastic characteristics of computing power demand, existing methods are unable to effectively analyze the dynamic coupling relationship between the two, resulting in low accuracy of coordinating feature extraction and the inability of the generated regulation strategy to take into account multiple objectives such as energy consumption, computing power utilization and power load stability.

[0004] Therefore, a clean energy-driven method for the coordinated regulation of computing power and electricity is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a clean energy-driven method for the coordinated regulation of computing power and electricity, which solves the problem mentioned in the background technology of deviation in the extraction of coordinated features affecting the generation accuracy of subsequent regulation strategies.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for coordinated control of computing power and electricity driven by clean energy, the method comprising the following steps: S1. Collect clean energy power generation data, computing power demand data, and power system load data; S2. Based on the clean energy power generation data, the computing power demand data, and the power system load data, perform multi-dimensional collaborative analysis and processing of clean energy, computing power, and electricity to generate clean energy-computing power-electricity collaborative feature data. S3. Based on the clean energy-computing power-electricity synergistic characteristic data, perform clean energy volatility and computing power elasticity demand matching processing to generate clean energy-driven computing power regulation strategy data; S4. Based on the clean energy-driven computing power control strategy data, perform dynamic load balance optimization processing of the power system to generate computing power and power coordinated control instruction data; S5. Execute the computing power and power coordinated control command data to form real-time coordinated control of computing power resources and power resources driven by clean energy.

[0007] Preferably, the collection of clean energy power generation data, computing power demand data, and power system load data in step S1 includes the following steps: S11. Use clean energy power generation monitoring equipment to collect real-time power generation data, power generation forecast data and weather impact parameters of solar power plants, wind power plants and hydropower plants, and generate clean energy power generation data. S12. Use the computing power resource monitoring platform to collect real-time computing task queues, processor utilization, memory usage and energy efficiency ratio data of cloud computing centers, edge computing nodes and high-performance computing clusters, and generate computing power demand data. S13. Use the power system SCADA platform to collect regional power grid load curves, distribution network power flow data, energy storage device charging and discharging status and power grid stability indicators, and generate power system load data; S14. Perform time series alignment and data standardization on the clean energy power generation data, the computing power demand data, and the power system load data to generate multi-source time series collaborative data. The data standardization uses the Z-score normalization algorithm to eliminate dimensional differences.

[0008] Preferably, the multi-dimensional collaborative analysis process in S2 includes the following steps: S21. Obtain the multi-source time-series collaborative data; S22. The principal component analysis algorithm is used to reduce the dimensionality of the multi-source time-series collaborative data, extract the collaborative principal components of clean energy power generation fluctuation characteristics, computing power demand elasticity characteristics and power load cycle characteristics, and generate a dimensionality-reduced feature matrix. S23. Based on the reduced-dimensional feature matrix, the gray relational analysis algorithm is applied to calculate the dynamic correlation between clean energy power generation, computing power demand and power load, and generate clean energy-computing power-power correlation data. S24. Based on the clean energy-computing power-electricity correlation data, construct a clean energy-computing power-electricity collaborative analysis model and output clean energy-computing power-electricity collaborative characteristic data. The collaborative analysis model uses multiple linear regression to fit the coupling relationship between computing power and electricity driven by clean energy.

[0009] Preferably, the process of matching clean energy volatility with computing power elasticity demand in S3 includes the following steps: S31. Based on the clean energy-computing power-electricity collaborative characteristic data, a time series prediction algorithm is used to predict the clean energy power generation fluctuation curve and the computing power demand elasticity curve, and clean energy fluctuation prediction data and computing power demand prediction data are generated. S32. Apply fuzzy logic control algorithm to calculate the matching degree of the clean energy fluctuation prediction data and the computing power demand prediction data, and generate a clean energy and computing power demand matching degree index. S33. Based on the matching degree index and combined with the power system stability constraints, a multi-objective optimization algorithm is used to generate clean energy-driven computing power regulation strategy data, wherein the multi-objective optimization algorithm takes minimizing energy waste, maximizing computing power utilization, and minimizing power load fluctuation variance as objective functions.

[0010] Preferably, the time series prediction algorithm in S31 adopts a long short-term memory network model, the mathematical expression of which is: ; in, For input data, For the previous moment The hidden state vector, In hidden state, and For model parameters, For the Gate of Oblivion For the previous moment The cell state vector, Let be the activation vector of the forget gate. Let be the bias vector of the forget gate. Let be the activation vector of the input gate. This is the current candidate cell state vector. and The input gate weight matrix, and This is the input gate bias vector. Let be the activation vector of the output gate. This is the output gate weight matrix. This is the bias vector for the output gate. It is the Sigmoid activation function. The hyperbolic tangent activation function is used. The hidden state from the previous moment Input at the current moment Perform vector concatenation.

[0011] Preferably, the power system dynamic load balancing optimization process in S4 includes the following steps: S41. Obtain the clean energy-driven computing power regulation strategy data; S42. Using the model predictive control algorithm, with the clean energy-driven computing power regulation strategy data as input, and combined with the real-time load data of the power system, dynamically optimize the allocation of computing power resources and the power dispatch plan to generate preliminary regulation instruction data. S43. Apply reinforcement learning algorithm to correct the preliminary control command data online, adjust the priority of computing power tasks and the charging and discharging strategy of energy storage equipment based on power system feedback data, and generate optimized control command data. S44. The optimized control command data is compared with the historical control effect data, and the Kalman filter algorithm is used to eliminate noise interference to generate computing power and power coordinated control command data.

[0012] Preferably, the reinforcement learning algorithm in S43 employs a deep network, and its objective function is: ; in, Indicates the system status. Indicates regulatory action, As a reward value, As a discount factor, This represents the estimated maximum future value. Indicates a new state. It indicates the value of a state and an action.

[0013] Preferably, the real-time coordinated control of computing resources and power resources in step S5 includes the following steps: S51. Parse the computing power and power coordinated control instruction data, and extract the computing power scheduling instruction and power control instruction; S52. Utilize the computing power management platform to execute computing power scheduling instructions and dynamically adjust the task allocation of the cloud computing center, the load balancing of edge nodes, and the energy efficiency mode of the high-performance computing cluster. S53. Utilize the power control system to execute power control commands, regulate the power flow of the distribution network, the charging and discharging rate of energy storage devices, and the grid-connected power of clean energy. S54. Monitor the control effect in real time, collect system operation data and feed it back to step S1 to form closed-loop optimization control.

[0014] Preferably, the method further includes an exception handling step S6: S61. Real-time monitoring of clean energy power generation anomalies, computing power overload, and power failure events, generating abnormal event data; S62. Apply the isolated forest algorithm to detect abnormal patterns and trigger backup control strategies.

[0015] Preferably, the method further includes a performance evaluation step S7: S71. Collect system data after regulation and control, and calculate clean energy utilization rate, computing power task completion rate and power system stability index. S72. Use the data envelopment analysis algorithm to evaluate the efficiency of coordinated control and generate a performance evaluation report; S73. Based on the performance evaluation report, generate the dynamic parameter configuration file for steps S2 to S5, and load the configuration file at the beginning of the next control cycle to complete the update of the system operating parameters.

[0016] Compared with existing technologies, this invention provides a method for coordinated control of computing power and electricity driven by clean energy, which has the following beneficial effects: 1. In this invention, when conducting coordinated regulation of computing power and electricity driven by clean energy, a unified standard is used to collect clean energy power generation data, computing power demand data, and power system load data. Multi-source time-series coordinated processing is performed to ensure the standardization and comparability of data from different sources. At the same time, based on a multi-dimensional coordinated analysis model, the coupling relationship between the volatility of clean energy and the elastic demand of computing power is analyzed in real time. This can identify the system operating status and potential imbalance risks, ensuring the foresight and accuracy of coordinated regulation of computing power and electricity resources driven by clean energy, and improving the overall energy efficiency of the system.

[0017] 2. In this invention, when coordinating the control of computing power and electricity driven by clean energy, an intelligent prediction algorithm is used to proactively capture the fluctuations in clean energy output and the changing trends in computing power demand. Multi-objective optimization technology is then used to dynamically generate control strategies that take into account economy, stability, and efficiency, enabling the system to proactively adapt to complex operating conditions in different regions and with different energy structures. During strategy execution, model predictive control and reinforcement learning algorithms are used to perform real-time rolling optimization and online correction of control commands, ensuring the speed and adaptability of control actions when facing sudden operating conditions, and enhancing the robustness and reliability of the coordinated control system.

[0018] 3. In this invention, when conducting collaborative regulation of computing power and electricity driven by clean energy, a complete performance evaluation and closed-loop feedback mechanism is established to perform multi-indicator quantitative analysis of the system's operational performance after regulation. Based on the evaluation results, the key parameters of each link in the data collaborative analysis, strategy generation, and instruction execution are automatically adjusted in reverse, enabling the system to form a complete autonomous cycle from perception, decision-making, execution to optimization. This results in the continuous self-improvement and dynamic optimization of the collaborative relationship between clean energy, computing power resources, and the power network, ensuring the long-term stable improvement of the collaborative regulation effect. Attached Figure Description

[0019] Figure 1 This is a flowchart of the clean energy-driven computing power and electricity coordinated control method of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] For specific implementation examples, please refer to: Figure 1A method for coordinated regulation of computing power and electricity driven by clean energy, comprising the following steps: S1. Collect clean energy power generation data, computing power demand data, and power system load data; S2. Based on clean energy power generation data, computing power demand data and power system load data, perform multi-dimensional collaborative analysis and processing of clean energy, computing power and power to generate clean energy-computing power-power collaborative characteristic data. S3. Based on the collaborative characteristic data of clean energy, computing power and electricity, perform matching processing of clean energy volatility and computing power elasticity demand to generate clean energy-driven computing power regulation strategy data. S4. Based on clean energy-driven computing power regulation strategy data, perform dynamic load balance optimization processing of the power system to generate computing power and power coordinated regulation command data; S5 executes commands and data for the coordinated control of computing power and electricity, forming a real-time coordinated control of computing power resources and electricity resources driven by clean energy.

[0022] The collection of clean energy generation data, computing power demand data, and power system load data in S1 includes the following steps: S11. Collect real-time power generation data, power generation forecast data, and weather impact parameters of solar power plants, wind power plants, and hydropower plants through clean energy power generation monitoring equipment, and generate clean energy power generation data. S12. Collect real-time computing task queues, processor utilization, memory usage and energy efficiency ratio data of cloud computing centers, edge computing nodes and high-performance computing clusters through the computing power resource monitoring platform, and generate computing power demand data. S13. Collect regional power grid load curves, distribution network power flow data, energy storage device charging and discharging status and power grid stability indicators through the power system SCADA platform, and generate power system load data; S14. Perform time series alignment and data standardization on clean energy power generation data, computing power demand data, and power system load data to generate multi-source time series collaborative data. The data standardization uses the Z-score normalization algorithm to eliminate dimensional differences, including the following steps: S141. Calculate the arithmetic mean of each data feature in the multi-source time-series collaborative data; S142. Calculate the standard deviation of each data feature; S143. Subtract the arithmetic mean of the corresponding feature from each data point, and then divide by the standard deviation to obtain the standardized data. The mathematical expression of the Z-score normalization algorithm is: ; in, For the original data points, The arithmetic mean of the corresponding features. The standard deviation of the corresponding feature. This is the standardized data.

[0023] The multi-dimensional collaborative analysis process in S2 includes the following steps: S21. Obtain multi-source time-series collaborative data; S22. Principal component analysis (PCA) is used to reduce the dimensionality of multi-source time-series collaborative data, extracting collaborative principal components of clean energy power generation fluctuation characteristics, computing power demand elasticity characteristics, and power load cycle characteristics to generate a dimensionality-reduced feature matrix. The specific steps of dimensionality reduction using PCA include: S221. Calculate the covariance matrix of multi-source time-series collaborative data to describe the correlation between each feature dimension; S222. Calculate the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvalues ​​reflect the variance contribution of the principal components. S223. Sort the eigenvalues ​​in descending order and select the eigenvectors corresponding to the first k eigenvalues ​​as principal components, where k is the preset number of principal components; the number of principal components k is selected as the minimum value that makes the cumulative variance contribution rate greater than 85%. S224. Project the original data onto the selected principal components to generate a dimension-reduced feature matrix; The mathematical expression for data projection is: ; in, The original data matrix, The projection matrix is ​​composed of the selected principal component vectors. The feature matrix after dimensionality reduction; S23. Based on the dimensionality-reduced feature matrix, the grey relational analysis algorithm is applied to calculate the dynamic correlation between clean energy power generation, computing power demand, and electricity load, generating clean energy-computing power-electricity correlation data. The grey relational analysis algorithm includes the following steps: S231. Determine the reference sequence and comparison sequence: Use clean energy power generation data as the reference sequence, and computing power demand data and power load data as the comparison sequence; S232. Data standardization: Initialize and mean the reference and comparison sequences to eliminate the influence of dimensions. S233. Calculate the correlation coefficient: Based on the standardized sequence, calculate the correlation coefficient between each comparison sequence and the reference sequence at each time step; The formula for calculating the grey relational coefficient is: ; in, Indicates the first The comparison sequence in the th ... The correlation coefficient at each time point, Indicates the reference sequence at the 1st The value at each moment. Indicates the first The comparison sequence in the th ... The value at each moment. This represents the resolution coefficient, which ranges from [0,1]. Indicates the sequence length. This represents the minimum difference between two levels. Indicates the maximum difference between two levels; S234. Calculate the correlation degree: Calculate the average value of the correlation coefficients at each time point to obtain the correlation degree between clean energy power generation and computing power demand, and between clean energy power generation and power load; The formula for calculating grey relational degree is: ; in, Indicates the first The correlation between the comparison sequence and the reference sequence, Indicates the first The comparison sequence in the th ... The correlation coefficient at each time point, Indicates the sequence length. This represents the summation of correlation coefficients over all time points. The arithmetic mean of the correlation coefficients; S235. Generate correlation data: Based on the calculated correlation values, generate correlation data between clean energy, computing power, and electricity. S24. Based on the correlation data of clean energy, computing power, and electricity, construct a collaborative analysis model of clean energy, computing power, and electricity, and output collaborative characteristic data of clean energy, computing power, and electricity. The collaborative analysis model uses multiple linear regression to fit the coupling relationship between computing power and electricity driven by clean energy. The multiple linear regression includes the following steps: S241. Use clean energy power generation data, computing power demand data, and power load data as independent variables, and use the coupling relationship index between computing power and power as the dependent variable. S242. Constructing a multiple linear regression model: ; in, As the dependent variable, to As the independent variable, to These are the regression coefficients; S243. Use the least squares method to estimate the regression coefficients and minimize the sum of squared prediction errors. The objective function of the least squares method is: ; in, For the first The dependent variable of the observations For the first The first observation value One independent variable, The number of independent variables is used to quantify the coupling relationship between computing power and electricity driven by clean energy through regression analysis; S244. Output clean energy-computing power-electricity collaborative characteristic data based on the estimated regression coefficients.

[0024] The process of matching clean energy volatility with computing power elasticity demand in S3 includes the following steps: S31. Based on the collaborative characteristic data of clean energy, computing power and electricity, a time series prediction algorithm is used to predict the fluctuation curve of clean energy power generation and the elasticity curve of computing power demand, and generate clean energy fluctuation prediction data and computing power demand prediction data. S32. Apply fuzzy logic control algorithm to calculate the matching degree of clean energy fluctuation prediction data and computing power demand prediction data, and generate a clean energy and computing power demand matching degree index. S33. Based on the matching degree index and combined with the power system stability constraints, a multi-objective optimization algorithm is used to generate clean energy-driven computing power regulation strategy data. The multi-objective optimization algorithm takes minimizing energy waste, maximizing computing power utilization, and minimizing power load fluctuation variance as its objective functions. The multi-objective optimization algorithm includes the following steps: S331. Define the objective function, including the energy waste function, computing power utilization function, and power load stability function; S332. Set constraints, including power system stability constraints; S333. A multi-objective optimization algorithm is used to solve for the Pareto optimal solution set, seeking a balance among multiple objectives. The mathematical description of the multi-objective optimization problem is as follows: ; ; ; in, arrive for The sum of several objective functions corresponds to the objectives of minimizing energy waste, maximizing computing power utilization, and optimizing power load stability, respectively. For inequality constraints, Equality constraint; S334. Select the optimal solution from the Pareto optimal solution set to generate clean energy-driven computing power regulation strategy data; The time series forecasting algorithm in S31 uses a long short-term memory network model, and its mathematical expression is: ; in, For input data, For the previous moment The hidden state vector, In hidden state, and For model parameters, For the Gate of Oblivion For the previous moment The cell state vector, Let be the activation vector of the forget gate. Let be the bias vector of the forget gate. Let be the activation vector of the input gate. This is the current candidate cell state vector. and The input gate weight matrix, and This is the input gate bias vector. Let be the activation vector of the output gate. This is the output gate weight matrix. This is the bias vector for the output gate. It is the Sigmoid activation function. The hyperbolic tangent activation function is used. The hidden state from the previous moment Input at the current moment Perform vector concatenation.

[0025] The dynamic load balancing optimization process for the power system in S4 includes the following steps: S41. Obtain data on clean energy-driven computing power regulation strategies; S42. Using model predictive control algorithm, with clean energy-driven computing power regulation strategy data as input, combined with real-time load data of power system, dynamically optimize computing power resource allocation and power dispatch plan, and generate preliminary regulation instruction data; S43. Apply reinforcement learning algorithms to correct the initial control command data online. Based on power system feedback data, adjust the priority of computing tasks and the charging and discharging strategies of energy storage devices to generate optimized control command data. The reinforcement learning algorithm uses a deep network, and its objective function is: ; in, Indicates the system status. Indicates regulatory action, As a reward value, As a discount factor, This represents the estimated maximum future value. Indicates a new state. Indicates the value of a state action; S44. Compare the optimized control command data with historical control effect data, use the Kalman filter algorithm to eliminate noise interference, and generate computing power and power coordinated control command data. The Kalman filter algorithm includes the following steps: S441. Define the system state equations and observation equations, and establish the system dynamic model; S442. Initialize the state estimate and covariance matrix; S443. Prediction steps: Predict the state and covariance at the next time step based on the state equation; S444. Perform the update step: use the observed data to correct the predicted value and obtain the optimal state estimate; S445. Repeat the prediction and update steps to eliminate noise interference; The recursive formulas for Kalman filtering include: Prediction steps: ; ; in, Indicates based on Time information Prior estimation of the state at time step. This represents the prior estimation error covariance matrix. Represents the state transition matrix. Represents the process noise covariance matrix; Update steps: ; ; ; in, Represents the Kalman gain matrix. Represents the observation matrix. Represents the observation noise covariance matrix. express The actual observed value at time , express Posterior estimation of the state at time step 1. Let represent the posterior estimation error covariance matrix.

[0026] The real-time coordinated control of computing and power resources in S5 includes the following steps: S51. Analyze the data of computing power and power coordinated regulation instructions, and extract computing power scheduling instructions and power control instructions; S52. Execute computing power scheduling instructions through the computing power management platform to dynamically adjust the task allocation of the cloud computing center, the load balancing of edge nodes, and the energy efficiency mode of the high-performance computing cluster. S53. Execute power control commands through the power control system to regulate the power flow of the distribution network, the charging and discharging rate of energy storage devices, and the grid-connected power of clean energy. S54. Monitor the control effect in real time, collect system operation data and feed it back to step S1 to form closed-loop optimization control.

[0027] The method also includes exception handling step S6: S61. Real-time monitoring of clean energy power generation anomalies, computing power overload, and power failure events, generating abnormal event data; S62. Apply the isolated forest algorithm to detect abnormal patterns and trigger backup control strategies. The isolated forest algorithm for detecting abnormal patterns includes the following steps: S621. Constructing an isolated tree forest: Randomly extract a subset of samples from the abnormal event data, and recursively select features and segmentation values ​​to construct multiple isolated trees; S622. Calculate path length: For each data point, calculate the path length from the root node to the leaf node in each isolated tree. S623. Calculate the anomaly score: Calculate the anomaly score for each data point based on the average path length. The shorter the path, the more abnormal it is. The formula for calculating the anomaly score is: ; in, Representing data points The abnormal score, with a value range of [0,1], Representing data points The average path length across all isolated trees. Represents a given number of samples The path length standardization factor at that time, when the score is abnormal When the data points exceed the preset threshold, It was identified as an outlier; S624. Detecting Abnormal Patterns: Determine whether an abnormal pattern exists based on the abnormal score threshold. The threshold is generally set to 0.6, and can be adjusted according to conventional engineering practices. S625, Trigger Backup Strategy: When an abnormal mode is detected, the preset backup control strategy is automatically triggered.

[0028] The method also includes a performance evaluation step S7: S71. Collect system data after regulation and control, and calculate clean energy utilization rate, computing power task completion rate and power system stability index. S72. Evaluate the efficiency of coordinated control using the Data Envelopment Analysis (DEA) algorithm and generate a performance evaluation report. The DEA algorithm includes the following steps: S721. Treat each regulation cycle as a decision-making unit and define input and output indicators; S722. Construct a data envelopment analysis model and calculate the relative efficiency of each decision-making unit; S723. Use linear programming to solve for the efficiency score and evaluate the efficiency of resource allocation. The mathematical expression for the linear programming problem is: ; , ; ; ; in, This represents the efficiency value of the decision-making unit. For the first The first decision-making unit Input metrics, For the first The first decision-making unit Output metrics, Indicates the first The weight of each decision-making unit, Indicates the total number of decision-making units. Indicates the number of input indicators. Indicates the number of output indicators; S724. Generate a performance evaluation report based on the efficiency score; S73. Based on the performance evaluation report, generate dynamic parameter configuration files for steps S2 to S5, and load these configuration files at the start of the next control cycle to complete the update of system operating parameters.

[0029] The operational steps of this clean energy-driven computing power and electricity coordinated control method are as follows: Step 1: Multi-source data acquisition and standardization processing This method first performs a data acquisition step, collecting real-time clean energy power generation data, including power generation and forecast data, through monitoring equipment deployed at clean energy power plants. Next, it collects computing power demand data through a computing resource monitoring platform, calculating task queues and processor utilization. Finally, it collects power system load data, grid load curves, and energy storage status through a power system SCADA platform. Subsequently, it performs time series alignment and data standardization, employing the Z-score normalization algorithm to eliminate dimensional differences between data from different sources, generating regular and comparable multi-source time-series collaborative data, laying the data foundation for subsequent analysis.

[0030] Step 2: Multi-dimensional collaborative feature extraction and analysis Based on the preprocessed multi-source time-series collaborative data, this method performs multi-dimensional collaborative analysis. First, principal component analysis is used to reduce the dimensionality of the high-dimensional data, extracting collaborative principal components that can represent the fluctuation characteristics of clean energy power generation, the elasticity characteristics of computing power demand, and the cyclical characteristics of power load, generating a dimensionality-reduced feature matrix. Then, the grey relational analysis algorithm is applied to calculate the dynamic correlation degree among the three, generating clean energy-computing power-power correlation degree data. Finally, based on this, a clean energy-computing power-power collaborative analysis model is constructed, typically using multiple linear regression to fit its coupling relationship, ultimately outputting clean energy-computing power-power collaborative feature data that can quantify the intrinsic connection of the system.

[0031] Step 3: Generation of Control Strategies Based on Prediction and Optimization The core decision-making step of this method is the matching of clean energy volatility and computing power elasticity demand. Based on the collaborative feature data generated in the previous step, this step first uses a time series prediction algorithm to predict the clean energy power generation fluctuation curve and the computing power demand elasticity curve, generating clean energy volatility prediction data and computing power demand prediction data. Subsequently, a fuzzy logic control algorithm is applied to calculate the matching degree of the two sets of prediction data, generating a clean energy and computing power demand matching degree index. Finally, combined with the power system stability constraints, a multi-objective optimization algorithm is used to generate optimal clean energy-driven computing power regulation strategy data with the objectives of minimizing energy waste, maximizing computing power utilization, and minimizing power load fluctuation variance.

[0032] Step 4: Dynamic Load Balancing Optimization and Command Correction After generating the initial strategy, this method performs dynamic load balancing optimization of the power system. This step takes clean energy-driven computing power control strategy data as input. First, a model predictive control algorithm is used to dynamically optimize the allocation of computing power resources and the power dispatch plan to generate initial control command data. To further improve adaptability, a reinforcement learning algorithm is applied based on real-time feedback data from the power system to correct the initial commands online, adjust the priority of computing power tasks and energy storage strategies, and generate optimized control command data. Finally, a Kalman filter algorithm is used to eliminate noise interference and output the final executable computing power and power coordinated control command data.

[0033] Step 5: Instruction Execution and Closed-Loop Feedback Optimization The final stage involves executing the collaborative control commands for computing power and power. The control system parses the commands and issues computing power scheduling commands and power control commands to the computing power management platform and the power control system, respectively. Specifically, it adjusts the task allocation of the cloud computing center, the load balancing of edge nodes, the power flow of the distribution network, and the grid-connected power of clean energy. At the same time, the system monitors the control effect in real time and feeds back the collected system operation data to the data acquisition stage in the first step, forming a complete closed-loop optimization control. In addition, the method also includes anomaly handling and performance evaluation steps. Anomalies are detected by the isolated forest algorithm and backup strategies are activated. Efficiency is evaluated and system parameters are dynamically updated by the data envelopment analysis algorithm to ensure the continuous, stable, and efficient operation of the system.

[0034] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for coordinated control of computing power and electricity driven by clean energy, characterized in that: The method includes the following steps: S1. Collect clean energy power generation data, computing power demand data, and power system load data; S2. Based on the clean energy power generation data, the computing power demand data, and the power system load data, perform multi-dimensional collaborative analysis and processing of clean energy, computing power, and electricity to generate clean energy-computing power-electricity collaborative feature data. S3. Based on the clean energy-computing power-electricity synergistic characteristic data, perform clean energy volatility and computing power elasticity demand matching processing to generate clean energy-driven computing power regulation strategy data; S4. Based on the clean energy-driven computing power control strategy data, perform dynamic load balance optimization processing of the power system to generate computing power and power coordinated control instruction data; S5. Execute the computing power and power coordinated control command data to form real-time coordinated control of computing power resources and power resources driven by clean energy.

2. The method for coordinated control of computing power and electricity driven by clean energy according to claim 1, characterized in that: The collection of clean energy power generation data, computing power demand data, and power system load data in S1 includes the following steps: S11. Clean energy power generation monitoring equipment collects real-time power generation data, power generation forecast data, and weather impact parameters from solar power plants, wind power plants, and hydropower plants, and generates clean energy power generation data. S12. Use the computing power resource monitoring platform to collect real-time computing task queues, processor utilization, memory usage and energy efficiency ratio data of cloud computing centers, edge computing nodes and high-performance computing clusters, and generate computing power demand data. S13. The power system SCADA platform collects regional power grid load curves, distribution network power flow data, energy storage device charging and discharging status and power grid stability indicators, and generates power system load data. S14. Perform time series alignment and data standardization on the clean energy power generation data, the computing power demand data, and the power system load data to generate multi-source time series collaborative data. The data standardization uses the Z-score normalization algorithm to eliminate dimensional differences.

3. The clean energy-driven computing power and power coordinated control method according to claim 1, characterized in that: The multi-dimensional collaborative analysis and processing in S2 includes the following steps: S21. Obtain the multi-source time-series collaborative data; S22. The principal component analysis algorithm is used to reduce the dimensionality of the multi-source time-series collaborative data, extract the collaborative principal components of clean energy power generation fluctuation characteristics, computing power demand elasticity characteristics and power load cycle characteristics, and generate a dimensionality-reduced feature matrix. S23. Based on the reduced-dimensional feature matrix, the gray relational analysis algorithm is applied to calculate the dynamic correlation between clean energy power generation, computing power demand and power load, and generate clean energy-computing power-power correlation data. S24. Based on the clean energy-computing power-electricity correlation data, construct a clean energy-computing power-electricity collaborative analysis model and output clean energy-computing power-electricity collaborative characteristic data. The collaborative analysis model uses multiple linear regression to fit the coupling relationship between computing power and electricity driven by clean energy.

4. The method for coordinated control of computing power and electricity driven by clean energy according to claim 1, characterized in that: The process of matching clean energy volatility with computing power elasticity demand in S3 includes the following steps: S31. Based on the clean energy-computing power-electricity collaborative characteristic data, a time series prediction algorithm is used to predict the clean energy power generation fluctuation curve and the computing power demand elasticity curve, and clean energy fluctuation prediction data and computing power demand prediction data are generated. S32. Apply fuzzy logic control algorithm to calculate the matching degree of the clean energy fluctuation prediction data and the computing power demand prediction data, and generate a clean energy and computing power demand matching degree index. S33. Based on the matching degree index and combined with the power system stability constraints, a multi-objective optimization algorithm is used to generate clean energy-driven computing power regulation strategy data, wherein the multi-objective optimization algorithm takes minimizing energy waste, maximizing computing power utilization, and minimizing power load fluctuation variance as objective functions.

5. The clean energy-driven computing power and power coordinated control method according to claim 4, characterized in that: The time series prediction algorithm in S31 uses a long short-term memory network model, and its mathematical expression is: ； in, For input data, For the previous moment The hidden state vector, In hidden state, and For model parameters, Here is the weight matrix for the forget gate. For the previous moment The cell state vector, Let be the activation vector of the forget gate. Let be the bias vector of the forget gate. Let be the activation vector of the input gate. This is the current candidate cell state vector. and The input gate weight matrix, and This is the input gate bias vector. Let be the activation vector of the output gate. This is the output gate weight matrix. This is the bias vector for the output gate. It is the Sigmoid activation function. The hyperbolic tangent activation function is used. The hidden state from the previous moment Input at the current moment Perform vector concatenation.

6. The clean energy-driven computing power and power coordinated control method according to claim 1, characterized in that: The dynamic load balancing optimization process for the power system in S4 includes the following steps: S41. Obtain the clean energy-driven computing power regulation strategy data; S42. Using the model predictive control algorithm, with the clean energy-driven computing power regulation strategy data as input, and combined with the real-time load data of the power system, dynamically optimize the allocation of computing power resources and the power dispatch plan to generate preliminary regulation instruction data. S43. Apply reinforcement learning algorithm to correct the preliminary control command data online, adjust the priority of computing power tasks and the charging and discharging strategy of energy storage equipment based on power system feedback data, and generate optimized control command data. S44. The optimized control command data is compared with the historical control effect data, and the Kalman filter algorithm is used to eliminate noise interference to generate computing power and power coordinated control command data.

7. The clean energy-driven computing power and power coordinated control method according to claim 6, characterized in that: The reinforcement learning algorithm in S43 uses a deep network, and its objective function is: ； in, Indicates the system status. Indicates regulatory action, As a reward value, As a discount factor, For learning rate, This represents the estimated maximum future value. Indicates a new state. Represents the state-action value.

8. The method for coordinated control of computing power and electricity driven by clean energy according to claim 1, characterized in that: The real-time coordinated control of computing resources and power resources in S5 includes the following steps: S51. Parse the computing power and power coordinated control instruction data, and extract the computing power scheduling instruction and power control instruction; S52, the computing power management platform executes computing power scheduling instructions and dynamically adjusts the task allocation of the cloud computing center, the load balancing of edge nodes, and the energy efficiency mode of the high-performance computing cluster; S53. The power control system executes power control commands to regulate the power flow of the distribution network, the charging and discharging rate of energy storage devices, and the grid-connected power of clean energy. S54. Monitor the control effect in real time, collect system operation data and feed it back to step S1 to form closed-loop optimization control.

9. The method for coordinated control of computing power and electricity driven by clean energy according to claim 1, characterized in that: The method further includes an exception handling step S6: S61. Real-time monitoring of clean energy power generation anomalies, computing power overload, and power failure events, generating abnormal event data; S62. Apply the isolated forest algorithm to detect abnormal patterns and trigger backup control strategies.

10. The method for coordinated control of computing power and electricity driven by clean energy according to claim 1, characterized in that: The method further includes a performance evaluation step S7: S71. Collect system data after regulation and control, and calculate clean energy utilization rate, computing power task completion rate and power system stability index. S72. Use the data envelopment analysis algorithm to evaluate the efficiency of coordinated control and generate a performance evaluation report; S73. Based on the performance evaluation report, generate the dynamic parameter configuration file for steps S2 to S5, and load the configuration file at the beginning of the next control cycle to complete the update of the system operating parameters.

Images