Intelligent data center energy management method, device and equipment and storage medium

Abstract

The invention relates to the technical field of data center energy management. The invention relates to an intelligent data center energy management method, device and equipment and a storage medium, and the method comprises the steps: obtaining energy data, collecting and monitoring the real-time power consumption data and real-time environment temperature data of a data center in real time, predicting the energy data through a trained LSTM model, and obtaining the energy data of the data center. The method comprises the steps of obtaining a first energy consumption prediction result, optimizing and adjusting the first energy consumption prediction result through a trained attribution model to obtain an energy consumption prediction result, comparing a preset energy consumption threshold value with the energy consumption prediction result to obtain an energy consumption comparison result, and dynamically adjusting the working states of an energy storage unit and a heat dissipation unit according to the energy consumption comparison result. The intelligent energy management system has the advantages that intelligent energy management is achieved, energy consumption is effectively reduced, and the energy use efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data center energy management, and in particular to an intelligent data center energy management method, device, equipment and storage medium. Background Art

[0002] With the rapid development of information technology, data centers, as crucial infrastructure supporting cloud computing, big data, and other information technology services, are facing an increasingly prominent energy consumption problem. High data center energy consumption not only increases operating costs but also places significant pressure on the environment. Therefore, intelligent data center energy management, reducing energy consumption, and improving efficiency have become urgent challenges for the industry. Summary of the Invention

[0003] In order to achieve intelligent energy management, effectively reduce energy consumption and improve energy utilization efficiency, the present application provides an intelligent data center energy management method, device, equipment and storage medium.

[0004] The above-mentioned invention objective of this application is achieved through the following technical solutions: An intelligent data center energy management method, the intelligent data center energy management method comprising: Acquiring energy data, including collecting and monitoring real-time power consumption data and real-time ambient temperature data of the data center; The energy data is predicted using the trained LSTM model to obtain a first energy consumption prediction result; Optimizing and adjusting the first energy consumption prediction result using the trained attribution model to obtain an energy consumption prediction result; The energy consumption prediction result is compared with the preset energy consumption threshold to obtain an energy consumption comparison result, and the working states of the energy storage unit and the heat dissipation unit are dynamically adjusted according to the energy consumption comparison result.

[0005] By adopting the above technical solution and leveraging artificial intelligence and machine learning algorithms, intelligent management of data center energy consumption is achieved. Future energy consumption can be predicted and optimized based on the predicted results. Based on the predicted results, the operating status of the energy storage and cooling units can be adjusted in advance to avoid energy waste. This optimizes energy efficiency and reduces unnecessary energy consumption. Energy consumption prediction using a trained LSTM model improves accuracy and reliability. The predicted results are optimized and adjusted using an attribution model, making energy management more precise. Predicted energy consumption is compared with the preset energy consumption threshold to generate an energy consumption comparison result. Based on this comparison result, the operating status of the energy storage and cooling units is dynamically adjusted to ensure efficient energy utilization and stable system operation.

[0006] In a preferred example, the present application may be further configured as follows: before acquiring energy data, which includes collecting and real-time monitoring of power consumption data and ambient temperature data of the data center, the process further includes: Acquire historical power consumption data and historical ambient temperature data, and divide the historical power consumption data and the historical ambient temperature data into time series to obtain a training data set and a validation data set; Using the training data set to adjust and optimize parameters of a preset LSTM model and a preset attribution model; The preset LSTM model and the preset attribution model are evaluated using the verification data set to obtain the trained LSTM model and the trained attribution model.

[0007] By employing this technical solution, we acquire historical power consumption and ambient temperature data and segment it into time series, enabling a comprehensive understanding of energy consumption and environmental changes in the data center over different time periods. This provides rich training data for model training. We use the training dataset to adjust and optimize the parameters of the pre-set LSTM model and attribution model, ensuring that the model accurately captures patterns in energy consumption and improves prediction accuracy. We evaluate the model using the validation dataset to ensure that it generalizes well and can accurately predict energy consumption in real-world applications.

[0008] In a preferred example, the present application may be further configured as follows: the intelligent data center energy management method includes: Inputting the real-time power consumption data and the real-time ambient temperature data into the trained LSTM model to obtain the first energy consumption prediction result; Inputting the first energy consumption prediction result, the historical power consumption data, and the ambient temperature data into the trained attribution model to obtain an attribution analysis result; According to the attribution analysis result, the first energy consumption prediction result is optimized and adjusted to obtain the energy consumption prediction result.

[0009] By employing the above technical solution, the LSTM model excels at processing time series data. By training with historical power consumption data and ambient temperature data, it can better capture time-series patterns in energy consumption and improve forecast accuracy. By inputting the first energy consumption forecast results and historical data into the trained attribution model for attribution analysis, key factors influencing energy consumption can be identified. Based on the attribution analysis results, the first energy consumption forecast results are optimized and adjusted to further improve the accuracy and reliability of energy consumption forecasts, resulting in more precise energy consumption forecasts.

[0010] In a preferred example, the present application can be further configured as follows: inputting the first energy consumption prediction result, the historical power consumption data, and the ambient temperature data into the trained attribution model to obtain an attribution analysis result, including: the attribution model uses the SHAP method to perform attribution analysis.

[0011] By employing the above technical solutions, the SHAP method calculates the contribution of each feature to the prediction results, providing accurate energy consumption forecasts and helping to better understand and optimize energy use. SHAP values identify key factors influencing energy consumption, allowing further optimization and adjustment of energy consumption forecasts to ensure the accuracy and effectiveness of energy management. Through attribution analysis using the SHAP method, the main factors influencing energy consumption are identified, and the configuration and scheduling of energy storage and cooling equipment are optimized to achieve optimal resource allocation.

[0012] In a preferred example, the present application may be further configured as follows: comparing the energy consumption prediction result with the preset energy consumption threshold to obtain an energy consumption comparison result, and dynamically adjusting the working states of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result, including: If the energy consumption prediction result exceeds the preset energy consumption threshold, the energy storage unit is activated in advance to reserve energy and the working efficiency of the heat dissipation unit is improved; If the energy consumption prediction result is lower than the preset energy consumption threshold, the energy reserve of the energy storage unit is reduced and the operating frequency of the heat dissipation unit is reduced.

[0013] By adopting the above technical solution, when the energy consumption forecast exceeds the preset energy consumption threshold, the energy storage unit is activated in advance to reserve energy. This allows for preparation before peak energy consumption arrives and avoids energy shortages in emergencies. This improves the efficiency of the cooling unit, proactively responding to the cooling needs caused by high energy consumption, ensuring that the data center temperature is within a controllable range and preventing equipment damage due to overheating. When the energy consumption forecast falls below the preset energy consumption threshold, the energy storage unit's energy reserves are reduced, eliminating unnecessary energy storage processes and avoiding energy waste. The cooling unit's operating frequency is reduced, reducing its energy consumption when energy demand is low, further saving energy.

[0014] In a preferred example, the present application may be further configured as follows: the intelligent data center energy management method includes: The working states of the energy storage unit and the heat dissipation unit are dynamically adjusted according to the energy consumption comparison result, and a DQN control algorithm is used to accurately control the energy storage unit and the heat dissipation unit.

[0015] By implementing the above technical solution and precisely controlling the operating status of the energy storage and cooling units through the DQN algorithm, optimal energy utilization efficiency can be achieved under varying energy consumption conditions. When high energy consumption is predicted, energy reserves are pre-stored and cooling efficiency is improved; when low energy consumption is predicted, energy reserves and cooling frequency are reduced to avoid energy waste. The DQN algorithm responds to energy consumption changes in real time, dynamically adjusting energy storage and cooling strategies to ensure stable data center operation despite load fluctuations. Precisely controlling the operating status of the energy storage and cooling units reduces unnecessary energy consumption, ultimately lowering the data center's energy consumption.

[0016] In a preferred example, the present application may be further configured as follows: dynamically adjusting the working states of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result using a DQN control algorithm to precisely control the energy storage unit and the heat dissipation unit, further comprising: Energy storage units, which store electricity during periods of low electricity demand and release it during periods of peak demand; The heat dissipation unit is used to dissipate heat when the server equipment in the data center is running.

[0017] By implementing this technical solution, the energy storage unit stores electricity during off-peak periods and releases it during peak periods, effectively utilizing power resources, optimizing power usage, and reducing power waste. The energy storage unit can provide additional power during peak demand periods, preventing unstable data center operations or downtime caused by power shortages. Effective heat dissipation from the cooling unit prevents server performance degradation or failures caused by overheating, ensuring stable data center operations.

[0018] The second object of the present invention is achieved through the following technical solutions: An intelligent data center energy management device, comprising: A data acquisition module is used to acquire energy data, including real-time power consumption data and real-time ambient temperature data of the data center. An LSTM model module is used to predict the energy data using a trained LSTM model to obtain a first energy consumption prediction result; an attribution model module, configured to optimize and adjust the first energy consumption prediction result using a trained attribution model to obtain an energy consumption prediction result; The result comparison and state adjustment module is used to compare the energy consumption prediction result with the preset energy consumption threshold to obtain the energy consumption comparison result, and dynamically adjust the working state of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result.

[0019] By adopting the above technical solution and leveraging artificial intelligence and machine learning algorithms, intelligent management of data center energy consumption is achieved. Future energy consumption can be predicted and optimized based on the predicted results. Based on the predicted results, the operating status of the energy storage and cooling units can be adjusted in advance to avoid energy waste. This optimizes energy efficiency and reduces unnecessary energy consumption. Energy consumption prediction using a trained LSTM model improves accuracy and reliability. The predicted results are optimized and adjusted using an attribution model, making energy management more precise. Predicted energy consumption is compared with the preset energy consumption threshold to generate an energy consumption comparison result. Based on this comparison result, the operating status of the energy storage and cooling units is dynamically adjusted to ensure efficient energy utilization and stable system operation.

[0020] The third objective of this application is achieved through the following technical solutions: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the intelligent data center energy management method are implemented.

[0021] The fourth objective of this application is achieved through the following technical solutions: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned intelligent data center energy management method.

[0022] In summary, this application includes at least one of the following beneficial technical effects: 1. By utilizing artificial intelligence and machine learning algorithms, intelligent management of data center energy consumption is achieved. It can predict future energy consumption and make optimization adjustments based on the prediction results. According to the energy consumption prediction results, the working status of the energy storage and heat dissipation units are adjusted in advance to avoid energy waste. Energy utilization efficiency is optimized and unnecessary energy consumption is reduced. Energy consumption is predicted using the trained LSTM model to improve the accuracy and reliability of the prediction. The prediction results are optimized and adjusted through the attribution model to make energy consumption management more accurate. The energy consumption comparison results are obtained by comparing the preset energy consumption threshold with the energy consumption prediction results. According to the energy consumption comparison results, the working status of the energy storage unit and the heat dissipation unit are dynamically adjusted to ensure efficient energy utilization and stable operation of the system; 2. The DQN algorithm precisely controls the working status of the energy storage unit and the heat dissipation unit, achieving optimal energy utilization efficiency under different energy consumption conditions. When high energy consumption is predicted, energy is reserved in advance and heat dissipation efficiency is improved; when low energy consumption is predicted, energy storage and heat dissipation frequency are reduced to avoid energy waste. The DQN algorithm can respond to energy consumption changes in real time and dynamically adjust energy storage and heat dissipation strategies to ensure that the data center maintains stable operation when the load fluctuates. Precisely controlling the working status of the energy storage unit and the heat dissipation unit reduces unnecessary energy consumption and reduces the energy consumption of the data center. 3. When the energy consumption forecast exceeds the preset threshold, the energy storage unit is activated in advance to reserve energy. This allows for preparation before peak energy consumption arrives and prevents energy shortages in emergencies. This improves the efficiency of the cooling unit, proactively responding to the cooling demands of high energy consumption, ensuring that the data center temperature remains within a controllable range and preventing equipment damage from overheating. When the energy consumption forecast falls below the preset threshold, the energy storage unit's energy reserves are reduced, eliminating unnecessary energy storage processes and avoiding energy waste. The cooling unit's operating frequency is reduced, reducing its energy consumption when energy demand is low, further saving energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of an intelligent data center energy management method in one embodiment of the present application; Figure 2 This is a flowchart of the implementation of the intelligent data center energy management method before step S10 in one embodiment of the present application; Figure 3 This is a flowchart of the implementation of the intelligent data center energy management method after step S10 in one embodiment of the present application; Figure 4 This is a flowchart for implementing step S201 of the intelligent data center energy management method in one embodiment of the present application; Figure 5 This is a flowchart for implementing step S40 of the intelligent data center energy management method in one embodiment of the present application; Figure 6 This is a flowchart of the implementation of the intelligent data center energy management method after step S42 in one embodiment of the present application; Figure 7 This is a flowchart for implementing step S421 of the intelligent data center energy management method in one embodiment of the present application; Figure 8 This is a principle block diagram of an intelligent data center energy management method in one embodiment of the present application; Figure 9 It is a schematic diagram of a device in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application is further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, if Figure 1 As shown, this application discloses an intelligent data center energy management method, which specifically includes the following steps: S10: Acquire energy data, which includes collecting and real-time monitoring of real-time power consumption data and real-time ambient temperature data of the data center.

[0026] In this embodiment, the real-time power consumption data refers to the actual power consumption of the data center at the current time point. The real-time ambient temperature data refers to the current temperature of the internal and external environments of the data center.

[0027] Specifically, power monitoring equipment is installed in the data center to collect real-time power consumption data. Temperature sensors are installed inside and outside the data center to monitor ambient temperature in real time. This provides a real-time display of the data center's power consumption and ambient temperature.

[0028] S20: Use the trained LSTM model to predict the energy data and obtain the first energy consumption prediction result.

[0029] In this embodiment, the trained LSTM model refers to a long short-term memory (LSTM) neural network model that has been trained and optimized using historical data. By training on historical data center power consumption data and historical ambient temperature data, the LSTM model effectively captures temporal dependencies and trends in the data, enabling accurate prediction of data center energy consumption over the next period of time. The first energy consumption forecast result refers to the preliminary energy consumption forecast obtained by using the trained LSTM model to predict real-time energy data.

[0030] Specifically, clean the real-time energy data to remove outliers and noise. Normalize the energy data and scale it to a certain range (e.g., 0 to 1) for better processing by the LSTM model. Construct the cleaned and normalized data into a time series, using a sliding window approach to generate the dataset required for training and prediction. Determine the input and output formats of the LSTM model. For example, the input could be historical data from the past 24 hours, and the output could be energy consumption forecasts for the next hour. Define the LSTM model architecture, including the input layer, LSTM layer, fully connected layer, and output layer. Load the pre-trained LSTM model and its weight parameters. Input the real-time energy data into the LSTM model. The input data includes real-time power consumption data and real-time ambient temperature data. The LSTM model receives the input data, performs forward propagation calculations, and outputs a prediction result. The model output is the first energy consumption forecast result, which represents the predicted energy consumption of the data center for a period of time.

[0031] S30: Optimizing and adjusting the first energy consumption prediction result through the trained attribution model to obtain an energy consumption prediction result.

[0032] In this embodiment, the trained attribution model refers to a model trained and optimized using historical power consumption data and historical ambient temperature data of the data center. The energy consumption prediction result refers to the final energy consumption prediction value after optimization and adjustment of the attribution model.

[0033] Specifically, the first energy consumption forecast result generated by the LSTM model and its corresponding characteristic variables are input into the trained attribution model. Based on the input data, the attribution model calculates the attribution value of each characteristic variable on the first energy consumption forecast result, quantifying the impact of each characteristic variable. An analysis report is generated based on the attribution value, detailing the magnitude and direction of the impact of each characteristic variable. Based on the attribution analysis report, an optimization and adjustment strategy is formulated. For example, if a characteristic variable (such as ambient temperature) has a significant impact on the energy consumption forecast result, corresponding adjustments can be made for that variable. The adjustment strategy is applied to the first energy consumption forecast result, optimizing and adjusting it to obtain a more accurate energy consumption forecast result. The optimized and adjusted forecast result is used as the final energy consumption forecast result.

[0034] S40: Compare the preset energy consumption threshold with the energy consumption prediction result to obtain an energy consumption comparison result, and dynamically adjust the working states of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result.

[0035] In this embodiment, the preset energy consumption threshold refers to the set upper and lower limits of energy consumption. The energy consumption comparison result refers to the result obtained by comparing the energy consumption prediction result with the preset energy consumption threshold.

[0036] Specifically, an upper threshold is set based on the data center's maximum energy consumption tolerance. For example, it can be based on the equipment's maximum power consumption and safe operating range. A lower threshold is set to indicate that low energy consumption may indicate resource waste or underutilization of equipment. The final energy consumption forecast is generated using the LSTM model and attribution model. The forecast is compared with the preset upper and lower thresholds to determine whether it exceeds the threshold range. An energy consumption comparison result is generated. If the forecast is within the threshold range, the data center's energy consumption is normal. If the forecast exceeds the upper threshold, high energy consumption may occur in the future, requiring action. If the forecast is below the lower threshold, low energy consumption may be expected in the future, requiring resource reduction. If the forecast exceeds the upper threshold, energy storage can be increased, and energy storage units can be activated in advance to reserve more energy to cope with high energy demand. Heat dissipation efficiency can be improved by increasing the efficiency of heat dissipation units to ensure normal heat dissipation even under high energy consumption conditions. If the forecast falls below the lower threshold, energy storage can be reduced, reducing the energy reserve of the energy storage units to avoid energy waste. Heat dissipation frequency can be reduced by reducing the operating frequency of the heat dissipation units to save energy. Normal: Maintain the current working status of the energy storage unit and heat dissipation unit, no adjustment is required.

[0037] In one embodiment, if Figure 2 As shown, before step S10, that is, obtaining energy data, the energy data includes collecting and real-time monitoring of the power consumption data and ambient temperature data of the data center, and further includes: S101: Acquire historical power consumption data and historical ambient temperature data, divide the historical power consumption data and historical ambient temperature data into time series, and obtain a training data set and a validation data set.

[0038] In this embodiment, historical power consumption data refers to data on power consumption over the data center's historical timeframe. Historical ambient temperature data refers to data on ambient temperature changes over the data center's historical timeframe. The training dataset refers to a portion of the historical power consumption data and historical ambient temperature data, used to train the LSTM model and attribution model. The validation dataset refers to another portion of the historical power consumption data and historical ambient temperature data, used to validate the performance of the LSTM model and attribution model.

[0039] Specifically, historical power consumption data and historical ambient temperature data are obtained from the data center's monitoring system, energy management system, or environmental monitoring system. Outliers and missing values are removed to ensure data integrity and consistency. Power consumption data and ambient temperature data are aligned at the same time interval, such as hourly, daily, or weekly. Time series partitioning: Data is divided into training and validation datasets based on time series. Common partitioning methods include proportional partitioning, such as 70% as the training dataset and 30% as the validation dataset, or partitioning based on chronological order, such as the first 80% of the data as the training dataset and the last 20% as the validation dataset. Time series data selected from the preprocessed data is used to train the LSTM model and attribution model. Contains historical records of power consumption and ambient temperature. Another portion of time series data selected from the preprocessed data is used to verify the performance of the LSTM model and attribution model. Ensure that the training and validation datasets do not overlap to test the generalization ability of the model.

[0040] S102: Use the training dataset to adjust and optimize parameters of the preset LSTM model and the preset attribution model.

[0041] In this embodiment, the preset LSTM model refers to a long short-term memory model with a predefined structure and initial parameters. The preset attribution model refers to a model with a predefined structure and initial parameters, which is used to analyze and optimize the prediction results of the LSTM model.

[0042] Specifically, Z-score normalization is performed on the energy consumption data and ambient temperature data in the training dataset to improve the stability and convergence speed of model training. The raw data in the training dataset is converted into a time series format acceptable to the LSTM model. Data from multiple consecutive time points are combined as the input sequence for the LSTM model. Based on the preset LSTM model structure, the number of neurons and the connection structure in the input, hidden, and output layers are defined. Model parameters, such as weights and biases, are initialized. The LSTM model is trained using the training dataset, and parameter updates are performed using the backpropagation algorithm and a gradient descent optimizer (such as Adam or SGD). The training process includes forward propagation, loss calculation, and backpropagation. Using cross-validation, the LSTM model's hyperparameters (such as the learning rate, number of hidden layer neurons, and batch size) are adjusted to optimize model performance. The LSTM model's prediction results are combined with actual energy consumption data to generate a training dataset for the attribution model. The training dataset includes the LSTM prediction results, actual energy consumption data, and other relevant features. Based on the preset attribution model structure, the number of neurons and the connection structure in the input, hidden, and output layers are defined. Model parameters are initialized. The attribution model is trained using the training dataset, and the backpropagation algorithm is used to update its parameters. The training process includes forward propagation, loss calculation, and backpropagation. Cross-validation is used to adjust the attribution model's hyperparameters (such as the learning rate, number of hidden layer neurons, and batch size) to optimize model performance. The validation dataset is used to evaluate the performance of the LSTM and attribution models, and the prediction errors (such as mean squared error and mean absolute error) are calculated. The trained LSTM and attribution models are deployed in a real-world data center environment to monitor and optimize energy consumption and heat dissipation in real time.

[0043] S103: Using the validation data set to evaluate the preset LSTM model and the preset attribution model, to obtain a trained LSTM model and a trained attribution model.

[0044] Specifically, a portion of historical power consumption data and historical ambient temperature data is allocated as a validation dataset. This data is not used in model training and is used only to evaluate model performance. The validation dataset is input into a pre-set LSTM model to obtain prediction results. The LSTM model predicts data center energy consumption for a period of time based on time series data. The LSTM model's prediction results are compared with actual energy consumption data to calculate the model's prediction error. Common error metrics include mean squared error (MSE) and mean absolute error (MAE). The LSTM model's prediction performance is evaluated based on the error calculation results. A low error indicates good predictive ability. The LSTM model's prediction results are combined with actual energy consumption data to generate a validation dataset for the attribution model. The validation dataset includes the LSTM prediction results, actual energy consumption data, and other relevant features. The validation dataset is input into the pre-set attribution model to obtain optimized energy consumption prediction results. The optimized attribution model's results are compared with actual energy consumption data to calculate the model's prediction error. Error calculations are also performed using metrics such as mean squared error (MSE) and mean absolute error (MAE). The optimized attribution model's performance is evaluated based on the error calculation results. If the error decreases further, it indicates that the attribution model has effectively improved prediction accuracy. Based on the evaluation results of the validation set, adjust the LSTM model and attribution model until the model performance reaches the expected target.

[0045] The mean square error is used to measure the square average of the difference between the predicted value and the actual value. The formula is: Where n is the number of samples, yi is the actual value of the i-th sample, is the predicted value of the i-th sample. The mean absolute error is used to measure the average absolute value of the difference between the predicted value and the actual value. The formula is: Where n is the number of samples, yi is the actual value of the i-th sample, is the predicted value of the i-th sample.

[0046] In one embodiment, if Figure 3 As shown, after step S10, the intelligent data center energy management method includes: S201: Input the real-time power consumption data and the real-time ambient temperature data into the trained LSTM model to obtain a first energy consumption prediction result.

[0047] Specifically, real-time power consumption data and real-time ambient temperature data are collected from the data center. The data is ensured to be time series data arranged in chronological order. The real-time data is normalized to make it suitable for LSTM model input. The power consumption data and ambient temperature data are combined into the LSTM model input format, forming an input sequence containing time steps. The trained LSTM model is used to predict the preprocessed input sequence, generating the first energy consumption forecast result.

[0048] S202: Input the first energy consumption prediction result, historical power consumption data, and ambient temperature data into the trained attribution model to obtain an attribution analysis result.

[0049] Specifically, real-time power consumption data and real-time ambient temperature data are collected and standardized or normalized to ensure that the input data is suitable for model processing. Data standardization can be achieved by converting the data to a mean of 0 and a standard deviation of 1, or by normalizing it to a range of 0-1. The preprocessed real-time power consumption data, real-time ambient temperature data, and the first energy consumption forecast result are input. The attribution model uses this data to perform calculations and analyze the contribution of each feature to the energy consumption forecast. The features include real-time power consumption data, real-time ambient temperature data, historical power consumption data, historical ambient temperature data, and the first energy consumption forecast result. The attribution analysis results will display the impact of each feature on the energy consumption forecast. For example, the attribution model may indicate the extent of the impact of the current ambient temperature on the energy consumption forecast, or how historical power consumption data affects future energy consumption forecasts. The output of the attribution model is analyzed to determine which factors have the greatest impact on the energy consumption forecast. For example, it may be found that a 1°C increase in ambient temperature leads to a 0.5% increase in energy consumption, or that historical power consumption data for a certain time period has a significant impact on the current energy consumption forecast.

[0050] S203: Optimize and adjust the first energy consumption prediction result according to the attribution analysis result to obtain an energy consumption prediction result.

[0051] Specifically, the historical power consumption data, historical ambient temperature data, and the first energy consumption prediction result are input into the trained attribution model to obtain the attribution analysis result. For example, the attribution analysis result may show that the current temperature has a 20% impact on the energy consumption prediction, and the power consumption trend in the past week has a 50% impact on the current prediction. Based on the attribution analysis results, the impact value of each feature on the first energy consumption prediction result is calculated. Example: Assume that the first energy consumption prediction result is 110kW, the ambient temperature increases energy consumption by 20% (i.e., 22kW), and the power consumption trend in the past week increases energy consumption by 50% (i.e., 55kW). Based on the attribution analysis results of each feature, the first energy consumption prediction result is adjusted to obtain the final energy consumption prediction result. Example: Final energy consumption prediction result = first energy consumption prediction result + ambient temperature impact value + power consumption trend impact value = 110kW + 22kW + 55kW = 187kW.

[0052] In one embodiment, if Figure 4 As shown, in step S201, the first energy consumption prediction result, historical power consumption data, and ambient temperature data are input into the trained attribution model to obtain attribution analysis results, including: S2011: The attribution model uses the SHAP method for attribution analysis.

[0053] In this embodiment, the SHAP method is a game-theoretic interpretive method used to explain the output of a machine learning model. It explains the importance of each feature by its contribution to the prediction result. The features are real-time power consumption data, real-time ambient temperature data, historical power consumption data, historical ambient temperature data, and the first energy consumption prediction result.

[0054] Specifically, real-time power consumption data and real-time ambient temperature data are obtained. The data is input into the trained LSTM model to obtain the first energy consumption prediction result. The first energy consumption prediction result and related historical data are input into the SHAP interpreter. The SHAP interpreter calculates the Shapley value of each feature and generates an attribution analysis result. The attribution analysis result shows the influence of each feature (such as power consumption data, ambient temperature data, etc.) on the first energy consumption prediction result. Example: The contribution of ambient temperature to the first energy consumption prediction result is +15kW, and the trend contribution of power consumption data is +25kW. Based on the attribution analysis results, the first energy consumption prediction result is adjusted to obtain the final energy consumption prediction result. Example: If the first energy consumption prediction result is 100kW, based on the attribution analysis results (+15kW and +25kW), the final energy consumption prediction result is 140kW.

[0055] The Shapley value is used to measure the contribution of each input feature to the model's prediction results. By calculating the marginal contribution of each feature to the model output under different combinations, the Shapley value can assign a unique and fair contribution value to each feature, reflecting its importance to the prediction result.

[0056] In one embodiment, if Figure 5 As shown, in step S40, the energy consumption prediction result is compared with the preset energy consumption threshold to obtain an energy consumption comparison result, and the working states of the energy storage unit and the heat dissipation unit are dynamically adjusted according to the energy consumption comparison result, including: S41: If the energy consumption prediction result exceeds the preset energy consumption threshold, the energy storage unit is turned on in advance to reserve energy and the working efficiency of the heat dissipation unit is improved.

[0057] Specifically, the energy consumption forecast result is compared with the preset energy consumption threshold. Determine whether the preset energy consumption threshold is exceeded, and check whether the energy consumption forecast result is greater than the preset energy consumption threshold. If the energy consumption forecast result is greater than the preset energy consumption threshold, it indicates that the energy consumption of the data center in the future will be high, and measures need to be taken in advance. Turn on the energy storage unit in advance: when the energy consumption forecast result exceeds the preset energy consumption threshold, turn on the energy storage unit in advance. The energy storage unit can store energy in the working state to cope with high energy consumption needs in the future. The specific operations of turning on the energy storage unit in advance include starting the energy storage device and storing enough energy within an appropriate time. Improve the working efficiency of the heat dissipation unit: in order to cope with the upcoming high energy consumption situation, improve the working efficiency of the heat dissipation unit to ensure that the temperature of the data center is controlled within a reasonable range. Improving the working efficiency of the heat dissipation unit includes increasing the working frequency of the heat dissipation unit and increasing the circulation speed of the cooling medium. S42: If the energy consumption prediction result is lower than the preset energy consumption threshold, the energy reserve of the energy storage unit is reduced, and the operating frequency of the heat dissipation unit is reduced.

[0058] Specifically, the energy consumption prediction result is compared with the preset energy consumption threshold in the control system to generate a comparison result. Determine whether it is lower than the preset energy consumption threshold. According to the comparison result, if the energy consumption prediction result is lower than the preset energy consumption threshold, the energy-saving mode is triggered. Reduce the energy reserve of the energy storage unit: The control system sends a signal to reduce the operating frequency of the energy storage device or shut down some energy storage units, and record the operation time and state changes. Reduce the operating frequency of the heat dissipation unit: The control system adjusts the operating parameters of the heat dissipation unit, such as reducing the circulation speed of the cooling medium, reducing the operating time of the heat dissipation device, or shutting down some heat dissipation units.

[0059] In one embodiment, if Figure 6 As shown, after step S42, the intelligent data center energy management method includes: S421: Dynamically adjust the working status of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result. Use the DQN control algorithm to accurately control the energy storage unit and the heat dissipation unit.

[0060] In this embodiment, the DQN control algorithm is to dynamically adjust the working states of the energy storage unit and the heat dissipation unit by continuously learning and optimizing the control strategy, thereby optimizing energy utilization efficiency and cooling effect.

[0061] Specifically, a state space is defined, including current power consumption (kW), ambient temperature (°C), humidity (%), server load (%), energy storage unit status (on / off), cooling unit status (on / off), etc. An action space is defined, including operations on the energy storage unit (on / off, increasing or decreasing the amount of stored energy) and operations on the cooling unit (on / off, adjusting cooling intensity). A reward function is defined: the reward function evaluates the effectiveness of each action, and its value is based on energy efficiency, cooling effectiveness, and system stability. The reward function formula is: R(s,a) = α·Energy_Efficiency + β·Cooling_Effectiveness - γ·Operational_Cost, where α, β, and γ are weighting coefficients. Various data center status parameters, such as power consumption, ambient temperature, and humidity, are monitored and recorded in real time. Actions are selected based on the current policy and executed in the actual environment, such as turning on or adjusting the energy storage unit and cooling unit. After the action is executed, system feedback, such as energy efficiency and cooling effectiveness, is recorded, and the corresponding reward value is calculated. The current state, selected action, reward, and next state are stored in the experience replay memory. A batch of samples are randomly drawn from the experience replay memory to break the data correlation and smooth the data distribution. The target network is used to calculate the maximum Q value of the next state and the target Q value is calculated based on the current reward. The target Q value is: Where γ is the discount factor. Update the Q network and minimize the gap between the current Q network's predicted Q value and the target Q value through the back propagation algorithm, and adjust the weight of the Q network: Minimize the gap between predicted and target Q values. Every fixed number of steps, the Q network weights are copied to the target network to stabilize the training process. At each time step, the optimized strategy is used to select the optimal action based on the current state. The operating states of the energy storage and cooling units are dynamically adjusted to achieve optimal energy utilization and cooling performance.

[0062] In one embodiment, if Figure 7 As shown, after step S421, the working states of the energy storage unit and the heat dissipation unit are dynamically adjusted according to the energy consumption comparison result using the DQN control algorithm, and the precise control of the energy storage unit and the heat dissipation unit also includes: S4211: The energy storage unit is used to store electric energy during the low power demand period and release electric energy during the peak power demand period.

[0063] Specifically, during periods of low electricity demand (such as at night or when electricity prices are low), the energy storage unit is started to convert electrical energy into energy storage media and store it. During periods of peak electricity demand (such as during the day or when electricity prices are high), the stored electrical energy is released for use by the data center to reduce the power load during peak periods. Based on real-time monitoring data (including real-time power consumption data and real-time ambient temperature data), the working status of the energy storage unit is adjusted: when it is predicted that future electricity demand will reach a peak, the stored electrical energy is released in advance to cope with the upcoming high load. When it is predicted that future electricity demand will decrease, the energy reserve of the energy storage unit is reduced to save energy and reduce operating costs.

[0064] S4212: Cooling unit, used to dissipate heat when data center server equipment is running.

[0065] Specifically, when the temperature exceeds a preset safety threshold, the control system automatically increases the working intensity of the heat dissipation unit, such as increasing the fan speed or increasing the flow rate of the coolant. When the temperature drops: When the temperature of the server equipment drops to a safe range, the control system reduces the working intensity of the heat dissipation unit to save energy. At the same time, the heat dissipation unit works in conjunction with the energy storage unit. During peak power demand periods, the heat dissipation unit continues to operate with the power provided by the energy storage unit to ensure the heat dissipation effect. During low power demand periods, the operating frequency of the heat dissipation unit can be appropriately reduced, while still ensuring that the server equipment is within a safe temperature range.

[0066] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0067] In one embodiment, an intelligent data center energy management device is provided, which corresponds to the intelligent data center energy management method in the above embodiment. Figure 8 As shown in Figure 1, the intelligent data center energy management device includes a data acquisition module, an LSTM model module, an attribution model module, and a result comparison and state adjustment module. The functional modules are described in detail as follows: The data acquisition module is used to acquire energy data, including real-time power consumption data and real-time ambient temperature data of the data center. The LSTM model module is used to predict energy data using the trained LSTM model to obtain the first energy consumption prediction result; an attribution model module, configured to optimize and adjust the first energy consumption prediction result by using a trained attribution model to obtain an energy consumption prediction result; The result comparison and state adjustment module is used to compare the energy consumption prediction result with the preset energy consumption threshold to obtain the energy consumption comparison result, and dynamically adjust the working state of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result.

[0068] Optionally, the data acquisition module also includes: A historical data acquisition module is used to obtain historical power consumption data and historical ambient temperature data, and divide the historical power consumption data and historical ambient temperature data into time series to obtain a training data set and a verification data set; The adjustment and optimization module is used to adjust and optimize the parameters of the preset LSTM model and the preset attribution model using the training dataset; The model training completion module is used to evaluate the preset LSTM model and the preset attribution model using the validation dataset to obtain the trained LSTM model and the trained attribution model.

[0069] Optionally, the data acquisition module includes: A first prediction submodule is used to input real-time power consumption data and real-time ambient temperature data into the trained LSTM model to obtain a first energy consumption prediction result; An attribution analysis result submodule, configured to input the first energy consumption prediction result, historical power consumption data, and ambient temperature data into a trained attribution model to obtain an attribution analysis result; The prediction result submodule is used to optimize and adjust the first energy consumption prediction result according to the attribution analysis result to obtain the energy consumption prediction result.

[0070] Optionally, the attribution analysis result submodule includes: Algorithm unit, used for attribution model to perform attribution analysis using the SHAP method.

[0071] Optionally, the result comparison and status adjustment module includes: The activation submodule is used to activate the energy storage unit in advance to reserve energy and improve the working efficiency of the heat dissipation unit if the energy consumption prediction result exceeds the preset energy consumption threshold; The reduction submodule is used to reduce the energy reserve of the energy storage unit and the operating frequency of the heat dissipation unit if the energy consumption prediction result is lower than the preset energy consumption threshold.

[0072] Optionally, the result comparison and status adjustment module also includes: The DQN control algorithm module is used to dynamically adjust the working status of the energy storage unit and the heat dissipation unit according to the energy consumption comparison results. The DQN control algorithm is used to accurately control the energy storage unit and the heat dissipation unit.

[0073] Optionally, the DQN control algorithm module also includes: Energy storage units, which store electricity during periods of low electricity demand and release it during periods of peak demand; The heat dissipation unit is used to dissipate heat when the server equipment in the data center is running.

[0074] The specific definition of the intelligent data center energy management device can be found in the definition of the intelligent data center energy management method above and will not be repeated here. Each module in the above-mentioned intelligent data center energy management device can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each of the above modules.

[0075] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 9 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used for data center management database. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements an intelligent data center energy management method.

[0076] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are performed: Acquire energy data, including collecting and monitoring the data center's real-time power consumption data and real-time ambient temperature data; The energy data is predicted through the trained LSTM model to obtain the first energy consumption prediction result; The first energy consumption prediction result is optimized and adjusted through the trained attribution model to obtain the energy consumption prediction result; The energy consumption comparison result is obtained by comparing the preset energy consumption threshold with the energy consumption prediction result, and the working status of the energy storage unit and the heat dissipation unit is dynamically adjusted according to the energy consumption comparison result.

[0077] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: Acquire energy data, including collecting and monitoring the data center's real-time power consumption data and real-time ambient temperature data; The energy data is predicted through the trained LSTM model to obtain the first energy consumption prediction result; The first energy consumption prediction result is optimized and adjusted through the trained attribution model to obtain the energy consumption prediction result; The energy consumption comparison result is obtained by comparing the preset energy consumption threshold with the energy consumption prediction result, and the working status of the energy storage unit and the heat dissipation unit is dynamically adjusted according to the energy consumption comparison result.

[0078] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0079] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0080] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. An intelligent data center energy management method, characterized in that: The intelligent data center energy management method includes: Acquiring energy data, including collecting and monitoring real-time power consumption data and real-time ambient temperature data of the data center; The energy data is predicted using the trained LSTM model to obtain a first energy consumption prediction result; Optimizing and adjusting the first energy consumption prediction result using the trained attribution model to obtain an energy consumption prediction result; The energy consumption prediction result is compared with the preset energy consumption threshold to obtain an energy consumption comparison result, and the working states of the energy storage unit and the heat dissipation unit are dynamically adjusted according to the energy consumption comparison result.

2. The intelligent data center energy management method according to claim 1, characterized in that: The energy data acquisition includes collecting and real-time monitoring of the power consumption data and ambient temperature data of the data center, and further includes: Acquire historical power consumption data and historical ambient temperature data, and divide the historical power consumption data and the historical ambient temperature data into time series to obtain a training data set and a validation data set; Using the training data set to adjust and optimize parameters of a preset LSTM model and a preset attribution model; The preset LSTM model and the preset attribution model are evaluated using the verification data set to obtain the trained LSTM model and the trained attribution model.

3. The intelligent data center energy management method according to claim 2, characterized in that: The intelligent data center energy management method includes: Inputting the real-time power consumption data and the real-time ambient temperature data into the trained LSTM model to obtain the first energy consumption prediction result; Inputting the first energy consumption prediction result, the historical power consumption data, and the ambient temperature data into the trained attribution model to obtain an attribution analysis result; According to the attribution analysis result, the first energy consumption prediction result is optimized and adjusted to obtain the energy consumption prediction result.

4. The intelligent data center energy management method according to claim 3, characterized in that: The step of inputting the first energy consumption prediction result, the historical power consumption data, and the ambient temperature data into the trained attribution model to obtain an attribution analysis result includes: The attribution model uses the SHAP method to perform attribution analysis.

5. The intelligent data center energy management method according to claim 1, characterized in that: The step of comparing the energy consumption prediction result with the preset energy consumption threshold to obtain an energy consumption comparison result, and dynamically adjusting the working states of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result includes: If the energy consumption prediction result exceeds the preset energy consumption threshold, the energy storage unit is activated in advance to reserve energy and the working efficiency of the heat dissipation unit is improved; If the energy consumption prediction result is lower than the preset energy consumption threshold, the energy reserve of the energy storage unit is reduced and the operating frequency of the heat dissipation unit is reduced.

6. The intelligent data center energy management method according to claim 5, characterized in that: The intelligent data center energy management method includes: The working states of the energy storage unit and the heat dissipation unit are dynamically adjusted according to the energy consumption comparison result, and a DQN control algorithm is used to accurately control the energy storage unit and the heat dissipation unit.

7. The intelligent data center energy management method according to claim 6, characterized in that: The method of dynamically adjusting the working states of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result and using a DQN control algorithm to accurately control the energy storage unit and the heat dissipation unit further includes: Energy storage units, which store electricity during periods of low electricity demand and release it during periods of peak demand; The heat dissipation unit is used to dissipate heat when the server equipment in the data center is running.

8. An intelligent data center energy management device, characterized in that: The intelligent data center energy management device includes: A data acquisition module is used to acquire energy data, including real-time power consumption data and real-time ambient temperature data of the data center. An LSTM model module is used to predict the energy data using a trained LSTM model to obtain a first energy consumption prediction result; an attribution model module, configured to optimize and adjust the first energy consumption prediction result using a trained attribution model to obtain an energy consumption prediction result; The result comparison and state adjustment module is used to compare the energy consumption prediction result with the preset energy consumption threshold to obtain the energy consumption comparison result, and dynamically adjust the working state of the energy storage unit and the heat dissipation unit according to the energy consumption comparison result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the intelligent data center energy management method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the intelligent data center energy management method according to any one of claims 1 to 7 are implemented.

Images