Remote intelligent operation and maintenance management method and system for data center

Abstract

The invention provides a remote intelligent operation and maintenance management method and system for a data center, and relates to the technical field of data center operation and maintenance, and the method comprises the steps: obtaining state parameter monitoring data of the data center, calculating an energy efficiency index, and screening out a strong correlation characteristic parameter; an initial population is randomly generated based on the strong correlation characteristic parameters, an energy efficiency index prediction model is constructed and trained based on a long short-term memory network, and an optimal characteristic parameter combination is determined by using a genetic algorithm based on a model training feedback result so as to be applied to the model; according to the method, various monitoring data are comprehensively collected to fully consider the influence of the environment, the battery and the cold and hot channel conditions on energy consumption, and based on the cooperative setting of the long and short term memory network and the genetic algorithm, the model calculation amount is reduced while the related factors are comprehensively considered, and the operation and maintenance management work order is generated. The accuracy of energy efficiency index prediction is improved, and a basis is provided for subsequent accurate operation and maintenance management.

Description

Technical Field

[0001] The present invention relates to the technical field of data center operation and maintenance, and specifically provides a remote intelligent operation and maintenance management method and system for a data center. Background Art

[0002] With the rapid development of the electronic information industry, behind the efficient and convenient industry application systems are countless servers, network devices, and data storage devices for support. The scale of the communication network is gradually expanding, and the communication network devices are also constantly increasing. The early data centers put into use are facing the problems of optimization and transformation and adapting to subsequent business development. When re-arranging devices such as servers on the cabinets of the original data center or arranging new cabinets in the original data center, the number of devices increases from one to many, and the operation and maintenance work becomes more and more complicated. To make the operation and maintenance management of the data center more effective, it is necessary to innovate the operation and maintenance management means. How to carry out scientific, reasonable, and efficient operation and maintenance management has become urgent and necessary.

[0003] In the prior art, a visualization operation and maintenance management method and system for a data center with the authorization announcement number of "CN114676862B" (the classification number is G06Q) includes: obtaining the device and facility information, pipeline information, and space layout information of the data center, and obtaining the three-dimensional visualization model of the data center; performing device deployment planning and wiring planning through the three-dimensional visualization model, and obtaining the operation and maintenance data information of the data center, performing energy consumption analysis according to the operation and maintenance data information, and formulating an intelligent control scheme through the energy consumption analysis; classifying the operation and maintenance data, and comparing and analyzing the data of the same type at different time periods to detect the faults of the data center; obtaining the type and location information of the faults to generate fault warnings and solutions, and marking and displaying the fault warnings and solutions in the three-dimensional visualization model. This method realizes the intelligent operation and maintenance of the data center through the three-dimensional visualization display of the operation and maintenance status and operation and maintenance data of the data center, reduces the human resource allocation, and improves the operation and maintenance work efficiency.

[0004] However, the prior art still has relatively large defects. For example, when performing energy consumption analysis in the prior art, only historical energy consumption data is used to predict future energy consumption data, and the environmental, battery, and cold and hot channel conditions affecting energy consumption are not fully considered, resulting in a certain deviation between the finally predicted energy consumption data and the true value. Furthermore, the accuracy of generating operation and maintenance management work orders according to the energy consumption analysis results decreases, causing waste of human and material resources and being unfavorable to the safe operation of the data center.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a remote intelligent operation and maintenance management method and system for a data center to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions: A remote intelligent operation and maintenance management method for a data center, comprising: S1, obtaining status parameter monitoring data of the data center during a monitoring time period, which includes energy consumption monitoring data, environmental monitoring data, battery monitoring data, and hot and cold aisle monitoring data; S2, calculating energy efficiency indicators of the data center during the monitoring time period based on the energy consumption monitoring data, including power usage effectiveness, partial power usage effectiveness, and water usage effectiveness; S3, performing feature extraction on the status parameter monitoring data to construct an initial feature library including multiple feature parameters, and the feature extraction includes descriptive statistics and combined interaction; S4, randomly splicing the feature parameters to obtain multiple combined feature parameters, and evaluating the correlation between the energy efficiency indicators and each feature parameter and combined feature parameter based on the Pearson correlation coefficient, and screening out strongly correlated feature parameters and combined feature parameters as strongly correlated feature parameters to construct a final feature library; S5, randomly generating an initial population based on the strongly correlated feature parameters in the final feature library, each individual in the initial population corresponds to a feature parameter combination, constructing and training an energy efficiency indicator prediction model based on a long short-term memory network, the input of the model is the energy efficiency indicators and feature parameter combinations of consecutive multiple monitoring time periods, the output is the energy efficiency indicator of the next monitoring time period, and based on the model training feedback result, using a genetic algorithm to optimize the feature parameter combination to obtain an optimal feature parameter combination for application on the model to complete model training; S6, obtaining the predicted value of the energy efficiency indicator of the future monitoring time period based on the trained energy efficiency indicator prediction model, and generating an operation and maintenance management work order based on the predicted value of the energy efficiency indicator.

[0008] Further, the energy consumption monitoring data includes the water consumption of the data center, the power consumption of each functional area of the data center, and the power consumption of IT devices in each functional area. The functional areas of the data center include a server room, a cooling system area, a power distribution room, a storage area, a network equipment area, a management and operation and maintenance area, and the IT devices include servers, storage devices, network devices, and other IT-related devices; The environmental monitoring data is the external environmental temperature time series data, the temperature time series data of each functional area, the humidity time series data, the air particulate matter concentration time series data, and the air velocity time series data during the monitoring time period; The battery monitoring data are the working temperature time series data, working voltage time series data, charging current time series data, discharging current time series data of each battery cell within the monitoring time period, as well as the cumulative charge and discharge cycle times and the maximum available capacity of each battery cell at the beginning and end moments of the monitoring time period; The hot and cold channel monitoring data include the temperature time series data at the outlets of the hot and cold channels, and the coolant flow rate time series data and temperature time series data at the inlets and outlets of the coolant channels.

[0009] Further, the calculation formula of the energy efficiency index is as follows: In the formula, is the power consumption of the data center within the monitoring time period, is the power consumption of the IT equipment in the data center within the monitoring time period, is the power usage effectiveness of the data center within the monitoring time period; In the formula, is the power consumption of the th functional area in the data center within the monitoring time period, is the power consumption of the IT equipment in the th functional area in the data center within the monitoring time period, is the partial power usage effectiveness of the th functional area in the data center within the monitoring time period, is the index of the functional area in the data center, and , is the number of functional areas in the data center; In the formula, is the water consumption of the data center within the monitoring time period, is the water usage effectiveness of the data center within the monitoring time period.

[0010] Further, the steps for constructing the initial feature library are as follows: Step S31: Extract the energy consumption monitoring data in the status parameter monitoring data as the primary energy consumption feature group, and perform data processing on the energy consumption monitoring data to obtain the secondary energy consumption feature group for characterizing the energy consumption distribution of each functional area in the data center; Step S32: Perform descriptive statistics on the environmental monitoring data to obtain the environmental statistical feature group, and combine the interactive temperature, humidity, and particulate matter concentration to obtain the environmental interaction feature group; Step S33: Perform descriptive statistics on the battery monitoring data to obtain the battery parameter statistical feature group, and combine the interactive current, voltage, charge and discharge cycle times, and capacity to obtain the battery parameter interaction feature group; Step S34: Conduct descriptive statistics on the monitoring data of the hot and cold channels to obtain a statistical feature group of the hot and cold channels, and combine temperature and coolant flow rate to obtain an interaction feature group of the hot and cold channels; Step S35: Statistically analyze the characteristic parameters in the primary energy consumption feature group, secondary energy consumption feature group, environmental statistical feature group, environmental interaction feature group, battery parameter statistical feature group, battery parameter interaction feature group, hot and cold channel statistical feature group, and hot and cold channel interaction feature group to form an initial feature library.

[0011] Furthermore, the method of randomly splicing the characteristic parameters is as follows: Randomly select no less than two characteristic parameters from the initial feature library for random calculation, and the random calculation includes, but is not limited to, calculating the selected characteristic parameters using addition, subtraction, multiplication, and division.

[0012] Furthermore, the judgment criterion for strong correlation is as follows: If the absolute value of the Pearson correlation coefficient between a characteristic parameter and any energy efficiency index is greater than the correlation threshold, it is determined that the definition of this characteristic parameter is strongly correlated with the energy efficiency index, and this characteristic parameter is a strongly correlated characteristic parameter. The method for judging whether a combined characteristic parameter is a strongly correlated characteristic parameter is the same.

[0013] Furthermore, each individual in the initial population consists of K gene positions, and each gene position corresponds to a strongly correlated characteristic parameter. If the digital label of the k-th gene position in an individual is 1, it means that the characteristic parameter combination corresponding to this individual includes the strongly correlated characteristic parameter corresponding to the k-th gene position. Conversely, if the digital label of the k-th gene position in an individual is 0, it means that the characteristic parameter combination corresponding to this individual does not include the strongly correlated characteristic parameter corresponding to the k-th gene position. k is the index of the gene position, and , where K represents the number of strongly correlated characteristic parameters in the final feature library.

[0014] Furthermore, the method for obtaining the optimal characteristic parameter combination is as follows: Step S51: Initialize the mutation probability of each gene position in an individual, and perform data preprocessing on the energy efficiency index and strongly correlated characteristic parameters in each historical monitoring time period to form a sample set. The data preprocessing includes removing outliers, filling in missing values, and data normalization processing; Step S52: Construct an energy efficiency index prediction model based on a long short-term memory network, and set the batch size, feedback training period, and total training period. The feedback training period is set between one-tenth and one-fifth of the total training period. The energy efficiency index prediction model includes an input layer, one or more LSTM layers, a fully connected layer, and an output layer; Step S53: Based on the characteristic parameter combinations corresponding to each individual, randomly select a training set, a validation set, and a test set from the sample set. Input the training sets corresponding to each individual into different energy efficiency index prediction models one by one for training. When the predetermined feedback training cycle is reached, use the validation sets corresponding to each individual to verify each energy efficiency index prediction model to obtain a model training feedback result. The model training feedback result includes the mean square error, the mean absolute error, and the coefficient of determination; Step S54: Based on the model training feedback result, calculate the fitness value of each individual. The larger the fitness value, the better the characteristic parameter combination corresponding to the individual. The calculation formula is as follows: In the formula, , , are the mean square error, the mean absolute error, and the coefficient of determination respectively, , , are all preset weights, and , , The specific values of are determined by the analytic hierarchy process, is the fitness value of the individual; Step S55: Sort each individual in descending order of fitness value, select the individuals in the top 50% as the parent generation, perform crossover and mutation operations between every two of the parent generation to generate offspring, and the mutation probability is determined based on the fitness value of the parent generation and the occurrence frequency of the digital labels of each gene locus. Calculate the fitness value of the offspring based on the same method, and repeat the selection, crossover, and mutation operations for the new population formed by the offspring and the parent generation to generate new parent generations and offspring until the predetermined number of iterative optimizations is reached. Select the individual with the largest fitness value as the optimal individual, and the characteristic parameter combination corresponding to the optimal individual is the optimal characteristic parameter combination.

[0015] Further, the method for determining the mutation probability is as follows: Step S551: Calculate the occurrence frequency of the digital labels of each gene locus in the parent generation. The calculation formula is as follows In the formula, represents the number of occurrences of the digital label 1 at the k-th gene locus in all parent generations during the t-th iterative optimization, represents the total number of parent generations during the t-th iterative optimization, represents the occurrence frequency of the digital label 1 at the k-th gene locus during the t-th iterative optimization, represents the number of occurrences of the digital label 0 at the k-th gene locus in all parent generations during the t-th iterative optimization, represents the occurrence frequency of the digital label 0 at the k-th gene position during the t-th iteration optimization, where t is the index of the iteration optimization times, and , is the total number of iteration optimizations; Step S552: Based on the fitness values of the parent generation, calculate the initial mutation probability of each gene position in each round of iteration optimization. The calculation formula is as follows: In the formula, is the initial mutation probability of the k-th gene position during the t-th iteration optimization, is the average fitness of all parent generations during the t-th iteration optimization, is the maximum value of the fitness values of all parent generations during the process from the 1st iteration optimization to the t-th iteration optimization, is the maximum value of the fitness values of all parent generations during the t-th iteration optimization, is a piecewise function, is the fitness evaluation threshold, is the randomly introduced perturbation amount during the t-th iteration optimization; Step S553: Based on the initial mutation probability and the occurrence frequency of the digital labels of each gene position in the parent generation, generate the final mutation probability. The calculation formula is as follows: In the formula, is the final mutation probability of changing the digital label of the k-th gene position from 0 to 1 during the t-th iteration optimization, is the final mutation probability of changing the digital label of the k-th gene position from 1 to 0 during the t-th iteration optimization.

[0016] A remote intelligent operation and maintenance management system for a data center, used to execute the above-mentioned remote intelligent operation and maintenance management method for a data center, including: A data monitoring module, used to obtain the status parameter monitoring data of the data center during the monitoring time period, including energy consumption monitoring data, environmental monitoring data, battery monitoring data, and hot and cold aisle monitoring data; An energy efficiency index calculation module, based on the energy consumption monitoring data, calculates the energy efficiency indexes of the data center during the monitoring time period, including power usage effectiveness, partial power usage effectiveness, and water usage effectiveness; An initial feature library construction module, used to extract features from the status parameter monitoring data to construct an initial feature library including multiple feature parameters, and the feature extraction includes descriptive statistics and combined interaction; The final feature library construction module is used to randomly splice feature parameters to obtain multiple combined feature parameters, and evaluate the correlation between the energy efficiency index and each feature parameter and combined feature parameter based on the Pearson correlation coefficient, and screen out strongly correlated feature parameters and combined feature parameters as strongly correlated feature parameters to construct the final feature library; The prediction model construction module randomly generates an initial population based on the strongly correlated feature parameters in the final feature library. Each individual in the initial population corresponds to a feature parameter combination. An energy efficiency index prediction model is constructed and trained based on the long short-term memory network. The input of this model is the energy efficiency index and feature parameter combination of multiple consecutive monitoring time periods, and the output is the energy efficiency index of the next monitoring time period. Based on the model training feedback results, the genetic algorithm is used to optimize the feature parameter combination to obtain the optimal feature parameter combination for application on the model to complete model training; The operation and maintenance work order generation module obtains the predicted value of the energy efficiency index for the future monitoring time period based on the trained energy efficiency index prediction model, and generates an operation and maintenance management work order based on the predicted value of the energy efficiency index.

[0017] Compared with the prior art, the beneficial effects of the present invention are: The remote intelligent operation and maintenance management method and system for a data center according to the present invention fully collects various monitoring data to fully consider the impact of the environment, battery, and cold and hot channel conditions on energy consumption. Then, strongly correlated feature parameters with the energy efficiency index are obtained based on various monitoring data. Furthermore, through the coordinated setting of the long short-term memory network and the genetic algorithm, the optimal feature parameter combination is obtained. Finally, the energy efficiency index prediction model with the optimal feature parameter combination as the input is used to predict the energy efficiency index. While comprehensively considering relevant factors, the model calculation amount is reduced, the accuracy of energy efficiency index prediction is greatly improved, and the computing power cost is saved, providing a basis for accurate operation and maintenance management in the future and ensuring the safe and efficient operation of the data center. Description of the Drawings

[0018] Figure 1 It is a flow schematic diagram of the overall method of the present invention; Figure 2 It is a module unit diagram of the overall system of the present invention. Detailed Embodiments

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0020] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those of ordinary skill in the field to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0021] Embodiment: Please refer to Figure 1 , the present invention provides a remote intelligent operation and maintenance management method for a data center, including: Step S1, obtaining status parameter monitoring data of the data center during a monitoring period, which includes energy consumption monitoring data, environmental monitoring data, battery monitoring data, and hot and cold aisle monitoring data; As an implementation manner, the energy consumption monitoring data includes the water consumption of the data center, the power consumption of each functional area of the data center, and the power consumption of IT devices in each functional area; Among them, the water consumption of the data center is obtained through a water meter, and since the water consumption of the data center mainly comes from the cooling system, the water consumption at the cooling system area can be used as the water consumption of the data center. The water consumption at the cooling system area is read through a water meter installed at the cooling system area, and the unit of the water consumption of the data center is liters; Among them, the functional areas of the data center include a server room, a cooling system area, a power distribution room, a storage area, a network equipment area, a management and operation and maintenance area. The general division criteria for functional areas are as follows. Of course, the specific division criteria can also be adjusted by the staff according to the actual situation; Server room: Mainly carry all key IT devices, usually the main source of energy consumption in the data center; Cooling system area: Includes cooling towers and air conditioning equipment, used to maintain the normal working temperature of the equipment; Power distribution room: The area responsible for power distribution and UPS (uninterruptible power supply) equipment, monitoring power quality and power supply stability; Storage area: The area dedicated to storage equipment, involving data backup and management; Network equipment area: The centralized area of network switches, routers and other equipment; Management and operation and maintenance area: The work area for operation and maintenance personnel to manage and monitor the data center; The power consumption of each of the above functional areas is obtained through the data center's own centralized monitoring system or through energy metering devices such as relevant electricity meters, and the unit of power consumption is set to kilowatt-hours; Among them, IT equipment includes servers, storage devices, network devices, and other IT-related devices. Specifically, servers include but are not limited to blade servers, rack servers, tower servers, etc., storage devices include but are not limited to network attached storage (NAS), storage area network (SAN), disk arrays, etc. for storing and managing data, network devices include but are not limited to switches, routers, firewalls, etc. for data transmission and network connection, and other IT-related devices include but are not limited to load balancers, backup devices, virtualization servers, etc. for assisting IT operations or supporting IT infrastructure. The method for obtaining the power consumption of IT equipment is the same as the method for obtaining the power consumption of each functional area in the above text. This is prior art and will not be elaborated here; As an implementation, the environmental monitoring data is the time-series data of the external environmental temperature, the time-series data of the temperature in each functional area, the time-series data of the humidity, the time-series data of the air particulate matter concentration, and the time-series data of the air flow rate during the monitoring period; Among them, the time-series data of the temperature, the time-series data of the humidity, the time-series data of the air particulate matter concentration, and the time-series data of the air flow rate can be obtained through temperature sensors (such as digital thermometers, thermocouples), humidity sensors (such as hygrometers), air quality monitors (such as laser particle counters), and flow sensors (such as anemometers, heat flow meters) installed in the corresponding functional areas. Of course, they can also be obtained through the data center's own centralized monitoring system. The time-series data of the external environmental temperature can be obtained through the weather information released by the meteorological bureau or by installing temperature sensors in the external environment; As an implementation, the battery monitoring data is the time-series data of the working temperature, the working voltage, the charging current, and the discharging current of each battery cell during the monitoring period, as well as the cumulative charge and discharge cycle times and the maximum available capacity of each battery cell at the beginning and end of the monitoring period. And all the above battery monitoring data can be directly obtained through the battery management system (BMS); As an implementation, the cold and hot aisle monitoring data includes the time-series data of the temperature at the cold and hot aisle exits, the time-series data of the coolant flow rate, and the time-series data of the temperature at the inlet and outlet of the coolant channel; It should be noted that the principle of cooling the equipment in the data center is as follows: The coolant channel inlet receives the coolant cooled by the cooling system. The coolant flows in the coolant pipeline and takes away the heat of the air in the hot and cold channels, causing the air in the hot and cold channels to cool down. The cooled air then flows past the equipment to be cooled to take away its heat. The air that has taken away the heat of the equipment to be cooled finally exits from the outlet of the hot and cold channels, and the coolant that has taken away the heat of the air in the hot and cold channels continues to flow until it re-enters the cooling system through the coolant channel outlet for cooling; Among them, the time-series data of the outlet temperature of the hot and cold channels can be specifically obtained through temperature sensors installed at the outlet of the hot and cold channels. This outlet temperature is the temperature of the air. The time-series data of the inlet and outlet temperatures of the coolant channel is obtained in the same way, and this inlet and outlet temperature is the temperature of the coolant. Flow meters are installed at both the inlet and outlet of the coolant channel to obtain the time-series data of the coolant flow rates at the inlet and outlet of the coolant respectively; It should be noted that the specific monitoring time period is set by the staff according to the actual situation, such as 30 minutes, 1 hour, 2 hours, 6 hours, etc., without limitation here. When collecting the above time-series data, different types of parameters can be collected at the same sampling frequency to unify the time stamps, or after collecting different types of parameters at different sampling frequencies, the time stamps can be unified through interpolation or data aggregation methods to facilitate subsequent processing and analysis.

[0022] Step S2, based on the energy consumption monitoring data, calculate the energy efficiency indicators of the data center during the monitoring time period, including power usage effectiveness, partial power usage effectiveness, and water usage effectiveness. The calculation formulas are as follows: In the formula, is the power consumption of the data center during the monitoring time period, specifically the sum of the power consumptions of each functional area during the monitoring time period, is the power consumption of the IT equipment in the data center during the monitoring time period, specifically the sum of the power consumptions of the IT equipment in each functional area during the monitoring time period, is the power usage effectiveness of the data center during the monitoring time period, The ideal value of is 1, which means that all the power in the data center is used for IT equipment without additional energy consumption. However, in reality, affected by the cooling and other auxiliary systems, it is generally higher than 1, usually between 1.5 and 2.5, and the higher it is, the lower the power usage efficiency of the data center, and the more operation and maintenance adjustment are needed; In the formula, is the power consumption of the th functional area of the data center during the monitoring time period, specifically the sum of the power consumptions of the th functional area during the monitoring time period, During the monitoring period, the power consumption of IT devices in the th functional area of the data center, specifically, during the monitoring period, it is the sum of the power consumption of IT devices in the th functional area, During the monitoring period, the partial power usage efficiency of the th functional area of the data center, The ideal value of which is 1, meaning that all the power in the th functional area of the data center is used for the IT devices in that area without additional energy consumption, while the higher it is, the lower the power usage efficiency of the th functional area of the data center, and the more operation and maintenance adjustment are needed. is the index of the functional area in the data center, and , is the number of functional areas in the data center; In the formula, is the water consumption of the data center during the monitoring period, is the water usage efficiency of the data center during the monitoring period, The lower the value, the higher the water usage efficiency of the data center while maintaining performance. Similarly, the higher it is, the lower the water usage efficiency of the data center during the monitoring period, and the more operation and maintenance adjustment are needed.

[0023] Step S3: Extract features from the status parameter monitoring data to construct an initial feature library including multiple feature parameters. The feature extraction includes descriptive statistics and combined interaction, as follows: Step S31: Extract the energy consumption monitoring data in the status parameter monitoring data as the primary energy consumption feature group, and perform data processing on the energy consumption monitoring data to obtain the secondary energy consumption feature group used to characterize the energy consumption distribution of each functional area in the data center. The primary energy consumption feature group and the secondary energy consumption feature group are represented as follows: In the formula, is the primary energy consumption feature group, which is used to characterize the power consumption of each functional area in the data center, the power consumption of IT devices in each functional area, and the water consumption of the data center during the monitoring period. The features in the primary energy consumption feature group are an important basis for calculating the energy efficiency index, and they provide an analysis and prediction basis for the model to predict future energy efficiency indexes; In the formula, It is a secondary energy consumption feature group, which is used to characterize the proportion of power consumption of each functional area in the power consumption of the data center during the monitoring period, as well as the proportion of power consumption of IT equipment in each functional area in the power consumption of IT equipment in the data center. The energy consumption distribution of each functional area in the data center is characterized by the characteristics in the secondary energy consumption feature group, which provides an analytical prediction basis for the model in the following text to predict future energy efficiency indicators, and The larger the The higher the proportion of power consumption of each functional area in the power consumption of the data center, the The larger the The higher the proportion of IT equipment in each functional area in the power consumption of IT equipment in the data center; Step S32, perform descriptive statistics on the environmental monitoring data to obtain an environmental statistical feature group, and combine the interactive temperature, humidity and particulate matter concentration to obtain an environmental interactive feature group. The environmental statistical feature group and the environmental interactive feature group are expressed as follows: In the formula, are the mean, maximum, minimum and standard deviation of the external ambient temperature during the monitoring period, They are respectively The mean, maximum, minimum and standard deviation of the temperature in each functional area. They are respectively The mean, maximum, minimum and standard deviation of humidity in each functional area. They are respectively The mean, maximum, minimum and standard deviation of the concentration of air particles in each functional area. They are respectively The mean, maximum, minimum and standard deviation of air velocity in each functional area, The environmental statistical feature group is used to characterize the environmental conditions that affect the operation of the data center during the monitoring period; In the formula, Used to characterize the monitoring period. The greater the temperature difference between the temperature in the functional area and the external environment temperature, the greater the temperature difference. Functional areas maintain temperature The more energy is consumed, Used to characterize the effect of temperature and humidity on the The impact of temperature and humidity on equipment performance is often interactive, so the multiplication result of the two is used as a combined feature to capture this complex relationship. For characterizing the influence of temperature and air particulate matter concentration factors on the equipment at the nth functional area during the monitoring time period. Since an increase in temperature usually leads to an increase in the concentration of particulate matter in the air, and particulate matter affects the scattering and absorption of thermal radiation, thereby affecting the temperature of the local area, and both will affect the operating performance of the equipment, the result of multiplying the two is used as a combined feature to facilitate capturing this complex relationship. It is an environmental interaction feature group, which is used to characterize the interactive influence of temperature, humidity, and air particulate matter concentration on the equipment operation during the monitoring time period; Step S33: Perform descriptive statistics on the battery monitoring data to obtain a battery parameter statistical feature group, and combine the interactive current, voltage, charge and discharge cycle times, and capacity to obtain a battery parameter interactive feature group. The battery parameter statistical feature group and the battery parameter interactive feature group are represented as follows: In the formula, are respectively the mean, maximum value, minimum value, and standard deviation of the working temperature of the nth battery cell during the monitoring time period, are respectively the mean, maximum value, minimum value, and standard deviation of the working voltage of the nth battery cell during the monitoring time period, are respectively the mean, maximum value, minimum value, and standard deviation of the charging current of the nth battery cell during the monitoring time period, are respectively the mean, maximum value, minimum value, and standard deviation of the discharging current of the nth battery cell during the monitoring time period, is the battery parameter statistical feature group, which is used to characterize the working state of each battery cell during the monitoring time period, is the index of the battery cell in the data center, and , is the number of battery cells in the data center; In the formula, , , , are respectively the rated working temperature, rated working voltage, rated charging current, and rated discharging current of the nth battery cell, The greater the deviation from 1, the less suitable the working temperature of the nth battery cell during the monitoring time period. Similarly, , , The greater the deviation from 1, the more it indicates that during the monitoring time period, the The more unsuitable the working voltage, charging current, and discharging current of a single battery cell are, the specific values of the rated working temperature, rated working voltage, rated charging current, and rated discharging current can be obtained from the product specification sheet of this single battery cell; In the formula, used to characterize the charging and discharging current difference situation of the th single battery cell during the monitoring period. If the discharging current is significantly lower than the charging current, it indicates an increase in the internal impedance of the battery or a decrease in capacity, and further indicates that the aging situation of this single battery cell is more serious. Therefore, through to quantify the charging and discharging current difference situation of the th single battery cell during the monitoring period, and the larger the value, the more serious the aging of the jth single battery cell and the shorter the remaining available life; In the formula, are respectively the maximum available capacities of the th single battery cell at the start and end moments of the monitoring period, used to characterize the capacity lost by the th single battery cell during the monitoring period. The smaller its value, the more capacity the th single battery cell has lost during the monitoring period, and the more serious the aging situation is, is the rated capacity of the th single battery cell, and the specific value can be obtained from the product specification sheet of this single battery cell, used to characterize the available capacity ratio of the th single battery cell after the monitoring period. The larger its value, the smaller the lost capacity of the th single battery cell after the monitoring period, and the longer the available life of this single battery cell; In the formula, are respectively the cumulative charge and discharge cycle numbers of the th single battery cell at the start and end moments of the monitoring period, used to characterize the charge and discharge cycle number of the th single battery cell during the monitoring period. The larger its value, the higher the charge and discharge intensity of the th single battery cell during the monitoring period, and the shorter the remaining available life, is the rated charge and discharge cycle number of the th single battery cell, and the specific value can be obtained from the product specification sheet of this single battery cell, used to characterize the remaining available charge and discharge cycle situation of the th single battery cell after the monitoring period. The larger its value, the The worse the remaining available charge-discharge cycles of a battery cell after the monitoring period, the shorter its remaining available life; Wherein, Used to characterize the remaining available life of the th battery cell after the monitoring period. Since both the battery capacity and the number of charge-discharge cycles can be used to quantitatively evaluate the remaining available life of the battery, combining the two to comprehensively evaluate the battery cell avoids the limitations of single evaluation. And as discussed above, the larger it is, the longer the corresponding remaining available life, the smaller it is, the shorter the corresponding remaining available life. Therefore, the form of is used to evaluate the remaining available life of the battery. The larger its value, the shorter the corresponding remaining available life; is the battery parameter interaction feature group, which is used to combine the interaction current, voltage, number of charge-discharge cycles and capacity to characterize the health state during the operation of the battery; Step S34, perform descriptive statistics on the hot and cold channel monitoring data to obtain the hot and cold channel statistical feature group, and combine the temperature and coolant flow rate to obtain the hot and cold channel interaction feature group. The representations of the hot and cold channel statistical feature group and the hot and cold channel interaction feature group are as follows: Wherein, are respectively the mean, maximum, minimum and standard deviation of the outlet temperature of the hot and cold channels during the monitoring period, are respectively the mean, maximum, minimum and standard deviation of the inlet temperature of the coolant channel during the monitoring period, are respectively the mean, maximum, minimum and standard deviation of the outlet temperature of the coolant channel during the monitoring period, are respectively the mean, maximum, minimum and standard deviation of the coolant flow rate at the inlet of the coolant channel during the monitoring period, are respectively the mean, maximum, minimum and standard deviation of the coolant flow rate at the outlet of the coolant channel during the monitoring period, is the hot and cold channel statistical feature group, which is used to characterize the working state of the cooling system during the monitoring period; Wherein, is used to characterize the temperature difference situation between the inlet and outlet of the coolant channel during the monitoring period. Within a reasonable range, the larger the temperature difference, the more thorough the heat exchange, but too large a temperature difference also indicates the problem of insufficient refrigeration capacity of the cooling system, It is used to characterize the evaporation or leakage of the coolant in the coolant channel during the monitoring period. The larger the value, the greater the evaporation of the coolant during the flow process, which further indicates the lack of refrigeration capacity of the cooling system. If the value is too large and exceeds the normal range, it indicates that there is a risk of leakage in the coolant channel. It is used to characterize the heat exchange effect during the monitoring period. The closer the value is to 0, the more thorough the heat exchange and the higher the heat exchange efficiency. It is used to characterize the heat pollution at the outlets of the hot and cold channels during the monitoring period. The larger the value, the more serious the heat pollution at the outlets of the hot and cold channels, and the more necessary it is to increase the refrigeration capacity of the cooling system. It is a hot and cold channel interaction feature group, which is used to combine the interaction temperature and the coolant flow rate to characterize the working state of the cooling system. Step S35: Statistically analyze the characteristic parameters in the primary energy consumption feature group, secondary energy consumption feature group, environmental statistics feature group, environmental interaction feature group, battery parameter statistics feature group, battery parameter interaction feature group, hot and cold channel statistics feature group, and hot and cold channel interaction feature group to form an initial feature library. The characteristic parameters are the elements in the feature groups. For example, the elements in the primary energy consumption feature group are characteristic parameters. By setting steps S31 - S35, multiple characteristic parameters affecting the energy efficiency index value are summarized, laying a foundation for further expanding the feature library and optimizing model training later.

[0024] S4: Randomly splice the characteristic parameters to obtain multiple combined characteristic parameters, and evaluate the correlation between the energy efficiency index and each characteristic parameter and combined characteristic parameter based on the Pearson correlation coefficient. Select the strongly correlated characteristic parameters and combined characteristic parameters as strongly correlated characteristic parameters to construct the final feature library. Among them, the method of randomly splicing the characteristic parameters is: randomly select no less than two characteristic parameters from the initial feature library for random calculation. The random calculation includes but is not limited to using addition, subtraction, multiplication, and division to perform random calculations on the selected characteristic parameters. For example, is and randomly spliced to obtain a combined characteristic parameter, which can evaluate the stability of the temperature at the th functional area compared with the external environment temperature during the monitoring period. The larger the value, the more stable the temperature at the th functional area compared with the external environment temperature during the monitoring period. The setting of randomly splicing the characteristic parameters to obtain combined characteristic parameters expands the selection range of features, facilitates a comprehensive analysis of relevant data when modeling and predicting the energy efficiency index later, and also reduces the workload of staff in artificially constructing characteristic parameters. Among them, the judgment criterion for strong correlation is as follows: if the absolute value of the Pearson correlation coefficient between a characteristic parameter and any energy efficiency index (such as power usage effectiveness, partial power usage effectiveness, or water usage effectiveness) is greater than the correlation threshold, it is determined that the definition of this characteristic parameter is strongly correlated with the energy efficiency index, and this characteristic parameter is a strongly correlated characteristic parameter. The specific value of the correlation threshold is set by the staff according to the actual situation. For example, considering that the energy efficiency index of the data center is affected by multiple factors in combination, to fully consider relevant characteristics, the value range of the correlation threshold can be set to a relatively low value between 0.1 and 0.3. The method for judging whether a combined characteristic parameter is a strongly correlated characteristic parameter is the same and will not be elaborated here; It should be noted that in step S4, by first randomly splicing to obtain multiple combined characteristic parameters and then screening out strongly correlated characteristic parameters from the characteristic parameters and combined characteristic parameters based on the Pearson correlation coefficient, high-quality characteristic parameters strongly correlated with the energy efficiency index are obtained, laying a foundation for determining the model input in the subsequent steps; Furthermore, to ensure the richness of strongly correlated characteristic parameters in the final feature library, the total number of strongly correlated characteristic parameters in the final feature library is generally set to more than 1.5 times the total number of characteristic parameters in the initial feature library. The specific upper limit can be set by the staff according to the actual situation, but generally it should not exceed 3 times to avoid excessive computational complexity.

[0025] In step S5, an initial population is randomly generated based on the strongly correlated characteristic parameters in the final feature library. Each individual in the initial population corresponds to a combination of characteristic parameters. An energy efficiency index prediction model is constructed and trained based on a long short-term memory network. The input of this model is the energy efficiency index and the combination of characteristic parameters for multiple consecutive monitoring time periods, and the output is the energy efficiency index for the next monitoring time period. Based on the model training feedback results, a genetic algorithm is used to optimize the combination of characteristic parameters to obtain the optimal combination of characteristic parameters for application to the model, thus completing the model training; Among them, each individual in the initial population consists of K gene positions, and each gene position corresponds to a strongly correlated characteristic parameter. If the digital label of the k-th gene position in an individual is 1, it means that the combination of characteristic parameters corresponding to this individual includes the strongly correlated characteristic parameter corresponding to the k-th gene position. Conversely, if the digital label of the k-th gene position in an individual is 0, it means that the combination of characteristic parameters corresponding to this individual does not include the strongly correlated characteristic parameter corresponding to the k-th gene position. k is the index of the gene position, and , where K represents the number of strongly correlated characteristic parameters in the final feature library; Among them, when constructing an energy efficiency index prediction model based on a long short-term memory network, experiments are generally carried out on different window lengths during the training process. The number of consecutive monitoring period data used to predict the data of the next monitoring period is determined by the performance of the validation set. This is the prior art and will not be elaborated here. Of course, the number of consecutive monitoring period data used to predict the data of the next monitoring period can also be set manually. For example, it is determined to use the data of 8 - 12 consecutive monitoring periods to predict the data of the next monitoring period. Taking 10 consecutive monitoring periods as an example, the input of the energy efficiency index prediction model is the combination of the energy efficiency index and characteristic parameters of each monitoring period in 10 consecutive monitoring periods, and the output is the energy efficiency index of the next monitoring period after these 10 consecutive monitoring periods. Of course, the specific number of consecutive monitoring periods can also be other values, which is not limited here; Among them, the method for obtaining the optimal characteristic parameter combination is as follows: Step S51, initialize the mutation probability of each gene position in the individual, and perform data preprocessing on the energy efficiency index and strongly correlated characteristic parameters in each historical monitoring period to form a sample set. The data preprocessing includes removing outliers, filling missing values, and data normalization; Among them, after initialization, the mutation probability of each gene position is between 0.01 - 0.1 to balance population diversity and avoid introducing too much randomness; Among them, conventional algorithms such as the Z-Score method, IQR method, or box method can be used to remove outliers, and conventional algorithms such as the mean filling method or median filling method can be used to fill missing values. However, the optimal choice is to fill with the mean of the previous and next monitoring periods to conform to the time series structure logic. Data normalization can use conventional algorithms such as the maximum-minimum normalization method or Robust normalization. This is the prior art and will not be elaborated here; Step S52, construct an energy efficiency index prediction model based on a long short-term memory network, and set the batch size, feedback training cycle, and total training cycle. The feedback training cycle is set between one-tenth and one-fifth of the total training cycle. The energy efficiency index prediction model includes an input layer, one or more LSTM layers, a fully connected layer, and an output layer; Among them, the input layer is used to receive the combination of the energy efficiency index and characteristic parameters of consecutive multiple monitoring periods. The LSTM layer is used to perform feature processing on the data received by the input layer. The fully connected layer is used to map the output of the LSTM layer to the predicted value, and the output layer is used to output the energy efficiency index of the next monitoring period; Among them, the batch size is the number of samples used in each training iteration, generally ranging from 32 to 256. The total training cycle is the total number of iterative training times, generally ranging from 30 to 100. The setting of the feedback training cycle is used to determine the time node for obtaining the model training feedback result. If this time node is too early, the model training feedback result will lack reference value, and if it is too late, the total model training time will be too long. Therefore, the feedback training cycle is set between one-tenth and one-fifth of the total training cycle to balance the reference value of the model training feedback result and the efficiency of model training; Step S53: Based on the characteristic parameter combinations corresponding to each individual, randomly select a training set, a validation set, and a test set from the sample set. Input the training sets corresponding to each individual into different energy efficiency index prediction models for training one by one. When the predetermined feedback training cycle is reached, use the validation sets corresponding to each individual to verify each energy efficiency index prediction model to obtain the model training feedback result. The model training feedback result includes the mean square error, the mean absolute error, and the coefficient of determination; It should be noted that the model training process, as well as the calculation of the mean square error, the mean absolute error, and the coefficient of determination, are all conventional technical means in this technical field and will not be elaborated here; Step S54: Based on the model training feedback result, calculate the fitness value of each individual. The larger the fitness value, the better the characteristic parameter combination corresponding to the individual. The calculation formula is as follows: In the formula, 、 、 are the mean square error, the mean absolute error, and the coefficient of determination respectively. The smaller the mean square error and the mean absolute error, the better the performance of the energy efficiency index prediction model after the feedback training cycle. And the closer the coefficient of determination is to 1, the better the performance of the energy efficiency index prediction model after the feedback training cycle. Therefore, the mean square error, the mean absolute error, and the coefficient of determination are used together to characterize the performance of the energy efficiency index prediction model; In the formula, 、 、 are all preset weights, which respectively represent the importance of the mean square error, the mean absolute error, and the coefficient of determination in the performance evaluation of the energy efficiency index prediction model. The larger their values, the higher the importance of the corresponding mean square error, mean absolute error, or coefficient of determination in the performance evaluation of the energy efficiency index prediction model, 、 、 The specific values of are determined by the analytic hierarchy process as follows: Determine the relative importance of three evaluation factors, namely mean square error, mean absolute error, and coefficient of determination, pairwise for the performance evaluation of the energy efficiency index prediction model through the nine-scale method to construct a judgment matrix. Among them, the indexes of mean square error, mean absolute error, and coefficient of determination are marked as 1, 2, and 3 respectively. The constructed judgment matrix is as follows: In the formula, both represent the indexes of the evaluation factors, and represents the relative importance of the evaluation factor with index x relative to the evaluation factor with index y for the performance evaluation of the energy efficiency index prediction model, The specific value of is determined by relevant experts using the 1-9 scoring method. The larger indicates that the evaluation factor with index x is more important relative to the evaluation factor with index y for the performance evaluation of the energy efficiency index prediction model, and the smaller In the formula, is the fitness value of an individual, which comprehensively considers three evaluation factors: mean square error, mean absolute error, and coefficient of determination, to evaluate the performance of the energy efficiency index prediction model. The larger the fitness value, the better the combination of characteristic parameters corresponding to the individual, and the more suitable it is as the input parameter of the energy efficiency index prediction model. It should be noted that the smaller the mean square error and mean absolute error, the better the performance of the energy efficiency index prediction model after the feedback training cycle. The closer the coefficient of determination is to 1, the better the performance of the energy efficiency index prediction model after the feedback training cycle. Therefore, through the method of weighted summation, is used as the main part of the fitness value to comprehensively evaluate the performance of the energy efficiency index prediction model. The setting of is used to quantify the deviation degree between the coefficient of determination and the ideal value 1. Also, because Taking 1 as the denominator and 1 as the numerator to establish the function expression of the fitness value. At the same time, the denominator is set to add 1 to avoid the situation where the denominator is 0, so that the finally formed fitness value function expression meets the logical requirement that the smaller the mean square error and the mean absolute error, and the closer the coefficient of determination is to 1, the larger the fitness value. Of course, other reasonable function expressions are also acceptable and are not restricted here; Step S55: Sort each individual in descending order of the fitness value, select the individuals in the top 50% as the parent generation, perform crossover and mutation operations pairwise on the parent generation to generate offspring, and determine the mutation probability based on the fitness value of the parent generation and the occurrence frequency of the digital tags at each gene position. Calculate the fitness value of the offspring based on the same method, and repeat the selection, crossover, and mutation operations on the new population formed by the offspring and the parent generation to generate new parent generations and offspring until the predetermined number of iterative optimizations is reached. Select the individual with the largest fitness value as the optimal individual, and the combination of characteristic parameters corresponding to the optimal individual is the optimal combination of characteristic parameters; Among them, the method for determining the mutation probability is as follows: Step S551: Calculate the occurrence frequency of the digital tags at each gene position in the parent generation. The calculation formula is as follows In the formula, represents the number of occurrences of the digital tag at the k-th gene position being 1 in all parent generations during the t-th iterative optimization; represents the total number of parent generations during the t-th iterative optimization; represents the occurrence frequency of the digital tag at the k-th gene position being 1 during the t-th iterative optimization. The larger its value, the higher the proportion of the digital tag at the k-th gene position being 1 in the parent generation. Since the parent generation is the relatively better individual in the population, it indicates that the possibility that the characteristic parameter corresponding to the k-th gene position is the optimal characteristic parameter is also higher; represents the number of occurrences of the digital tag at the k-th gene position being 0 in all parent generations during the t-th iterative optimization; represents the occurrence frequency of the digital tag at the k-th gene position being 0 during the t-th iterative optimization. The larger its value, the higher the proportion of the digital tag at the k-th gene position being 0 in the parent generation. Since the parent generation is the relatively better individual in the population, it indicates that the possibility that the characteristic parameter corresponding to the k-th gene position is the optimal characteristic parameter is lower. t is the index of the number of iterative optimizations, and , is the total number of iterative optimizations; Step S552: Based on the fitness value of the parent generation, calculate the initial mutation probability of each gene position in each round of iterative optimization. The calculation formula is as follows: In the formula, is the initial mutation probability of the k-th gene position during the t-th iteration optimization. The larger its value, the greater the probability that the digital label of the k-th gene position changes from 0 to 1 or from 1 to 0 during the t-th iteration optimization. Similarly, is the initial mutation probability of the k-th gene position during the (t - 1)-th iteration optimization, where represents the initial mutation probability of the k-th gene position after initialization in step S51; In the formula, is the average fitness of all parents during the t-th iteration optimization, is the maximum fitness value of all parents during the process from the 1st iteration optimization to the t-th iteration optimization, is the maximum fitness value of all parents during the t-th iteration optimization; It should be noted that the larger it is, the closer the selected parents as a whole are to the historical optimum during the t-th iteration optimization, that is, the higher the quality of the selected parents during the t-th iteration optimization. A lower mutation probability should be set to avoid introducing too much randomness and ensure the quality of the offspring generated during the t-th iteration optimization. Similarly, if is smaller, it means that the quality of the selected parents during the t-th iteration optimization is lower. A higher mutation probability should be set to introduce higher randomness to ensure the diversity of the offspring, and then search for the optimal solution in a larger space; On this basis, the larger it is, the closer the average fitness value of the selected parents is to the maximum value during the t-th iteration optimization, that is, it means that the quality of the selected parents is more uniform during the t-th iteration optimization. On the contrary, if is smaller, it means that the quality of the selected parents during the t-th iteration optimization is more uneven; Therefore, when and are both relatively high, it means that the quality of the selected parents during the t-th iteration optimization is high and relatively uniform, that is, it means that the selected parents are basically close to the optimal solution and there is not much room for exploration. Therefore, a lower mutation probability is set through to avoid introducing too much ineffective randomness. And when is relatively high and is relatively low, it means that the quality of the selected parents during the t-th iteration optimization is high and relatively uneven, that is, it means that the selected parents are relatively good but there is still room for optimization. Similarly, a slightly lower mutation probability is set through to introduce an appropriate amount of randomness; On the contrary, when When it is relatively high, it indicates that when performing the t-th iteration optimization, the selected parental generation has poor quality and is relatively uniform, that is, it means that the selected parental generation is far from the optimal solution and lacks diversity at this time, and there is a large exploration space. Therefore, through to set a relatively high mutation probability to introduce a large amount of randomness, and then search for the optimal solution in a larger space. While in and are both relatively low, it indicates that when performing the t-th iteration optimization, the selected parental generation has low quality and is relatively non-uniform, that is, it means that although the selected parental generation is far from the optimal solution at this time, it has high diversity. Therefore, through to set a slightly higher mutation probability to appropriately enrich randomness; Therefore, a piecewise function is set, and in the form of to obtain the mutation probability for the t-th iteration optimization. In the formula, is the fitness evaluation threshold, which is used to evaluate excellence. When , it indicates that the selected parental generation has excellent quality when performing the t-th iteration optimization, while when , it indicates that the selected parental generation has poor quality when performing the t-th iteration optimization. The specific value of the fitness evaluation threshold is set by the staff according to the actual situation. For example, the value range of the fitness evaluation threshold can be set between 0.4 - 0.7. It should be noted that setting the fitness evaluation threshold too high will introduce too much randomness, and too low is likely to fall into a local optimal solution. It is recommended to adjust it dynamically according to the actual situation; In the formula, is the random perturbation amount introduced when performing the t-th iteration optimization, and its value range is between 0.8 - 1.2, which is used to randomly change the mutation probability to ensure randomness. Of course, taking other values close to 1 is also acceptable and is not restricted here; Step S553, based on the initial mutation probability and the occurrence frequency of the digital tags of each gene position in the parental generation, generate the final mutation probability. The calculation formula is as follows: In the formula, is the final mutation probability of changing the digital tag of the k-th gene position from 0 to 1 when performing the t-th iteration optimization. The larger it is, the greater the possibility that the characteristic parameter corresponding to the k-th gene position is the optimal characteristic parameter. Therefore, to guide the generation of high-quality offspring, is used to correct , and in the form of to calculate the final mutation probability of changing the digital tag of the k-th gene position from 0 to 1 when performing the t-th iteration optimization. The larger it is, The larger it is, the higher the probability that the digital label of the k-th gene locus will be changed from 0 to 1 during the t-th iterative optimization, achieving the technical effect of guiding the generation of high-quality offspring. is the final mutation probability value for changing the digital label of the k-th gene locus from 1 to 0 during the t-th iterative optimization, based on to calculate Similarly, it will not be elaborated here; It should be noted that the input of the finally constructed energy efficiency index prediction model is the energy efficiency index and the optimal feature parameter combination of consecutive multiple monitoring time periods, and the output is the energy efficiency index of the next monitoring time period. Randomly selecting the training set, validation set, and test set from the sample set for model training based on the determined optimal feature parameter combination is the prior art and will not be elaborated here; Step S6, obtaining the predicted value of the energy efficiency index for the future monitoring time period based on the trained energy efficiency index prediction model, and generating an operation and maintenance management work order based on the predicted value of the energy efficiency index; Among them, the method for obtaining the predicted value of the energy efficiency index for the future monitoring time period is: inputting the energy efficiency index and the optimal feature parameter combination of the current monitoring time period and consecutive multiple monitoring time periods before it into the trained model to obtain the predicted value of the energy efficiency index for the future monitoring time period; It should be noted that generating an operation and maintenance management work order based on the predicted value of the energy efficiency index can adopt the prior art. For example, the power usage efficiency is generally between 1.5 and 2.5. When the power usage efficiency exceeds 2.5, it usually means that the load of the data center is too high or there is a fault in the cooling system. Therefore, when the predicted value of the power usage efficiency in the future monitoring time period exceeds 2.5, an operation and maintenance management work order for checking the load status of the data center and the operation status of the cooling system is automatically generated, so as to achieve the effect of discovering and handling problems in advance and ensuring the safe and efficient operation of the data center.

[0026] Embodiment 2: Please refer to Figure 2 , this embodiment provides a remote intelligent operation and maintenance management system for a data center, which is used to execute the remote intelligent operation and maintenance management method for a data center in the above Embodiment 1, including: A data monitoring module, which is used to obtain the status parameter monitoring data of the data center during the monitoring time period, including energy consumption monitoring data, environmental monitoring data, battery monitoring data, and hot and cold aisle monitoring data; An energy efficiency index calculation module, which calculates the energy efficiency index of the data center during the monitoring time period based on the energy consumption monitoring data, including power usage efficiency, partial power usage efficiency, and water usage efficiency; An initial feature library construction module, which is used to extract features from the status parameter monitoring data to construct an initial feature library including multiple feature parameters, and the feature extraction includes descriptive statistics and combined interaction; The final feature library construction module is used to randomly splice feature parameters to obtain multiple combined feature parameters, and evaluate the correlation between the energy efficiency index and each feature parameter and combined feature parameter based on the Pearson correlation coefficient, and screen out strongly correlated feature parameters and combined feature parameters as strongly correlated feature parameters to construct the final feature library; The prediction model construction module randomly generates an initial population based on the strongly correlated feature parameters in the final feature library. Each individual in the initial population corresponds to a combination of feature parameters. An energy efficiency index prediction model is constructed and trained based on the long short-term memory network. The input of this model is the energy efficiency index and feature parameter combinations of multiple consecutive monitoring time periods, and the output is the energy efficiency index of the next monitoring time period. Based on the model training feedback results, the genetic algorithm is used to optimize the feature parameter combinations to obtain the optimal feature parameter combinations for application on the model to complete the model training; The operation and maintenance work order generation module obtains the predicted value of the energy efficiency index for the future monitoring time period based on the trained energy efficiency index prediction model, and generates an operation and maintenance management work order based on the predicted value of the energy efficiency index.

[0027] The above formulas are all dimensionless and take their numerical values for calculation. The formula is a formula obtained by collecting a large amount of data for software simulation to get the closest to the real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0028] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by the combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0029] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0030] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application.

Claims

1. A remote intelligent operation and maintenance management method for a data center, characterized in that, Including: S1. Obtain the status parameter monitoring data of the data center during the monitoring period, including energy consumption monitoring data, environmental monitoring data, battery monitoring data, and cold and hot aisle monitoring data; S2. Based on the energy consumption monitoring data, calculate the energy efficiency indicators of the data center during the monitoring period, including power usage effectiveness (PUE), partial power usage effectiveness (PPUE), and water usage effectiveness (WUE); S3. Extract features from the status parameter monitoring data to construct an initial feature library including multiple feature parameters. The feature extraction includes descriptive statistics and combinatorial interaction; S4. Randomly splice the feature parameters to obtain multiple combined feature parameters, and evaluate the correlation between the energy efficiency indicators and each feature parameter and combined feature parameter based on the Pearson correlation coefficient. Select the strongly correlated feature parameters and combined feature parameters as strongly correlated feature parameters to construct the final feature library; S5. Randomly generate an initial population based on the strongly correlated feature parameters in the final feature library. Each individual in the initial population corresponds to a feature parameter combination. Construct and train an energy efficiency indicator prediction model based on the long short-term memory network. The input of the model is the energy efficiency indicators and feature parameter combinations of consecutive multiple monitoring periods, and the output is the energy efficiency indicator of the next monitoring period. Based on the model training feedback results, use the genetic algorithm to optimize the feature parameter combination to obtain the optimal feature parameter combination for application on the model to complete the model training; S6. Obtain the predicted value of the energy efficiency indicator for the future monitoring period based on the trained energy efficiency indicator prediction model, and generate an operation and maintenance management work order based on the predicted value of the energy efficiency indicator.

2. The remote intelligent operation and maintenance management method for a data center according to claim 1, wherein: The energy consumption monitoring data includes the water consumption of the data center, the power consumption of each functional area of the data center, and the power consumption of IT equipment in each functional area. The functional areas of the data center include the server room, the cooling system area, the power distribution room, the storage area, the network equipment area, and the management and operation and maintenance area. The IT equipment includes servers, storage devices, network equipment, and other IT-related equipment; The environmental monitoring data is the time series data of the external environmental temperature, the time series data of the temperature, humidity, air particulate matter concentration, and air velocity in each functional area during the monitoring period; The battery monitoring data is the time series data of the working temperature, working voltage, charging current, and discharging current of each battery cell during the monitoring period, as well as the cumulative charge and discharge cycle times and the maximum available capacity of each battery cell at the beginning and end of the monitoring period; The cold and hot aisle monitoring data includes the time series data of the temperature at the cold and hot aisle exits, the time series data of the coolant flow rate and temperature at the inlet and outlet of the coolant channel.

3. The remote intelligent operation and maintenance management method for a data center according to claim 2, characterized in that: The calculation formula of the energy efficiency indicator is as follows: Wherein, is the power consumption of the data center during the monitoring period, is the power consumption of the IT equipment in the data center during the monitoring period, is the power usage effectiveness of the data center during the monitoring period; Wherein, is the power consumption of the th functional area in the data center during the monitoring period, is the power consumption of the IT equipment in the th functional area in the data center during the monitoring period, is the partial power usage effectiveness of the th functional area in the data center during the monitoring period, is the index of the functional area in the data center, and , is the number of functional areas in the data center; In the formula, is the water consumption of the data center during the monitoring period, is the water use efficiency of the data center during the monitoring period.

4. The remote intelligent operation and maintenance management method for a data center according to claim 2, wherein: The steps for constructing the initial feature library are as follows: Step S31. Extract the energy consumption monitoring data in the status parameter monitoring data as the first-level energy consumption feature group, and perform data processing on the energy consumption monitoring data to obtain the second-level energy consumption feature group for characterizing the energy consumption distribution of each functional area in the data center; Step S32, perform descriptive statistics on the environmental monitoring data to obtain an environmental statistical feature group, and combine the interactive temperature, humidity, and particulate matter concentration to obtain an environmental interactive feature group; Step S33, perform descriptive statistics on the battery monitoring data to obtain a battery parameter statistical feature group, and combine the interactive current, voltage, charge and discharge cycle times, and capacity to obtain a battery parameter interactive feature group; Step S34, perform descriptive statistics on the hot and cold channel monitoring data to obtain a hot and cold channel statistical feature group, and combine the temperature and coolant flow rate to obtain a hot and cold channel interactive feature group; Step S35, count the characteristic parameters in the primary energy consumption feature group, secondary energy consumption feature group, environmental statistical feature group, environmental interactive feature group, battery parameter statistical feature group, battery parameter interactive feature group, hot and cold channel statistical feature group, and hot and cold channel interactive feature group to form an initial feature library.

5. The remote intelligent operation and maintenance management method for a data center according to claim 1, characterized in that: The method for randomly splicing the characteristic parameters is as follows: randomly select no less than two characteristic parameters from the initial feature library for random calculation, and the random calculation includes but is not limited to calculating the selected characteristic parameters using addition, subtraction, multiplication, and division.

6. The remote intelligent operation and maintenance management method for a data center according to claim 1, characterized in that: The judgment criterion for strong correlation is as follows: if the absolute value of the Pearson correlation coefficient between a characteristic parameter and any energy efficiency index is greater than the correlation threshold, it is determined that the definition of this characteristic parameter is strongly correlated with the energy efficiency index, and this characteristic parameter is a strongly correlated characteristic parameter. The method for judging whether a combined characteristic parameter is a strongly correlated characteristic parameter is the same.

7. The remote intelligent operation and maintenance management method for a data center according to claim 1, wherein: Each individual of the initial population consists of K gene positions, and each gene position corresponds to a strongly correlated feature parameter. If the digital label of the k-th gene position in an individual is 1, it indicates that the feature parameter combination corresponding to this individual includes the strongly correlated feature parameter corresponding to the k-th gene position. Conversely, if the digital label of the k-th gene position in an individual is 0, it indicates that the feature parameter combination corresponding to this individual does not include the strongly correlated feature parameter corresponding to the k-th gene position. k is the index of the gene position, and , where K represents the number of strongly correlated feature parameters in the final feature library.

8. The remote intelligent operation and maintenance management method for a data center according to claim 7, characterized in that: The method for obtaining the optimal characteristic parameter combination is as follows: Step S51, initialize the mutation probability of each gene position in the individual, and perform data preprocessing on the energy efficiency index and strongly correlated characteristic parameters in each historical monitoring time period to form a sample set. The data preprocessing includes removing outliers, filling missing values, and data normalization processing; Step S52, construct an energy efficiency index prediction model based on the long short-term memory network, and set the batch size, feedback training period, and total training period. The feedback training period is set between one-tenth and one-fifth of the total training period. The energy efficiency index prediction model includes an input layer, one or more LSTM layers, a fully connected layer, and an output layer; Step S53, based on the characteristic parameter combination corresponding to each individual, randomly select a training set, a validation set, and a test set from the sample set, and input the training set corresponding to each individual into different energy efficiency index prediction models for training one by one. When the predetermined feedback training period is reached, use the validation set corresponding to each individual to verify each energy efficiency index prediction model to obtain a model training feedback result. The model training feedback result includes the mean square error, mean absolute error, and determination coefficient; Step S54, calculate the fitness value of each individual based on the model training feedback result. The larger the fitness value, the better the characteristic parameter combination corresponding to the individual. The calculation formula is as follows: In the formula, , , are the mean square error, mean absolute error, and coefficient of determination respectively, , , are all preset weights, and , , are determined by the analytic hierarchy process, is the fitness value of the individual; Step S55: Sort each individual in descending order of fitness value, select the individuals in the top 50% as the parent generation, perform crossover and mutation operations pairwise on the parent generation to generate offspring, and determine the mutation probability based on the fitness value of the parent generation and the occurrence frequency of the digital tags of each gene locus. Calculate the fitness value of the offspring based on the same method, and repeatedly perform selection, crossover, and mutation operations on the new population formed by the offspring and the parent generation to generate new parent generations and offspring until the predetermined number of iterative optimizations is reached. Select the individual with the largest fitness value as the optimal individual, and the combination of characteristic parameters corresponding to the optimal individual is the optimal combination of characteristic parameters.

9. The remote intelligent operation and maintenance management method for a data center according to claim 8, characterized in that: The method for determining the mutation probability is as follows: Step S551: Calculate the occurrence frequency of the digital tags of each gene locus in the parent generation. The calculation formula is as follows Wherein, represents the occurrence times of the digital label 1 at the k-th gene position in all parents during the t-th iterative optimization, represents the total number of all parents during the t-th iterative optimization, represents the occurrence frequency of the digital label 1 at the k-th gene position during the t-th iterative optimization, represents the occurrence times of the digital label 0 at the k-th gene position in all parents during the t-th iterative optimization, represents the occurrence frequency of the digital label 0 at the k-th gene position during the t-th iterative optimization, t is the index of the iterative optimization times, and , is the total number of iterative optimizations; Step S552: Based on the fitness value of the parent generation, calculate the initial mutation probability of each gene locus in each round of iterative optimization. The calculation formula is as follows: wherein, is the initial mutation probability of the k-th gene position during the t-th iteration optimization, is the average fitness of all parents during the t-th iteration optimization, is the maximum fitness value of all parents during the process from the 1st iteration optimization to the t-th iteration optimization, is the maximum fitness value of all parents during the t-th iteration optimization, is a piecewise function, is the fitness evaluation threshold, is the introduced random perturbation amount during the t-th iteration optimization; Step S553: Based on the initial mutation probability and the occurrence frequency of the digital tags of each gene locus in the parent generation, generate the final mutation probability value. The calculation formula is as follows: In the formula, is the final mutation probability of changing the digital label of the k-th gene position from 0 to 1 during the t-th iterative optimization. is the final mutation probability of changing the digital label of the k-th gene position from 1 to 0 during the t-th iterative optimization.

10. A remote intelligent operation and maintenance management system for a data center, which is used to execute the remote intelligent operation and maintenance management method for a data center according to any one of the above claims 1-9, characterized in that , including: A data monitoring module for obtaining the status parameter monitoring data of the data center during the monitoring time period, which includes energy consumption monitoring data, environmental monitoring data, battery monitoring data, and hot and cold aisle monitoring data; An energy efficiency index calculation module for calculating the energy efficiency indexes of the data center during the monitoring time period based on the energy consumption monitoring data, including power usage effectiveness, partial power usage effectiveness, and water usage effectiveness; An initial feature library construction module for extracting features from the status parameter monitoring data to construct an initial feature library including multiple characteristic parameters. The feature extraction includes descriptive statistics and combined interaction; A final feature library construction module for randomly splicing the characteristic parameters to obtain multiple combined characteristic parameters, and evaluating the correlation between the energy efficiency indexes and each characteristic parameter and combined characteristic parameter based on the Pearson correlation coefficient, and screening out the strongly correlated characteristic parameters and combined characteristic parameters as the strongly correlated characteristic parameters to construct the final feature library; A prediction model construction module for randomly generating an initial population based on the strongly correlated characteristic parameters in the final feature library. Each individual in the initial population corresponds to a combination of characteristic parameters. Construct and train an energy efficiency index prediction model based on the long short-term memory network. The input of this model is the energy efficiency indexes and the combination of characteristic parameters in multiple consecutive monitoring time periods, and the output is the energy efficiency index in the next monitoring time period. And based on the model training feedback results, use the genetic algorithm to optimize the combination of characteristic parameters to obtain the optimal combination of characteristic parameters for application on the model to complete the model training; An operation and maintenance work order generation module for obtaining the predicted value of the energy efficiency index in the future monitoring time period based on the trained energy efficiency index prediction model, and generating an operation and maintenance management work order based on the predicted value of the energy efficiency index.

Images