Energy efficiency analysis system and method based on difference drive

Abstract

The present invention relates to the technical field of energy efficiency analysis, and invents a difference-driven energy efficiency analysis system and method, including obtaining historical energy consumption data of equipment in a computer room, extracting energy consumption time series features from the historical energy consumption data; obtaining historical equipment monitoring data corresponding to the historical energy consumption data, respectively extracting constraint weight features and structural weight features from the historical equipment monitoring data; performing feature weighting and full-connection mapping on energy consumption time series features according to constraint weight features to obtain energy consumption analysis data; obtaining real-time energy consumption data of equipment in a computer room, calculating energy consumption difference features between energy consumption analysis data and real-time energy consumption data; performing weighted mapping on energy consumption difference features according to structural weight features to obtain weighted energy consumption difference features, performing energy efficiency analysis on equipment in a computer room according to weighted energy consumption difference features, and obtaining energy efficiency analysis results. The present invention can improve the accuracy of energy efficiency analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of energy efficiency analysis, and in particular to an energy efficiency analysis system and method based on difference drive. Background Art

[0002] A data center room is a dedicated facility for centralized storage, management, and distribution of data. It contains a large number of computing and storage devices, such as servers, storage devices, network devices, etc., to support the operation of various information technologies and business applications. A data center room consumes a large amount of energy when in use. In order to reduce energy consumption and operating costs, it is necessary to conduct energy efficiency analysis on the data center room.

[0003] Most of the existing energy efficiency analysis methods are based on simple energy consumption comparison, that is, simple energy consumption result calculation is performed according to the energy consumption value per unit time, the real-time calculated energy consumption result is compared with the historical calculated energy consumption result, and the energy efficiency of the equipment with a large difference is optimized. However, in actual application, the energy efficiency analysis method based on simple energy consumption comparison can only mechanically reflect the average energy consumption value in the current calculated time interval, and cannot reflect more energy consumption information, and does not consider the correlation between energy consumption and usage cycle, which leads to low accuracy in energy efficiency analysis. Summary of the invention

[0004] The present invention provides an energy efficiency analysis system and method based on difference driving, the main purpose of which is to solve the problem of low accuracy when performing energy efficiency analysis.

[0005] To achieve the above-mentioned purpose, the present invention provides an energy efficiency analysis system based on difference drive, the system comprising a web page acquisition module, a visual acquisition module, a structure analysis module, a semantic analysis module and a merge storage module, wherein:

[0006] The time series analysis module is used to obtain historical energy consumption data of equipment in the computer room and extract energy consumption time series characteristics from the historical energy consumption data;

[0007] The weight analysis module is used to obtain the historical equipment monitoring data corresponding to the historical energy consumption data, and extract the constraint weight characteristics and the structure weight characteristics from the historical equipment monitoring data according to the historical energy consumption data. When the weight analysis module extracts the constraint weight characteristics and the structure weight characteristics from the historical equipment monitoring data according to the historical energy consumption data, it includes: extracting the equipment operation monitoring data, the equipment environment monitoring data and the equipment structure monitoring data from the historical equipment monitoring data respectively; extracting the equipment operation temperature and the equipment utilization rate from the equipment operation monitoring data respectively; extracting the ambient temperature from the equipment environment monitoring data; and calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate and the ambient temperature using the following physical constraint formula:

[0008] P=P1+γ1U+γ2T1

[0009] T1=T2+γ3P

[0010] Wherein, P refers to the equipment energy consumption in the historical energy consumption data, P1 is the basic energy consumption in the historical energy consumption data, γ1, γ2 and γ3 are the constraint weight coefficients, U refers to the equipment utilization rate, T1 refers to the equipment operating temperature, and T2 refers to the ambient temperature; the constraint weight coefficient vector is converted into a constraint weight feature; the structural weight feature of the historical energy consumption data is extracted from the equipment structure monitoring data;

[0011] The energy consumption analysis module is used to perform feature weighting and full connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data;

[0012] The difference driving module is used to obtain the real-time energy consumption data of the equipment in the computer room and calculate the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data;

[0013] The energy efficiency analysis module is used to perform weighted mapping on the energy consumption difference characteristics according to the structural weight characteristics to obtain weighted energy consumption difference characteristics, and perform energy efficiency analysis on the equipment in the computer room according to the weighted energy consumption difference characteristics to obtain energy efficiency analysis results.

[0014] Optionally, when the time series analysis module extracts energy consumption time series features from the historical energy consumption data, the time series analysis module includes:

[0015] Sorting the historical energy consumption data into a historical energy consumption sequence in chronological order;

[0016] respectively extracting the long-term time series characteristics of energy consumption and the short-term time series characteristics of energy consumption from the historical energy consumption sequence;

[0017] The energy consumption long-term time series characteristics and the energy consumption short-term time series characteristics are integrated into energy consumption long-short time series characteristics;

[0018] Extracting energy consumption attention time series features from the historical energy consumption sequence;

[0019] Fully connect and fuse the energy consumption length time series feature and the energy consumption attention time series feature into a fully connected time series feature;

[0020] Performing feature dimension reduction mapping on the fully connected time series features to obtain energy consumption time series features.

[0021] Optionally, when the time series analysis module extracts the energy consumption attention time series feature from the historical energy consumption sequence, it includes:

[0022] Performing matrix operation on the historical energy consumption sequence to obtain an energy consumption characteristic matrix;

[0023] Performing attention feature mapping on the energy consumption feature matrix to obtain an energy consumption query feature matrix, an energy consumption key feature matrix, and an energy consumption value feature matrix;

[0024] The energy consumption attention weight is calculated according to the energy consumption query feature matrix and the energy consumption key feature matrix using the following energy consumption attention weight formula:

[0025]

[0026] Among them, W ij It refers to the energy consumption attention weight corresponding to the energy consumption query feature in the i-th row in the energy consumption query feature matrix and the energy consumption key feature in the j-th row in the energy consumption key feature matrix, i refers to the i-th row, j refers to the j-th row, exp is the exponential function symbol, Q i refers to the energy consumption query feature of the i-th row in the energy consumption query feature matrix, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, T is the transposition symbol, col() is the column number symbol, row() is the row number symbol, K refers to the energy consumption key feature matrix, col(K) refers to the total number of columns of the energy consumption key feature matrix, row(K) refers to the total number of rows of the energy consumption key feature matrix, r refers to the rth row, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, K r Refers to the energy consumption bond feature of the rth row in the energy consumption bond feature matrix;

[0027] The energy consumption attention weight is used to perform weighted mapping on the energy consumption value feature matrix to obtain the energy consumption attention time series feature.

[0028] Optionally, when the timing analysis module performs feature dimension reduction mapping on the fully connected timing features to obtain the energy consumption timing features, the timing analysis module includes:

[0029] Centralizing the fully connected time series features to obtain centralized time series features;

[0030] Performing covariance calculation on the centralized time series features to obtain covariance time series features;

[0031] Performing eigenvalue decomposition on the covariance time series feature to obtain a time series eigenvalue and a time series eigenvector;

[0032] Filtering out a reduced-dimensional time series feature from the time series feature vector according to the time series feature value;

[0033] The dimension reduction time series feature is used to perform dimension reduction projection on the fully connected time series feature to obtain an energy consumption time series feature.

[0034] Optionally, when the weight analysis module extracts the structural weight features of the historical energy consumption data from the equipment structure monitoring data, it includes:

[0035] Extracting a computer room equipment set and a device interaction data set corresponding to the computer room equipment set from the equipment structure monitoring data respectively;

[0036] Sorting the historical energy consumption data into a historical energy consumption sequence in chronological order, and sorting the device interaction data set into an interaction data sequence in chronological order;

[0037] Initialize a device node set according to the equipment set in the computer room;

[0038] Using the historical energy consumption sequence to perform node feature mapping on the device node set, to obtain a node energy consumption feature sequence;

[0039] Performing edge feature mapping on the device node set using the interaction data sequence to obtain an edge interaction feature sequence;

[0040] The node energy consumption feature sequence and the edge interaction feature sequence are used to perform multi-layer graph convolution on the device node set to obtain structural weight features.

[0041] Optionally, when the energy consumption analysis module performs feature weighting and full-connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data, it includes:

[0042] Using the constraint weight feature to perform weighted mapping on the energy consumption time series feature to obtain a weighted energy consumption time series feature;

[0043] Performing full connection mapping on the weighted energy consumption time series features to obtain standard energy consumption time series features;

[0044] The standard energy consumption time series characteristics are linearly activated to obtain energy consumption analysis data.

[0045] Optionally, when the difference driving module calculates the energy consumption difference characteristics between the energy consumption analysis data and the energy consumption real-time data, it is specifically used to:

[0046] Converting the energy consumption analysis data vector into an energy consumption analysis feature;

[0047] Converting the energy consumption real-time data vector into energy consumption real-time features;

[0048] The energy consumption difference characteristic between the energy consumption analysis characteristic and the energy consumption real-time characteristic is calculated using the following difference driving formula:

[0049]

[0050] Among them, L n It refers to the nth characteristic element in the energy consumption difference feature, n is the characteristic element index, Max() is the maximum value symbol, Min() is the minimum value symbol, P n refers to the nth characteristic element in the real-time characteristic of energy consumption, It refers to the nth characteristic element in the energy consumption analysis feature, and ‖‖ is the absolute value symbol.

[0051] Optionally, when the energy efficiency analysis module performs weighted mapping on the energy consumption difference feature according to the structural weight feature to obtain a weighted energy consumption difference feature, the energy efficiency analysis module includes:

[0052] Selecting characteristic elements in the structural weight feature one by one as target weight features, and taking the device node corresponding to the target weight feature as the target device node;

[0053] Taking the characteristic element corresponding to the target device node in the energy consumption difference characteristic as the target device difference characteristic;

[0054] The product of the target weight feature and the target device difference feature is used as the target weighted difference feature;

[0055] A weighted energy consumption difference feature is generated based on all target weighted difference features.

[0056] Optionally, when the energy efficiency analysis module performs energy efficiency analysis on the equipment in the computer room according to the weighted energy consumption difference feature to obtain the energy efficiency analysis result, the energy efficiency analysis module includes:

[0057] Selecting characteristic elements in the weighted energy consumption difference feature one by one as the target weighted difference feature, and taking the device node corresponding to the target weighted difference feature as the target device node;

[0058] Determining whether the target weighted difference feature is greater than a preset difference threshold;

[0059] If not, returning to the step of selecting characteristic elements in the weighted energy consumption difference feature one by one as the target weighted difference feature;

[0060] If so, the target device node is abnormally energy-efficiently labeled according to the target weighted difference feature to obtain the target device analysis result, until the target weighted difference feature is the last feature element in the weighted energy consumption difference feature, and then an energy efficiency analysis result is generated according to all target device analysis results.

[0061] In order to solve the above problems, the present invention further provides an energy efficiency analysis method based on difference driving, the method comprising:

[0062] Obtaining historical energy consumption data of equipment in the computer room, and extracting energy consumption time series characteristics from the historical energy consumption data;

[0063] Obtain historical equipment monitoring data corresponding to the historical energy consumption data, and extract constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data, wherein the extracting constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data includes: extracting equipment operation monitoring data, equipment environment monitoring data, and equipment structure monitoring data from the historical equipment monitoring data; extracting equipment operation temperature and equipment utilization rate from the equipment operation monitoring data; extracting ambient temperature from the equipment environment monitoring data; and calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate, and the ambient temperature using the following physical constraint formula:

[0064] P=P1+γ1U+γ2T1

[0065] T1=T2+γ3P

[0066] Wherein, P refers to the equipment energy consumption in the historical energy consumption data, P1 is the basic energy consumption in the historical energy consumption data, γ1, γ2 and γ3 are the constraint weight coefficients, U refers to the equipment utilization rate, T1 refers to the equipment operating temperature, and T2 refers to the ambient temperature; the constraint weight coefficient vector is converted into a constraint weight feature; the structural weight feature of the historical energy consumption data is extracted from the equipment structure monitoring data;

[0067] Perform feature weighting and full-connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data;

[0068] Acquire the real-time energy consumption data of the equipment in the computer room, and calculate the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data;

[0069] The energy consumption difference feature is weighted mapped according to the structural weight feature to obtain a weighted energy consumption difference feature, and energy efficiency analysis is performed on the equipment in the computer room according to the weighted energy consumption difference feature to obtain an energy efficiency analysis result.

[0070] The embodiment of the present invention can capture and utilize the time series variation characteristics of historical energy consumption data in different time periods by extracting energy consumption time series characteristics, and highlight the impact of key time steps on energy consumption in combination with the self-attention mechanism, thereby improving the accuracy of energy consumption analysis. By extracting constraint weight characteristics and structural weight characteristics from the historical equipment monitoring data according to the historical energy consumption data, the physical model and the graph neural network can be combined to set weights for the historical energy consumption data, thereby improving the intuitiveness of energy efficiency analysis. By performing feature weighting and fully connected mapping, the energy consumption time series characteristics can be weighted and assigned in combination with the physical model, thereby improving the accuracy of energy consumption analysis.

[0071] By calculating the energy consumption difference characteristics, it is possible to extract equipment with abnormal energy consumption through the difference between real-time energy consumption data and energy consumption analysis data, thereby improving the efficiency of energy efficiency analysis. By calculating the weighted energy consumption difference characteristics, it is possible to combine the communication frequency and connection relationship between the equipment to determine the weighted processing of the energy consumption difference characteristics, thereby improving the intuitiveness and accuracy of energy efficiency analysis. Through energy efficiency analysis, the energy efficiency of each equipment in the data center computer room can be determined, thereby improving the accuracy of energy efficiency analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A functional module diagram of an energy efficiency analysis system based on difference drive provided by an embodiment of the present invention;

[0073] Figure 2 A flowchart of extracting energy consumption attention time series characteristics provided by an embodiment of the present invention;

[0074] Figure 3 A flow chart of generating energy consumption timing characteristics provided by an embodiment of the present invention;

[0075] Figure 4 A schematic diagram of a flow chart of an energy efficiency analysis method based on difference drive provided in one embodiment of the present invention;

[0076] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0077] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0078] The embodiment of the present application provides an energy efficiency analysis system based on difference driving. The execution subject of the energy efficiency analysis system based on difference driving includes but is not limited to at least one of the electronic devices such as the server, the terminal, etc. that can be configured to execute the system provided by the embodiment of the present application. In other words, the energy efficiency analysis system based on difference driving can be executed by software or hardware installed on the terminal device or the server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0079] like Figure 1 , which is a functional module diagram of an energy efficiency analysis system based on difference drive provided in one embodiment of the present invention.

[0080] The energy efficiency analysis system 100 based on difference drive of the present invention can be installed in an electronic device. According to the functions to be implemented, the energy efficiency analysis system 100 based on difference drive can include a timing analysis module 101, a weight analysis module 102, an energy consumption analysis module 103, a difference drive module 104 and an energy efficiency analysis module 105. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0081] In this embodiment, the functions of each module / unit are as follows:

[0082] The time series analysis module 101 is used to obtain historical energy consumption data of equipment in the computer room and extract energy consumption time series features from the historical energy consumption data.

[0083] In detail, the computer room equipment refers to various equipment in the computer room of a data center, and the computer room equipment includes servers, storage devices, network equipment, air-conditioning equipment, and power equipment, etc. The historical energy consumption data refers to the equipment energy consumption and basic energy consumption of each equipment in the computer room under normal working conditions in the past time period, wherein the basic energy consumption refers to the power consumption of the equipment under no load or minimum load conditions, and the basic energy consumption usually indicates the basic energy consumption of the equipment, does not change with the load, and is a relatively fixed value.

[0084] Specifically, the historical energy consumption data can be obtained through a smart meter and an energy monitoring system, and the energy consumption time series feature is a feature used to characterize the relationship between the energy consumption of each device in the computer room and the time variation.

[0085] In detail, when the timing analysis module 101 extracts the energy consumption timing characteristics from the historical energy consumption data, it includes: sorting the historical energy consumption data into a historical energy consumption sequence in chronological order; extracting the long-term timing characteristics of energy consumption and the short-term timing characteristics of energy consumption from the historical energy consumption sequence respectively; fusing the long-term timing characteristics of energy consumption and the short-term timing characteristics of energy consumption into the long and short timing characteristics of energy consumption; extracting the energy consumption attention timing characteristics from the historical energy consumption sequence; fully connecting and fusing the long and short timing characteristics of energy consumption and the energy consumption attention timing characteristics into a fully connected timing characteristic; performing feature dimensionality reduction mapping on the fully connected timing characteristics to obtain energy consumption timing characteristics.

[0086] In detail, the cell state in the gate structure of the pre-trained Long Short-Term Memory (LSTM) network can be used to extract the long-term time series features of energy consumption, and the hidden state in the gate structure can be used to extract the short-term time series features of energy consumption.

[0087] Specifically, refer to Figure 2 As shown, when the time series analysis module 101 extracts the energy consumption attention time series features from the historical energy consumption sequence, it includes:

[0088] S21, performing matrix operation on the historical energy consumption sequence to obtain an energy consumption characteristic matrix;

[0089] S22, performing attention feature mapping on the energy consumption feature matrix to obtain an energy consumption query feature matrix, an energy consumption key feature matrix, and an energy consumption value feature matrix;

[0090] S23, calculating the energy consumption attention weight according to the energy consumption query feature matrix and the energy consumption key feature matrix;

[0091] S24. Use the energy consumption attention weight to perform weighted mapping on the energy consumption value feature matrix to obtain energy consumption attention time series characteristics.

[0092] In detail, the matrixing refers to using the energy consumption data of each time step in the historical energy consumption sequence as the row vector of the energy consumption characteristic matrix, and splicing all the row vectors into the energy consumption characteristic matrix.

[0093] Specifically, the attention feature mapping can be performed using a linear transformation method, the energy consumption query feature matrix is ​​the product of the energy consumption feature matrix and the query vector, the energy consumption key feature matrix is ​​the product of the energy consumption feature matrix and the key feature, and the energy consumption value feature matrix refers to the product of the energy consumption feature matrix and the value feature.

[0094] In detail, the energy consumption attention weight is calculated using the following energy consumption attention weight formula:

[0095]

[0096] Among them, W ij It refers to the energy consumption attention weight corresponding to the energy consumption query feature in the i-th row in the energy consumption query feature matrix and the energy consumption key feature in the j-th row in the energy consumption key feature matrix, i refers to the i-th row, j refers to the j-th row, exp is the exponential function symbol, Q i refers to the energy consumption query feature of the i-th row in the energy consumption query feature matrix, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, T is the transposition symbol, col() is the column number symbol, row() is the row number symbol, K refers to the energy consumption key feature matrix, col(K) refers to the total number of columns of the energy consumption key feature matrix, row(K) refers to the total number of rows of the energy consumption key feature matrix, r refers to the rth row, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, K r Refers to the energy consumption key feature of the rth row in the energy consumption key feature matrix.

[0097] Specifically, the energy consumption attention weight formula can directly compare each element in the sequence with all other elements, so it can capture long-distance dependencies, unlike traditional convolutional neural networks and recurrent neural networks that are limited by fixed receptive fields or sequence processing orders. It can capture detailed energy consumption timing characteristics and improve the accuracy of energy consumption analysis.

[0098] For details, refer to Figure 3 As shown, when the timing analysis module 101 performs feature dimension reduction mapping on the fully connected timing features to obtain energy consumption timing features, it includes:

[0099] S31, performing centralization processing on the fully connected time series features to obtain centralized time series features;

[0100] S32, performing covariance calculation on the centralized time series features to obtain covariance time series features;

[0101] S33, performing eigenvalue decomposition on the covariance time series feature to obtain a time series eigenvalue and a time series eigenvector;

[0102] S34, filtering out a reduced-dimensional time series feature from the time series feature vector according to the time series feature value;

[0103] S35. Perform dimension reduction projection on the fully connected time series feature using the dimension reduction time series feature to obtain an energy consumption time series feature.

[0104] Specifically, the centralization processing refers to subtracting the average characteristic element from each characteristic element in the fully connected time series feature, and the eigenvalue decomposition (Eigenvalue Decomposition) is a linear algebra technique used to decompose a matrix into the form of eigenvalues ​​and eigenvectors.

[0105] In detail, the reduced-dimensionality time series feature is a feature composed of several feature vectors with higher corresponding time series feature values ​​in the time series feature vector. The reduced-dimensionality projection refers to multiplying the reduced-dimensionality time series feature by the fully-connected time series feature to obtain the energy consumption time series feature. The feature reduced-dimensionality mapping can reduce the redundant features of the fully-connected time series feature and improve the computational efficiency of the subsequent extraction of feature weights.

[0106] In the embodiment of the present invention, by extracting the time series characteristics of energy consumption, it is possible to capture and utilize the time series change characteristics of historical energy consumption data in different time periods, and combine the self-attention mechanism to highlight the impact of key time steps on energy consumption, thereby improving the accuracy of energy consumption analysis.

[0107] The weight analysis module 102 is used to obtain historical equipment monitoring data corresponding to the historical energy consumption data, and extract constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data.

[0108] In detail, the historical equipment monitoring data refers to the relevant data obtained by monitoring each device in the computer room equipment in the past time period. The historical equipment monitoring data includes equipment operation monitoring data, equipment environment monitoring data and equipment structure monitoring data. The historical equipment monitoring data can be obtained by using the equipment work log or the sensor equipment in the data center computer room.

[0109] Specifically, the equipment operation monitoring data includes data such as equipment utilization, fan speed, and equipment operating temperature when the equipment is running; the equipment environment monitoring data includes data such as ambient temperature and ambient humidity when the equipment is running; and the equipment structure monitoring data includes data such as connection relationships when the equipment is running and interaction behavior data between various devices.

[0110] In an embodiment of the present invention, when the weight analysis module 102 extracts constraint weight features and structural weight features from the historical equipment monitoring data respectively according to the historical energy consumption data, it includes: extracting equipment operation monitoring data, equipment environment monitoring data and equipment structure monitoring data respectively from the historical equipment monitoring data; extracting equipment operation temperature and equipment utilization rate respectively from the equipment operation monitoring data; extracting ambient temperature from the equipment environment monitoring data; calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate and the ambient temperature; converting the constraint weight coefficient vector into constraint weight features; and extracting the structural weight features of the historical energy consumption data from the equipment structure monitoring data.

[0111] In detail, the constraint weight coefficient corresponding to the historical energy consumption data is calculated using the following physical constraint formula:

[0112] P=P1+γ1U+γ2T1

[0113] T1=T2+γ3P

[0114] Among them, P refers to the equipment energy consumption in the historical energy consumption data, P1 is the basic energy consumption in the historical energy consumption data, γ1, γ2 and γ3 are the constraint weight coefficients, U refers to the equipment utilization rate, T1 refers to the equipment operating temperature, and T2 refers to the ambient temperature.

[0115] Specifically, the physical constraint formula restricts the historical energy consumption data through a power consumption mathematical model and a thermodynamic model, so that the constraint weight coefficient can improve the accuracy of energy consumption analysis.

[0116] In detail, when the weight analysis module 102 extracts the structural weight features of the historical energy consumption data from the equipment structure monitoring data, it includes: extracting the computer room equipment set and the equipment interaction data set corresponding to the computer room equipment set from the equipment structure monitoring data respectively; sorting the historical energy consumption data into a historical energy consumption sequence in chronological order, and sorting the equipment interaction data set into an interaction data sequence in chronological order; initializing the equipment node set according to the computer room equipment set; using the historical energy consumption sequence to perform node feature mapping on the equipment node set to obtain a node energy consumption feature sequence; using the interaction data sequence to perform edge feature mapping on the equipment node set to obtain an edge interaction feature sequence; using the node energy consumption feature sequence and the edge interaction feature sequence to perform multi-layer graph convolution on the equipment node set to obtain structural weight features.

[0117] In detail, the computer room equipment set is a collection of each device in the computer room equipment monitored in the equipment structure monitoring data, and each device interaction data in the equipment interaction data set corresponds to the data transmission and data interaction related data of each device within a period of time recorded by the equipment structure monitoring data.

[0118] Specifically, each device node in the device node set corresponds to each device in the computer room equipment set, the node feature mapping refers to mapping each energy consumption data in the historical energy consumption sequence as a node energy consumption feature to the corresponding node of the device node set, and the edge feature mapping refers to mapping each interaction data in the interaction data sequence as an edge interaction feature to the corresponding two nodes of the device node set, wherein the edge feature mapping can be performed using the edge weighted averaging method.

[0119] In detail, performing multi-layer graph convolution on the device node set using the node energy consumption feature sequence and the edge interaction feature sequence refers to performing convolution operations on each node energy consumption feature in the node energy consumption feature sequence and each edge interaction feature in the edge interaction feature using multiple graph convolution layers in a pre-trained graph neural network.

[0120] In an embodiment of the present invention, by extracting constraint weight features and structural weight features from the historical equipment monitoring data respectively according to the historical energy consumption data, the historical energy consumption data can be weighted in combination with the physical model and the graph neural network, thereby improving the intuitiveness of energy efficiency analysis.

[0121] The energy consumption analysis module 103 is used to perform feature weighting and full-connection mapping on the energy consumption time series features according to the constraint weight features to obtain energy consumption analysis data.

[0122] In detail, the energy consumption analysis data is energy consumption data of each device in the data center in a future time period analyzed based on the energy consumption time series characteristics and constraint weight characteristics of the historical energy consumption data.

[0123] In an embodiment of the present invention, when the energy consumption analysis module 103 performs feature weighting and full-connection mapping on the energy consumption timing characteristics according to the constraint weight characteristics to obtain energy consumption analysis data, it includes: using the constraint weight characteristics to perform weighted mapping on the energy consumption timing characteristics to obtain weighted energy consumption timing characteristics; performing full-connection mapping on the weighted energy consumption timing characteristics to obtain standard energy consumption timing characteristics; and performing linear activation on the standard energy consumption timing characteristics to obtain energy consumption analysis data.

[0124] Specifically, the weighted mapping can be performed using a bit-by-bit multiplication method, the fully connected layer of a pre-trained long short-term memory network can be used for fully connected mapping, and linear activation functions such as RULE or sigmoid can be used for linear activation.

[0125] In the embodiment of the present invention, by performing feature weighting and full-connection mapping, the energy consumption time series features can be weighted and assigned in combination with the physical model, thereby improving the accuracy of energy consumption analysis.

[0126] The difference driving module 104 is used to obtain the real-time energy consumption data of the equipment in the computer room and calculate the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data.

[0127] In detail, the real-time energy consumption data refers to the energy consumption data of the equipment in the computer room obtained through real-time monitoring, and the energy consumption difference feature is a feature used to reflect the data gap between the energy consumption analysis data and the real-time energy consumption data.

[0128] In an embodiment of the present invention, when the difference driving module 104 is executing the calculation of the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data, it is specifically used to: convert the energy consumption analysis data vector into an energy consumption analysis characteristic; convert the energy consumption real-time data vector into an energy consumption real-time characteristic; calculate the energy consumption difference characteristics between the energy consumption analysis characteristic and the real-time energy consumption characteristic.

[0129] Specifically, the energy consumption difference characteristics are calculated using the following difference driving formula:

[0130]

[0131] Among them, L n It refers to the nth characteristic element in the energy consumption difference feature, n is the characteristic element index, Max() is the maximum value symbol, Min() is the minimum value symbol, P nrefers to the nth characteristic element in the energy consumption real-time characteristic, It refers to the nth characteristic element in the energy consumption analysis feature, and ‖‖ is the absolute value symbol.

[0132] Specifically, the difference-driven formula can standardize the differences among various devices, so that the energy consumption differences of devices of different units and different orders of magnitude are expanded according to the same standard, thereby facilitating subsequent energy consumption analysis.

[0133] In the embodiment of the present invention, by calculating the energy consumption difference characteristics, it is possible to extract devices with abnormal energy consumption through the difference between real-time energy consumption data and energy consumption analysis data, thereby improving the efficiency of energy efficiency analysis.

[0134] The energy efficiency analysis module 105 is used to perform weighted mapping on the energy consumption difference characteristics according to the structural weight characteristics to obtain weighted energy consumption difference characteristics, and perform energy efficiency analysis on the equipment in the computer room according to the weighted energy consumption difference characteristics to obtain energy efficiency analysis results.

[0135] In an embodiment of the present invention, when the energy efficiency analysis module 105 performs weighted mapping of the energy consumption difference feature according to the structural weight feature to obtain a weighted energy consumption difference feature, it includes: selecting the characteristic elements in the structural weight feature one by one as the target weight feature, and taking the device node corresponding to the target weight feature as the target device node; taking the characteristic element corresponding to the target device node in the energy consumption difference feature as the target device difference feature; taking the product of the target weight feature and the target device difference feature as the target weighted difference feature; and generating a weighted energy consumption difference feature based on all the target weighted difference features.

[0136] In detail, the device node corresponds to a device in the data center computer room. By calculating the weighted energy consumption difference characteristics, the communication frequency and connection relationship between the devices can be combined to determine the weighted processing of the energy consumption difference characteristics, thereby improving the intuitiveness and accuracy of energy efficiency analysis. The energy efficiency analysis result is log data that records the energy efficiency status of each device in the data center computer room.

[0137] Specifically, when the energy efficiency analysis module 105 performs energy efficiency analysis on the computer room equipment according to the weighted energy consumption difference feature to obtain the energy efficiency analysis result, it includes: selecting characteristic elements in the weighted energy consumption difference feature one by one as the target weighted difference feature, and taking the device node corresponding to the target weighted difference feature as the target device node; judging whether the target weighted difference feature is greater than a preset difference threshold; if not, returning to the step of selecting characteristic elements in the weighted energy consumption difference feature one by one as the target weighted difference feature; if so, marking the target device node for abnormal energy efficiency according to the target weighted difference feature to obtain the target device analysis result, until the target weighted difference feature is the last characteristic element in the weighted energy consumption difference feature, and then generating the energy efficiency analysis result according to all the target device analysis results.

[0138] In an embodiment of the present invention, by calculating the weighted energy consumption difference characteristics, the energy consumption difference characteristics can be weighted in combination with the AC frequency and connection relationship between the devices, thereby improving the intuitiveness and accuracy of the energy efficiency analysis. The energy efficiency of each device in the data center computer room can be determined through energy efficiency analysis, thereby improving the accuracy of the energy efficiency analysis.

[0139] The embodiment of the present invention can capture and utilize the time series variation characteristics of historical energy consumption data in different time periods by extracting energy consumption time series characteristics, and highlight the impact of key time steps on energy consumption in combination with the self-attention mechanism, thereby improving the accuracy of energy consumption analysis. By extracting constraint weight characteristics and structural weight characteristics from the historical equipment monitoring data according to the historical energy consumption data, the physical model and the graph neural network can be combined to set weights for the historical energy consumption data, thereby improving the intuitiveness of energy efficiency analysis. By performing feature weighting and fully connected mapping, the energy consumption time series characteristics can be weighted and assigned in combination with the physical model, thereby improving the accuracy of energy consumption analysis.

[0140] By calculating the energy consumption difference characteristics, it is possible to extract equipment with abnormal energy consumption through the difference between real-time energy consumption data and energy consumption analysis data, thereby improving the efficiency of energy efficiency analysis. By calculating the weighted energy consumption difference characteristics, it is possible to combine the communication frequency and connection relationship between the equipment to determine the weighted processing of the energy consumption difference characteristics, thereby improving the intuitiveness and accuracy of energy efficiency analysis. Through energy efficiency analysis, the energy efficiency of each equipment in the data center computer room can be determined, thereby improving the accuracy of energy efficiency analysis.

[0141] Reference Figure 4 FIG. 1 is a flow chart of an energy efficiency analysis method based on difference driving provided in one embodiment of the present invention. In this embodiment, the energy efficiency analysis method based on difference driving includes:

[0142] S1. Obtain historical energy consumption data of equipment in the computer room, and extract energy consumption time series characteristics from the historical energy consumption data;

[0143] S2. Obtain historical equipment monitoring data corresponding to the historical energy consumption data, and extract constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data, wherein the extracting constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data includes: extracting equipment operation monitoring data, equipment environment monitoring data, and equipment structure monitoring data from the historical equipment monitoring data; extracting equipment operation temperature and equipment utilization rate from the equipment operation monitoring data; extracting ambient temperature from the equipment environment monitoring data; and calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate, and the ambient temperature using the following physical constraint formula:

[0144] P=P1+γ1U+γ2T1

[0145] T1=T2+γ3P

[0146] Wherein, P refers to the equipment energy consumption in the historical energy consumption data, P1 is the basic energy consumption in the historical energy consumption data, γ1, γ2 and γ3 are the constraint weight coefficients, U refers to the equipment utilization rate, T1 refers to the equipment operating temperature, and T2 refers to the ambient temperature; the constraint weight coefficient vector is converted into a constraint weight feature; the structural weight feature of the historical energy consumption data is extracted from the equipment structure monitoring data;

[0147] S3, performing feature weighting and full-connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data;

[0148] S4, obtaining the real-time energy consumption data of the equipment in the computer room, and calculating the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data;

[0149] S5. Perform weighted mapping on the energy consumption difference feature according to the structural weight feature to obtain a weighted energy consumption difference feature, and perform energy efficiency analysis on the equipment in the computer room according to the weighted energy consumption difference feature to obtain an energy efficiency analysis result.

[0150] In detail, the energy efficiency analysis method based on difference driving in the embodiment of the present invention is used in the same manner as above. Figure 1 The same technical means as the difference-driven energy efficiency analysis system described in the text can produce the same technical effects, so I will not go into details here.

[0151] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0152] Obtaining historical energy consumption data of equipment in the computer room, and extracting energy consumption time series characteristics from the historical energy consumption data;

[0153] Obtain historical equipment monitoring data corresponding to the historical energy consumption data, and extract constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data, wherein the extracting constraint weight features and structure weight features from the historical equipment monitoring data according to the historical energy consumption data includes: extracting equipment operation monitoring data, equipment environment monitoring data, and equipment structure monitoring data from the historical equipment monitoring data; extracting equipment operation temperature and equipment utilization rate from the equipment operation monitoring data; extracting ambient temperature from the equipment environment monitoring data; and calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate, and the ambient temperature using the following physical constraint formula:

[0154] P=P1+γ1U+γ2T1

[0155] T1=T2+γ3P

[0156] Wherein, P refers to the equipment energy consumption in the historical energy consumption data, P1 is the basic energy consumption in the historical energy consumption data, γ1, γ2 and γ3 are the constraint weight coefficients, U refers to the equipment utilization rate, T1 refers to the equipment operating temperature, and T2 refers to the ambient temperature; the constraint weight coefficient vector is converted into a constraint weight feature; the structural weight feature of the historical energy consumption data is extracted from the equipment structure monitoring data;

[0157] Perform feature weighting and full-connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data;

[0158] Acquire the real-time energy consumption data of the equipment in the computer room, and calculate the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data;

[0159] The energy consumption difference feature is weighted mapped according to the structural weight feature to obtain a weighted energy consumption difference feature, and energy efficiency analysis is performed on the equipment in the computer room according to the weighted energy consumption difference feature to obtain an energy efficiency analysis result.

[0160] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and systems can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0161] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0162] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0163] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0164] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0165] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, system, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0166] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or systems stated in the system embodiment can also be implemented by one unit or system through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any specific order.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An energy efficiency analysis system based on difference drive, characterized in that: The system includes a timing analysis module, a weight analysis module, an energy consumption analysis module, a difference driving module and an energy efficiency analysis module, wherein: The timing analysis module is used to obtain the historical energy consumption data of the equipment in the computer room and extract the energy consumption timing characteristics from the historical energy consumption data. When the timing analysis module extracts the energy consumption timing characteristics from the historical energy consumption data, it includes: sorting the historical energy consumption data into a historical energy consumption sequence in chronological order; extracting the long-term timing characteristics of energy consumption and the short-term timing characteristics of energy consumption from the historical energy consumption sequence respectively; fusing the long-term timing characteristics of energy consumption and the short-term timing characteristics of energy consumption into the long-short timing characteristics of energy consumption; extracting the energy consumption attention timing characteristics from the historical energy consumption sequence; fusing the long-short timing characteristics of energy consumption and the short-short timing characteristics of energy consumption into the long-short timing characteristics of energy consumption; The energy consumption attention time series features are fully connected and fused into fully connected time series features; the fully connected time series features are mapped to feature dimension reduction to obtain energy consumption time series features, wherein the time series analysis module, when executing to extract the energy consumption attention time series features from the historical energy consumption sequence, includes: performing matrix operation on the historical energy consumption sequence to obtain an energy consumption feature matrix; performing attention feature mapping on the energy consumption feature matrix to obtain an energy consumption query feature matrix, an energy consumption key feature matrix and an energy consumption value feature matrix; and using the following energy consumption attention weight formula to calculate the energy consumption attention weight according to the energy consumption query feature matrix and the energy consumption key feature matrix: Among them, W ij It refers to the energy consumption attention weight corresponding to the energy consumption query feature in the i-th row in the energy consumption query feature matrix and the energy consumption key feature in the j-th row in the energy consumption key feature matrix, i refers to the i-th row, j refers to the j-th row, exp is the exponential function symbol, Q i refers to the energy consumption query feature of the i-th row in the energy consumption query feature matrix, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, T is the transposition symbol, col() is the column number symbol, row() is the row number symbol, K refers to the energy consumption key feature matrix, col(K) refers to the total number of columns of the energy consumption key feature matrix, row(K) refers to the total number of rows of the energy consumption key feature matrix, r refers to the rth row, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, K r Refers to the energy consumption key feature of the rth row in the energy consumption key feature matrix; the energy consumption value feature matrix is ​​weighted mapped using the energy consumption attention weight to obtain the energy consumption attention time series feature; The weight analysis module is used to obtain the historical equipment monitoring data corresponding to the historical energy consumption data, and extract the constraint weight features and the structure weight features from the historical equipment monitoring data according to the historical energy consumption data, wherein the weight analysis module, when executing the extraction of the constraint weight features and the structure weight features from the historical equipment monitoring data according to the historical energy consumption data, includes: extracting the equipment operation monitoring data, the equipment environment monitoring data and the equipment structure monitoring data from the historical equipment monitoring data respectively; extracting the equipment operation temperature and the equipment utilization rate from the equipment operation monitoring data respectively; extracting the ambient temperature from the equipment environment monitoring data; calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate and the ambient temperature; converting the constraint weight coefficient vector into the constraint weight feature; extracting the structure weight feature of the historical energy consumption data from the equipment structure monitoring data; The energy consumption analysis module is used to perform feature weighting and full connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data; The difference driving module is used to obtain the real-time energy consumption data of the equipment in the computer room and calculate the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data; The energy efficiency analysis module is used to perform weighted mapping on the energy consumption difference characteristics according to the structural weight characteristics to obtain weighted energy consumption difference characteristics, and perform energy efficiency analysis on the equipment in the computer room according to the weighted energy consumption difference characteristics to obtain energy efficiency analysis results.

2. The difference-driven energy efficiency analysis system according to claim 1, characterized in that: When the timing analysis module performs feature dimension reduction mapping on the fully connected timing features to obtain energy consumption timing features, it includes: Centralizing the fully connected time series features to obtain centralized time series features; Performing covariance calculation on the centralized time series features to obtain covariance time series features; Performing eigenvalue decomposition on the covariance time series feature to obtain a time series eigenvalue and a time series eigenvector; Filtering out a reduced-dimensional time series feature from the time series feature vector according to the time series feature value; The dimension reduction time series feature is used to perform dimension reduction projection on the fully connected time series feature to obtain an energy consumption time series feature.

3. The difference-driven energy efficiency analysis system according to claim 1, characterized in that: When the weight analysis module extracts the structural weight features of the historical energy consumption data from the equipment structure monitoring data, the weight analysis module includes: Extracting a computer room equipment set and a device interaction data set corresponding to the computer room equipment set from the equipment structure monitoring data respectively; Sorting the historical energy consumption data into a historical energy consumption sequence in chronological order, and sorting the device interaction data set into an interaction data sequence in chronological order; Initialize a device node set according to the equipment set in the computer room; Using the historical energy consumption sequence to perform node feature mapping on the device node set, to obtain a node energy consumption feature sequence; Performing edge feature mapping on the device node set using the interaction data sequence to obtain an edge interaction feature sequence; The node energy consumption feature sequence and the edge interaction feature sequence are used to perform multi-layer graph convolution on the device node set to obtain structural weight features.

4. The energy efficiency analysis system based on difference drive according to claim 1, characterized in that: When the energy consumption analysis module performs feature weighting and full connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data, it includes: Using the constraint weight feature to perform weighted mapping on the energy consumption time series feature to obtain a weighted energy consumption time series feature; Performing full connection mapping on the weighted energy consumption time series features to obtain standard energy consumption time series features; The standard energy consumption time series characteristics are linearly activated to obtain energy consumption analysis data.

5. The difference-driven energy efficiency analysis system according to claim 1, characterized in that: When the difference driving module calculates the energy consumption difference characteristics between the energy consumption analysis data and the energy consumption real-time data, it is specifically used to: Converting the energy consumption analysis data vector into an energy consumption analysis feature; Converting the energy consumption real-time data vector into energy consumption real-time features; The energy consumption difference characteristic between the energy consumption analysis characteristic and the energy consumption real-time characteristic is calculated using the following difference driving formula: Among them, L n It refers to the nth characteristic element in the energy consumption difference feature, n is the characteristic element index, Max() is the maximum value symbol, Min() is the minimum value symbol, P n refers to the nth characteristic element in the real-time characteristic of energy consumption, It refers to the nth characteristic element in the energy consumption analysis feature, and ‖‖ is the absolute value symbol.

6. The difference-driven energy efficiency analysis system according to claim 1, characterized in that: When the energy efficiency analysis module performs weighted mapping on the energy consumption difference feature according to the structural weight feature to obtain the weighted energy consumption difference feature, the energy efficiency analysis module includes: Selecting characteristic elements in the structural weight feature one by one as target weight features, and taking the device node corresponding to the target weight feature as the target device node; Taking the characteristic element corresponding to the target device node in the energy consumption difference characteristic as the target device difference characteristic; The product of the target weight feature and the target device difference feature is used as the target weighted difference feature; A weighted energy consumption difference feature is generated based on all target weighted difference features.

7. The difference-driven energy efficiency analysis system according to claim 1, characterized in that: When the energy efficiency analysis module performs energy efficiency analysis on the equipment in the computer room according to the weighted energy consumption difference characteristics to obtain the energy efficiency analysis result, it includes: Selecting characteristic elements in the weighted energy consumption difference feature one by one as the target weighted difference feature, and taking the device node corresponding to the target weighted difference feature as the target device node; Determining whether the target weighted difference feature is greater than a preset difference threshold; If not, returning to the step of selecting characteristic elements in the weighted energy consumption difference feature one by one as the target weighted difference feature; If so, the target device node is abnormally energy-efficiently labeled according to the target weighted difference feature to obtain the target device analysis result, until the target weighted difference feature is the last feature element in the weighted energy consumption difference feature, and then an energy efficiency analysis result is generated according to all target device analysis results.

8. A difference-driven energy efficiency analysis method, characterized in that: The method comprises: Obtain historical energy consumption data of equipment in a computer room, and extract energy consumption time series features from the historical energy consumption data, wherein the energy consumption time series features extracted from the historical energy consumption data include: sorting the historical energy consumption data into a historical energy consumption sequence in chronological order; extracting long-term energy consumption time series features and short-term energy consumption time series features from the historical energy consumption sequence respectively; fusing the long-term energy consumption time series features and the short-term energy consumption time series features into long-short energy consumption time series features; extracting energy consumption attention time series features from the historical energy consumption sequence; fusing the long-short energy consumption time series features and the energy consumption attention time series features into a long-short energy consumption time series feature ... The energy time series features are fully connected and fused into fully connected time series features; the fully connected time series features are mapped to feature dimension reduction to obtain energy consumption time series features, wherein the energy consumption attention time series features are extracted from the historical energy consumption sequence, including: performing matrix operation on the historical energy consumption sequence to obtain an energy consumption feature matrix; performing attention feature mapping on the energy consumption feature matrix to obtain an energy consumption query feature matrix, an energy consumption key feature matrix and an energy consumption value feature matrix; and calculating the energy consumption attention weight according to the energy consumption query feature matrix and the energy consumption key feature matrix using the following energy consumption attention weight formula: Among them, W ij It refers to the energy consumption attention weight corresponding to the energy consumption query feature in the i-th row in the energy consumption query feature matrix and the energy consumption key feature in the j-th row in the energy consumption key feature matrix, i refers to the i-th row, j refers to the j-th row, exp is the exponential function symbol, Q i refers to the energy consumption query feature of the i-th row in the energy consumption query feature matrix, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, T is the transposition symbol, col() is the column number symbol, row() is the row number symbol, K refers to the energy consumption key feature matrix, col(K) refers to the total number of columns of the energy consumption key feature matrix, row(K) refers to the total number of rows of the energy consumption key feature matrix, r refers to the rth row, K j refers to the energy consumption key feature of the jth row in the energy consumption key feature matrix, K r Refers to the energy consumption key feature of the rth row in the energy consumption key feature matrix; the energy consumption value feature matrix is ​​weighted mapped using the energy consumption attention weight to obtain the energy consumption attention time series feature; Obtain historical equipment monitoring data corresponding to the historical energy consumption data, and extract constraint weight features and structural weight features from the historical equipment monitoring data according to the historical energy consumption data, wherein the extracting constraint weight features and structural weight features from the historical equipment monitoring data according to the historical energy consumption data includes: extracting equipment operation monitoring data, equipment environment monitoring data, and equipment structure monitoring data from the historical equipment monitoring data; extracting equipment operation temperature and equipment utilization rate from the equipment operation monitoring data; extracting ambient temperature from the equipment environment monitoring data; calculating the constraint weight coefficient corresponding to the historical energy consumption data according to the equipment operation temperature, the equipment utilization rate, and the ambient temperature; converting the constraint weight coefficient vector into a constraint weight feature; extracting the structural weight feature of the historical energy consumption data from the equipment structure monitoring data; Perform feature weighting and full-connection mapping on the energy consumption time series feature according to the constraint weight feature to obtain energy consumption analysis data; Acquire the real-time energy consumption data of the equipment in the computer room, and calculate the energy consumption difference characteristics between the energy consumption analysis data and the real-time energy consumption data; The energy consumption difference feature is weighted mapped according to the structural weight feature to obtain a weighted energy consumption difference feature, and energy efficiency analysis is performed on the equipment in the computer room according to the weighted energy consumption difference feature to obtain an energy efficiency analysis result.

Images