Multi-dimensional machine room management method and system
Through the multi-dimensional computer room management method, the improved CW-PSO-CNN algorithm is used to process video, environment and power consumption data, which solves the problem that traditional computer room management methods cannot meet the security and intelligence requirements of modern data centers, and realizes intelligent management and The effect of cost reduction.
Patent Information
- Application Number
- CN202410983055.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional computer room management methods cannot meet the security and intelligent management needs of modern data centers. They have problems such as diverse equipment, complex structures and information islands, making it difficult to effectively manage the security and operating environment of the computer room.
Using a multi-dimensional computer room management method, by acquiring and preprocessing video timing data, environmental information timing data and power consumption timing data, the improved CW-PSO-CNN algorithm is used to manage personnel intrusion, operating environment and power consumption abnormalities, and generate intelligent Manage results to automatically identify device abnormalities and potential risks.
It improves the intelligence level of computer room management, reduces operation and maintenance costs, improves equipment usage efficiency, and realizes real-time early warning and abnormal management.
Smart Images

Figure CN118917605A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of computer room management, and in particular relates to a multi-dimensional computer room management method and system. Background Art
[0002] With the rapid development of information technology, data centers and computer rooms are becoming increasingly important as the core infrastructure of enterprise information systems. In the Internet industry, computer rooms generally refer to places where servers are stored. They provide IT services to their users or employees. Large computer rooms may house thousands or even more cabinets, while small computer rooms generally house dozens of cabinets. Computer rooms usually house various small computers and servers. Computer rooms have high requirements for temperature, humidity, and anti-static measures, and non-professionals are generally not allowed to enter. However, computer room management still faces many challenges, such as numerous equipment, complex structure, and information islands. Traditional computer room management methods often only conduct video surveillance and manual inspections, which cannot meet the security and intelligent management needs of modern data centers. Summary of the invention
[0003] The present invention provides a multi-dimensional computer room management method and system to solve the technical problems existing in the prior art.
[0004] On the one hand, the present invention provides a multi-dimensional computer room management method, comprising:
[0005] Acquire multi-dimensional time series data corresponding to the managed computer room, and pre-process the multi-dimensional time series data to obtain multi-dimensional time series data after pre-processing;
[0006] Wherein, the multi-dimensional time series data includes video time series data, environmental information time series data and power consumption time series data;
[0007] Based on the video time series data, personnel intrusion management is performed on the managed computer room to obtain a personnel intrusion management result;
[0008] Based on the environmental information time series data, the operating environment management of the managed computer room is performed to obtain the operating environment management result;
[0009] Based on the power consumption time series data, abnormal power consumption management is performed on the managed computer room to obtain abnormal power consumption management results;
[0010] According to the personnel intrusion management results, the operating environment management results and the abnormal power consumption management results, a multi-dimensional computer room management result is obtained.
[0011] Further, the multi-dimensional time series data corresponding to the managed computer room is obtained, and the multi-dimensional time series data is preprocessed to obtain the multi-dimensional time series data after preprocessing, including:
[0012] Obtain video time series data, environmental information time series data, and power consumption time series data corresponding to the managed computer room, and obtain multi-dimensional time series data corresponding to the managed computer room;
[0013] The environmental information time series data includes environmental temperature time series data, equipment temperature time series data, environmental humidity time series data and environmental smoke concentration time series data; the power consumption time series data includes power consumption time series data, overvoltage record time series data, overcurrent record time series data, undervoltage record time series data and undercurrent record time series data; the data lengths of power consumption, overvoltage record, overcurrent record, undervoltage record and undercurrent record at each moment are the same; when there is no overvoltage, overcurrent, undervoltage or undercurrent at a certain time point, the data fields in the corresponding time series data are filled with zeros;
[0014] Performing denoising processing on the video time series data in the multi-dimensional time series data to obtain the pre-processed video time series data;
[0015] Processing abnormal values and missing values of the environmental information time series data and the power consumption time series data in the multi-dimensional time series data to obtain the pre-processed environmental information time series data and the power consumption time series data;
[0016] The preprocessed multi-dimensional time series data is obtained according to the preprocessed video time series data, the environmental information time series data and the power consumption time series data.
[0017] Furthermore, based on the video time series data, personnel intrusion management is performed on the managed computer room to obtain a personnel intrusion management result, including:
[0018] Pre-training video frame recognition rules, wherein the video frame recognition rules are used to identify whether there is an illegal intruder in the video frame;
[0019] Perform frame extraction on the video time series data to obtain the target video frame;
[0020] The target video frame is identified by using the pre-trained video frame recognition rules to obtain the illegal intruder detection result corresponding to the target video frame;
[0021] The illegal intruder detection result corresponding to the target video frame is attached with a timestamp and used as the intrusion management result.
[0022] Furthermore, the pre-trained video frame recognition rules include:
[0023] Obtaining historical video frames that are pre-stored or input by staff through human-computer interaction and illegal intruder detection results corresponding to the historical video frames;
[0024] The improved CW-PSO-CNN algorithm is used to learn the relationship between historical video frames and the illegal intruder detection results corresponding to the historical video frames, and the video frame recognition rules are obtained.
[0025] Furthermore, the improved CW-PSO-CNN algorithm is used to learn the relationship between the historical video frames and the illegal intruder detection results corresponding to the historical video frames, and the video frame recognition rules are obtained, including:
[0026] Based on the CNN algorithm, the chaotic mapping rule is used to initialize the particle swarm, and the best particle recorder of the swarm is enabled to record and update the best particle of the swarm in real time, and the best individual history recorder is enabled to record and update the best individual history of the particle in real time;
[0027] The historical video frames are used as the input data of the CNN algorithm, and the illegal intruder detection results corresponding to the historical video frames are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the particle swarm;
[0028] According to the fitness of each particle, the search times of each particle are determined, and according to the search times of each particle, a jump search is performed on the particle to obtain the particle search value;
[0029] If the fitness of the particle search value increases, the particle search value is used to replace the original particle, and the original particle is used as the first target particle;
[0030] If the fitness of the particle search value decreases, the original particle is used as the second target particle and the corresponding particle search value is used as the third target particle;
[0031] The number of particles to be updated is obtained by subtracting the first target number of particles from the total number of particles in the particle group;
[0032] Determine the population discreteness corresponding to the first target particle, the second target particle, and the third target particle, and select the fourth target particles having the same number as the number of particles to be updated according to the population discreteness;
[0033] The fourth target particle is used to randomly replace particles whose fitness of the particle search value decreases one by one to obtain an updated particle swarm;
[0034] The historical video frames are used as the input data of the CNN algorithm, and the illegal intruder detection results corresponding to the historical video frames are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the updated particle swarm;
[0035] According to the fitness of each particle in the updated particle swarm, a fast grouping value corresponding to each particle is determined, and the particle swarm is divided into a first particle subgroup and a second particle subgroup according to the fast grouping value corresponding to the particle;
[0036] Pair the particles in the first particle subgroup with the particles in the second particle subgroup one by one to obtain particle pairs;
[0037] Read the best particle of the population from the best particle recorder of the population and read the best historical individual of each particle from the best individual historical recorder;
[0038] The particles of the second particle subgroup are updated according to the best historical individual of the particle and the particles in the first particle subgroup to obtain an updated second particle subgroup;
[0039] The particles in the first particle subgroup are updated according to the best particle in the population and the best historical individual of the particle to obtain an updated first particle subgroup;
[0040] Determine whether the current number of training times is greater than or equal to the maximum number of training times. If so, read the best particle of the population from the best particle recorder of the population, and fuse the best particle of the population with the CNN algorithm to obtain the video frame recognition rule. Otherwise, return to the step of executing the jump search.
[0041] Furthermore, the video time series data is subjected to frame extraction processing to obtain a target video frame, including:
[0042] For any two consecutive video frames, obtain the difference in R value, G value and B value of the same pixel, and determine whether any one of the difference in R value, G value and B value is greater than the set difference threshold. If so, mark the pixel as a target pixel, otherwise mark the pixel as a non-target pixel;
[0043] When the number of target pixel points between any two consecutive video frames is greater than a preset number threshold, it is determined that a dynamic transformation occurs in the next video frame, and the next video frame is subjected to frame extraction processing;
[0044] Traverse all two consecutive video frames in the video time series data, and perform frame extraction to obtain the target video frame.
[0045] Further, based on the environmental information time series data, the operating environment of the managed computer room is managed to obtain an operating environment management result, including:
[0046] Constructing a first regular expression corresponding to the ambient temperature, a second regular expression corresponding to the device temperature, a third regular expression corresponding to the ambient humidity, and a fourth regular expression corresponding to the ambient smoke concentration;
[0047] Performing an over-limit detection on the ambient temperature time series data using a first regular expression corresponding to the ambient temperature to obtain an ambient temperature over-limit detection result;
[0048] Perform over-limit detection on the device temperature time series data using the second regular expression corresponding to the device temperature to obtain the device temperature over-limit detection result;
[0049] Perform over-limit detection on the ambient humidity time series data using the third regular expression corresponding to the ambient humidity to obtain an ambient humidity over-limit detection result;
[0050] Perform over-limit detection on the time series data of ambient smoke concentration by using the fourth regular expression corresponding to the ambient smoke concentration to obtain an over-limit detection result of the ambient smoke concentration;
[0051] The operating environment management results are obtained based on the ambient temperature over-limit detection results, equipment temperature over-limit detection results, ambient humidity over-limit detection results and ambient smoke concentration over-limit detection results.
[0052] Furthermore, based on the power consumption time series data, abnormal power consumption management is performed on the managed computer room to obtain abnormal power consumption management results, including:
[0053] Obtaining historical power consumption time series data stored in advance or input by staff through human-computer interaction and power consumption anomaly detection results corresponding to the historical power consumption time series data;
[0054] The improved CW-PSO-CNN algorithm is used to learn the relationship between the historical power consumption time series data and the power consumption anomaly detection results corresponding to the historical power consumption time series data to obtain the power consumption anomaly detection rules;
[0055] The power consumption time series data is identified through power consumption anomaly detection rules to obtain power consumption anomaly management results.
[0056] Furthermore, the improved CW-PSO-CNN algorithm is used to learn the relationship between the historical power consumption time series data and the power consumption anomaly detection results corresponding to the historical power consumption time series data, and the power consumption anomaly detection rules are obtained, including:
[0057] Based on the CNN algorithm, the chaotic mapping rule is used to initialize the particle swarm, and the best particle recorder of the swarm is enabled to record and update the best particle of the swarm in real time, and the best individual history recorder is enabled to record and update the best individual history of the particle in real time;
[0058] The historical electricity consumption time series data is used as the input data of the CNN algorithm, and the power consumption anomaly detection results corresponding to the historical electricity consumption time series data are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the particle swarm;
[0059] According to the fitness of each particle, the search times of each particle are determined, and according to the search times of each particle, a jump search is performed on the particle to obtain the particle search value;
[0060] If the fitness of the particle search value increases, the particle search value is used to replace the original particle, and the original particle is used as the fourth target particle;
[0061] If the fitness of the particle search value decreases, the original particle is used as the fifth target particle and the corresponding particle search value is used as the sixth target particle;
[0062] The number of particles to be updated is obtained by subtracting the fourth target number of particles from the total number of particles in the particle group;
[0063] Determine the population dispersion corresponding to the fourth target particle, the fifth target particle, and the sixth target particle, and select the seventh target particle having the same number as the number of particles to be updated according to the population dispersion;
[0064] The seventh target particle is used to randomly replace particles whose fitness of the particle search value decreases one by one to obtain an updated particle swarm;
[0065] The historical electricity consumption time series data is used as the input data of the CNN algorithm, and the power consumption anomaly detection results corresponding to the historical electricity consumption time series data are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the updated particle swarm;
[0066] According to the fitness of each particle in the updated particle swarm, a fast grouping value corresponding to each particle is determined, and the particle swarm is divided into a first particle subgroup and a second particle subgroup according to the fast grouping value corresponding to the particle;
[0067] Pair the particles in the first particle subgroup with the particles in the second particle subgroup one by one to obtain particle pairs;
[0068] Read the best particle of the population from the best particle recorder of the population and read the best historical individual of each particle from the best individual historical recorder;
[0069] The particles of the second particle subgroup are updated according to the best historical individual of the particle and the particles in the first particle subgroup to obtain an updated second particle subgroup;
[0070] The particles in the first particle subgroup are updated according to the best particle in the population and the best historical individual of the particle to obtain an updated first particle subgroup;
[0071] Determine whether the current number of training times is greater than or equal to the maximum number of training times. If so, read the best particle of the population from the best particle recorder of the population, and fuse the best particle of the population with the CNN algorithm to obtain the power consumption anomaly detection rule, otherwise return to the step of executing the jump search.
[0072] On the other hand, the present invention provides a multi-dimensional computer room management system, including: a multi-dimensional data acquisition module, a first management module, a second management module, a third management module and a management result acquisition module;
[0073] The multi-dimensional data acquisition module is used to acquire the multi-dimensional time series data corresponding to the managed computer room, and pre-process the multi-dimensional time series data to obtain the pre-processed multi-dimensional time series data;
[0074] Wherein, the multi-dimensional time series data includes video time series data, environmental information time series data and power consumption time series data;
[0075] The first management module is used to perform personnel intrusion management on the managed computer room based on the video time series data to obtain a personnel intrusion management result;
[0076] The second management module is used to perform operating environment management on the managed computer room based on the environmental information time series data to obtain an operating environment management result;
[0077] The third management module is used to perform abnormal power consumption management on the managed computer room based on the power consumption time series data to obtain abnormal power consumption management results;
[0078] The management result acquisition module is used to obtain multi-dimensional computer room management results based on the personnel intrusion management results, the operating environment management results and the power consumption abnormality management results.
[0079] The present invention provides a multi-dimensional computer room management method and system, which obtains multi-dimensional time series data corresponding to the managed computer room, and adopts intelligent rules to intelligently identify the multi-dimensional time series data corresponding to the managed computer room, so as to obtain multi-dimensional computer room management results, automatically identify equipment anomalies and potential risks, improve the intelligence level of computer room management, effectively reduce operation and maintenance costs, and improve the utilization efficiency of computer room equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0080] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0081] Figure 1 A flowchart of a multi-dimensional computer room management method provided by an embodiment of the present invention.
[0082] Figure 2 A schematic diagram of the structure of a multi-dimensional computer room management system provided by an embodiment of the present invention.
[0083] Among them, 201 - multi-dimensional data acquisition module, 202 - first management module, 203 - second management module, 204 - third management module, 205 - management result acquisition module.
[0084] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0085] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.
[0086] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0087] like Figure 1 As shown, an embodiment of the present invention provides a multi-dimensional computer room management method, including:
[0088] S101. Acquire multi-dimensional time series data corresponding to a managed computer room, and pre-process the multi-dimensional time series data to obtain pre-processed multi-dimensional time series data.
[0089] The multi-dimensional time series data includes video time series data, environmental information time series data and power consumption time series data.
[0090] It is worth noting that the multi-dimensional time series data in the embodiment of the present invention is only a preferred implementation method, and time series data can also be added or deleted to form different analysis and management effects.
[0091] S102: Based on the video time series data, perform personnel intrusion management on the managed computer room to obtain a personnel intrusion management result.
[0092] For example, deep learning technology can be used to identify video frames in video time series data to determine whether the video frames contain illegal intruders.
[0093] Optionally, when a video frame contains an illegal intruder, an intrusion warning message may be generated and transmitted to a terminal device of the staff to achieve a real-time early warning effect.
[0094] S103: Based on the environmental information time series data, perform operating environment management on the managed computer room to obtain an operating environment management result.
[0095] Environmental information is important data that reflects the operating environment of the computer room, such as temperature, humidity, etc. By analyzing the time series data of environmental information, it can be determined whether the operating environment of the computer room is abnormal, thereby ensuring the safety of the computer room.
[0096] S104. Based on the power consumption time series data, abnormal power consumption management is performed on the managed computer room to obtain abnormal power consumption management results.
[0097] In addition to the operating environment of the computer room, its own operating status is also very important. Therefore, by analyzing its power consumption timing data, it is possible to determine whether there are any operating abnormalities and achieve multi-dimensional management of the computer room.
[0098] S105. Obtain a multi-dimensional computer room management result based on the personnel intrusion management result, the operating environment management result, and the abnormal power consumption management result.
[0099] The present invention provides a multi-dimensional computer room management method and system, which obtains multi-dimensional time series data corresponding to the managed computer room, and adopts intelligent rules to intelligently identify the multi-dimensional time series data corresponding to the managed computer room, so as to obtain multi-dimensional computer room management results, automatically identify equipment anomalies and potential risks, improve the intelligence level of computer room management, effectively reduce operation and maintenance costs, and improve the utilization efficiency of computer room equipment.
[0100] In an embodiment of the present invention, multi-dimensional time series data corresponding to a managed computer room is obtained, and the multi-dimensional time series data is preprocessed to obtain the preprocessed multi-dimensional time series data, including:
[0101] The video time series data, environmental information time series data and power consumption time series data corresponding to the managed computer room are obtained to obtain the multi-dimensional time series data corresponding to the managed computer room.
[0102] Among them, the environmental information timing data includes environmental temperature timing data, equipment temperature timing data, environmental humidity timing data and environmental smoke concentration timing data, and the power consumption timing data includes power consumption timing data, overvoltage record timing data, overcurrent record timing data, undervoltage record timing data and undercurrent record timing data; the data length of power consumption, overvoltage record, overcurrent record, undervoltage record and undercurrent record at each moment is the same. When there is no overvoltage, overcurrent, undervoltage and undercurrent at a certain point in time, the data field in the corresponding timing data will be filled with zeros.
[0103] It is worth noting that, in addition to the above-mentioned specific data, other relevant data in the computer room can also be used as time series data for analysis to achieve more management effects.
[0104] Optionally, in order to make the length of each time series data the same, a maximum data length can be set for the data at each time point, and if the data length is less than the maximum data length, zeros are filled in front of it for processing, which can ensure that the data length is the same and will not affect the data identification process. The embodiment of the present invention mainly performs length uniform processing on the power consumption time series data.
[0105] Performing denoising processing on the video time series data in the multi-dimensional time series data to obtain the pre-processed video time series data;
[0106] The environmental information time series data and the power consumption time series data in the multi-dimensional time series data are processed for abnormal values and missing values to obtain the pre-processed environmental information time series data and the power consumption time series data.
[0107] For example: when the data at a certain moment in the time series data exceeds the limit range of the data (such as electricity consumption is obviously impossible to be a negative number, when the electricity consumption is negative, it can be considered an outlier) or when the data field at a certain moment is blank, the data mean of the previous moment and the next moment can be used for replacement processing to eliminate outliers or missing values.
[0108] The preprocessed multi-dimensional time series data is obtained according to the preprocessed video time series data, the environmental information time series data and the power consumption time series data.
[0109] In an embodiment of the present invention, based on the video time series data, personnel intrusion management is performed on the managed computer room to obtain a personnel intrusion management result, including:
[0110] Pre-training video frame recognition rules, wherein the video frame recognition rules are used to identify whether there is an illegal intruder in the video frame;
[0111] Perform frame extraction on the video time series data to obtain the target video frame;
[0112] The target video frame is identified by using the pre-trained video frame recognition rules to obtain the illegal intruder detection result corresponding to the target video frame;
[0113] Optionally, the illegal intruder detection result corresponding to the target video frame may include the presence of an illegal intruder or the absence of an illegal intruder. The absence of an illegal intruder may be further subdivided into: no person and classification labels of related persons, thereby realizing full-cycle personnel recording.
[0114] The illegal intruder detection result corresponding to the target video frame is attached with a timestamp and used as the intrusion management result.
[0115] In an embodiment of the present invention, the pre-trained video frame recognition rule includes:
[0116] Obtaining historical video frames that are pre-stored or input by staff through human-computer interaction and illegal intruder detection results corresponding to the historical video frames;
[0117] The improved CW (cooperate with)-PSO (Particle Swarm Optimization)-CNN (Convolutional Neural Networks) algorithm is used to learn the relationship between historical video frames and the illegal intruder detection results corresponding to the historical video frames, and the video frame recognition rules are obtained.
[0118] In the embodiment of the present invention, the improved CW-PSO-CNN algorithm is used to learn the relationship between the historical video frames and the illegal intruder detection results corresponding to the historical video frames to obtain the video frame recognition rules, including:
[0119] Based on the CNN algorithm, the chaotic mapping rule is used to initialize the particle swarm, and the best particle recorder of the swarm is enabled to record and update the best particle of the swarm in real time, and the best individual history recorder is enabled to record and update the best individual history of the particle in real time;
[0120] Optionally, the particle swarm is initialized using a chaotic mapping rule, which may include: generating a particle swarm using a chaotic mapping rule based on a hyperparameter upper limit and a hyperparameter lower limit in a CNN algorithm.
[0121] The historical video frames are used as the input data of the CNN algorithm, and the illegal intruder detection results corresponding to the historical video frames are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the particle swarm;
[0122] An embodiment of the present invention provides a fitness acquisition method, which may include: acquiring a root mean square error, a mean square error, or a cross entropy error of a particle, and then adding the error to a very small constant (such as 0.0001) and taking the reciprocal to obtain the fitness of the particle.
[0123] According to the fitness of each particle, the search times of each particle are determined, and according to the search times of each particle, a jump search is performed on the particle to obtain the particle search value;
[0124] According to the fitness of each particle, the search times of each particle are determined, including:
[0125]
[0126] s i =round(S i )
[0127] Among them, f max represents the fitness of the best particle in the particle swarm, f i represents the fitness corresponding to the i-th particle, i = 1, 2, ..., I, I represents the total number of particles in the particle swarm, ε represents a constant term, which is generally set to 0.0001, S i Indicates the intermediate parameter, round(*) indicates the rounding function, s i represents the number of searches corresponding to the i-th particle;
[0128] The number of searches corresponding to the i-th particle is s i Based on, execute i Searches, and in each search process, if the fitness increases after the search, then the search is accepted, otherwise the search is rejected; each search is:
[0129] x i,d '=x i,d +rand(-R i ,R i )
[0130]
[0131] Among them, x i,d represents the d-th dimension parameter of the ith particle, d = 1, 2, ..., D, D represents the total number of dimensions of the particle, x i,d 'Indicates the x after the search i,d , R i Indicates the search radius, rand(0,R i ) means that in (0,R i ) to pick a random number, f min represents the minimum fitness, R max Indicates the preset maximum search radius.
[0132] Optionally, each time the particle's position changes, the particle can be processed as follows:
[0133]
[0134] in, represents the particle before out-of-bounds processing, represents the lower limit of the particle's d-th dimension parameter, The upper limit corresponding to the particle's d-th dimension parameter, Represents the particle after out-of-bounds processing, and % represents the modulo operation.
[0135] Through the above search method, the particles can achieve a stronger global search function, thereby ensuring that the entire algorithm will not fall into the local optimum, thus ensuring the functionality of the entire algorithm.
[0136] If the fitness of the particle search value increases, the particle search value is used to replace the original particle, and the original particle is used as the first target particle;
[0137] If the fitness of the particle search value decreases, the original particle is used as the second target particle and the corresponding particle search value is used as the third target particle;
[0138] The number of particles to be updated is obtained by subtracting the first target number of particles from the total number of particles in the particle group;
[0139] Determine the population discreteness corresponding to the first target particle, the second target particle, and the third target particle, and select the fourth target particles having the same number as the number of particles to be updated according to the population discreteness;
[0140] For example, if the particle search value of H particles increases, the original values of the H particles are first taken out to obtain the first target particle, and then the particle search value is used to update the H particles. Then, the original values of IH particles are used as the second target particle, the particle search value of IH particles is used as the third target particle, and IH particles are taken out from the first target particle, the second target particle and the third target particle as the remaining IH particles.
[0141] The population discreteness corresponding to the first target particle, the second target particle, and the third target particle is determined as:
[0142] Based on the first target particle, the second target particle and the third target particle, for any particle to be calculated whose population discreteness is to be obtained, determine J neighboring particles with the smallest Euclidean distance to the particle to be calculated;
[0143] According to the fitness of the neighborhood particles, the population discreteness of the particles to be calculated is determined as:
[0144]
[0145] Among them, ((f k ) j represents the fitness of the kth particle among the first target particle, the second target particle, and the third target particle, (f k-1 ) j represents the fitness of the k-1th particle among the first target particle, the second target particle, and the third target particle, (f k+1 ) jrepresents the fitness of the first target particle, the second target particle, and the k+1th particle among the third target particles;
[0146] When k=1 or k=H+2*(IH)=2I-H, the particle is directly determined to be the fourth target particle;
[0147] When k is another number, particles with small population discreteness are removed in turn, and when the population discreteness is the same, any one of the particles with the same population discreteness is randomly removed until the number of the fourth target particles meets IH.
[0148] The fourth target particle is used to randomly replace particles whose fitness of the particle search value decreases one by one to obtain an updated particle swarm;
[0149] Through the selection process of the above example, not only can the particles with larger fitness be retained, but also the excessive aggregation of the population can be effectively avoided, the collective jump search can be performed, and the global search can be realized without affecting the execution efficiency of the algorithm in the later stage.
[0150] The historical video frames are used as the input data of the CNN algorithm, and the illegal intruder detection results corresponding to the historical video frames are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the updated particle swarm;
[0151] According to the fitness of each particle in the updated particle swarm, a fast grouping value corresponding to each particle is determined, and the particle swarm is divided into a first particle subgroup and a second particle subgroup according to the fast grouping value corresponding to the particle;
[0152] Determine the fast grouping value corresponding to each particle:
[0153]
[0154] Among them, ξ i represents the fast grouping value corresponding to the i-th particle, f n,i represents the fitness obtained by the i-th particle under the n-th error function, represents the maximum fitness obtained by the nth error function, The minimum fitness obtained by the nth error function.
[0155] Optional error functions such as root mean square error, mean square error, and cross entropy error should generally have values greater than or equal to 3.
[0156] Pair the particles in the first particle subgroup with the particles in the second particle subgroup one by one to obtain particle pairs;
[0157] Read the best particle of the population from the best particle recorder of the population and read the best historical individual of each particle from the best individual historical recorder;
[0158] The particles of the second particle subgroup are updated according to the best historical individual of the particle and the particles in the first particle subgroup to obtain an updated second particle subgroup;
[0159] According to the best historical individual of the particle and the particles in the first particle subgroup, the particles in the second particle subgroup are updated as follows:
[0160]
[0161] in, represents the mth particle of the second particle subgroup in the tth training process, m = I / 2, represents the updated value of the mth particle of the second particle subgroup during the t+1th training process, represents the mth particle of the updated second particle subgroup, ω represents the inertia weight, represents the updated value of the mth particle of the second particle subgroup during the tth training process (if there is none, it is zero), c1 represents the first update coefficient, r1 represents a random number between (0,1), and x mbest Indicated The best historical individual, c2 represents the second update coefficient, r2 represents a random number between (0,1), x ml represents a particle in the first particle subgroup that is paired with the mth particle in the second particle subgroup.
[0162] By guiding the search through the first particle subpopulation with better fitness, particles can search in different and better-positioned areas, which is more helpful to find the best location. When the algorithm is in the later stage, all particles gather together, which does not affect the search ability of the overall algorithm.
[0163] The particles in the first particle subgroup are updated according to the best particle in the population and the best historical individual of the particle to obtain an updated first particle subgroup;
[0164] According to the best particle in the population and the best historical individual of the particle, the particles in the first particle sub-swarm are updated as follows:
[0165]
[0166] in, represents the lth particle of the first particle subgroup in the tth training process, l = I / 2, represents the updated value of the lth particle of the first particle subgroup during the t+1th training process, represents the lth particle of the first particle subgroup after update, represents the updated value of the lth particle of the first particle subgroup during the tth training process, x gbestRepresents the best particle in the population.
[0167] Through the above-mentioned best particles in the population, the particles in the first particle sub-population with better fitness can move toward the optimal position, and the particles with better positions can move further toward the current optimal position, which helps to improve the convergence accuracy and accelerate the convergence speed.
[0168] Determine whether the current number of training times is greater than or equal to the maximum number of training times. If so, read the best particle of the population from the best particle recorder of the population, and fuse the best particle of the population with the CNN algorithm to obtain the video frame recognition rule. Otherwise, return to the step of executing the jump search.
[0169] Fusing the best particle of the population with the CNN algorithm may include: applying the hyperparameters included in the best particle of the population to the CNN algorithm.
[0170] It is worth noting that the above-mentioned data relationship learning algorithm is only a preferred implementation of this embodiment. Conventional algorithms such as particle swarm algorithm, gradient descent method, genetic algorithm, etc. can also be directly used to optimize the CNN algorithm to achieve data relationship learning.
[0171] The improved CW-PSO-CNN algorithm provided in the embodiment of the present invention adopts the idea of collaborative evolution to improve the particle swarm algorithm, which can effectively improve the global search capability of the particle algorithm, thereby achieving better training effect.
[0172] In an embodiment of the present invention, the video time sequence data is subjected to frame extraction processing to obtain a target video frame, including:
[0173] For any two consecutive video frames, obtain the difference in R value, G value and B value of the same pixel, and determine whether any one of the difference in R value, G value and B value is greater than the set difference threshold. If so, mark the pixel as a target pixel, otherwise mark the pixel as a non-target pixel;
[0174] When the number of target pixel points between any two consecutive video frames is greater than a preset number threshold, it is determined that a dynamic transformation occurs in the next video frame, and a frame extraction process is performed on the next video frame;
[0175] Traverse all two consecutive video frames in the video time series data, and perform frame extraction to obtain the target video frame.
[0176] Optionally, in addition to the above frame extraction processing method, a preset interval frame extraction method can also be used for recognition, for example: extracting a video frame every 0.1 seconds to obtain a target video frame.
[0177] In an embodiment of the present invention, based on the environment information time series data, the operation environment management of the managed computer room is performed to obtain the operation environment management result, including:
[0178] Constructing a first regular expression corresponding to the ambient temperature, a second regular expression corresponding to the device temperature, a third regular expression corresponding to the ambient humidity, and a fourth regular expression corresponding to the ambient smoke concentration;
[0179] The regular expressions in this embodiment are essentially conditional judgment expressions. Each regular expression limits the data range of its corresponding data. When the corresponding data range is exceeded, the data is abnormal.
[0180] Optionally, when there is an abnormality in the data, an abnormality warning may be generated and transmitted to the terminal device of the staff so that the staff can promptly discover the abnormality of the operating environment.
[0181] Performing an over-limit detection on the ambient temperature time series data using a first regular expression corresponding to the ambient temperature to obtain an ambient temperature over-limit detection result;
[0182] Perform over-limit detection on the device temperature time series data using the second regular expression corresponding to the device temperature to obtain the device temperature over-limit detection result;
[0183] Perform over-limit detection on the ambient humidity time series data using the third regular expression corresponding to the ambient humidity to obtain an ambient humidity over-limit detection result;
[0184] Perform over-limit detection on the time series data of ambient smoke concentration by using the fourth regular expression corresponding to the ambient smoke concentration to obtain an over-limit detection result of the ambient smoke concentration;
[0185] The operating environment management results are obtained based on the ambient temperature over-limit detection results, equipment temperature over-limit detection results, ambient humidity over-limit detection results and ambient smoke concentration over-limit detection results.
[0186] In an embodiment of the present invention, based on the power consumption time series data, power consumption abnormality management is performed on the managed computer room to obtain a power consumption abnormality management result, including:
[0187] Obtaining historical power consumption time series data stored in advance or input by staff through human-computer interaction and power consumption anomaly detection results corresponding to the historical power consumption time series data;
[0188] The improved CW-PSO-CNN algorithm is used to learn the relationship between the historical power consumption time series data and the power consumption anomaly detection results corresponding to the historical power consumption time series data to obtain the power consumption anomaly detection rules;
[0189] The power consumption time series data is identified through power consumption anomaly detection rules to obtain power consumption anomaly management results.
[0190] The electricity consumption time series data after length unification can form a data matrix, so the CNN algorithm can be used to identify the data matrix.
[0191] In the embodiment of the present invention, the improved CW-PSO-CNN algorithm is used to learn the relationship between the historical power consumption time series data and the power consumption anomaly detection results corresponding to the historical power consumption time series data to obtain the power consumption anomaly detection rules, including:
[0192] Based on the CNN algorithm, the chaotic mapping rule is used to initialize the particle swarm, and the best particle recorder of the swarm is enabled to record and update the best particle of the swarm in real time, and the best individual history recorder is enabled to record and update the best individual history of the particle in real time;
[0193] The historical electricity consumption time series data is used as the input data of the CNN algorithm, and the power consumption anomaly detection results corresponding to the historical electricity consumption time series data are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the particle swarm;
[0194] According to the fitness of each particle, the search times of each particle are determined, and according to the search times of each particle, a jump search is performed on the particle to obtain the particle search value;
[0195] If the fitness of the particle search value increases, the particle search value is used to replace the original particle, and the original particle is used as the fourth target particle;
[0196] If the fitness of the particle search value decreases, the original particle is used as the fifth target particle and the corresponding particle search value is used as the sixth target particle;
[0197] The number of particles to be updated is obtained by subtracting the fourth target number of particles from the total number of particles in the particle group;
[0198] Determine the population dispersion corresponding to the fourth target particle, the fifth target particle, and the sixth target particle, and select the seventh target particle having the same number as the number of particles to be updated according to the population dispersion;
[0199] The seventh target particle is used to randomly replace particles whose fitness of the particle search value decreases one by one to obtain an updated particle swarm;
[0200] The historical electricity consumption time series data is used as the input data of the CNN algorithm, and the power consumption anomaly detection results corresponding to the historical electricity consumption time series data are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the updated particle swarm;
[0201] According to the fitness of each particle in the updated particle swarm, a fast grouping value corresponding to each particle is determined, and the particle swarm is divided into a first particle subgroup and a second particle subgroup according to the fast grouping value corresponding to the particle;
[0202] Pair the particles in the first particle subgroup with the particles in the second particle subgroup one by one to obtain particle pairs;
[0203] Read the best particle of the population from the best particle recorder of the population and read the best historical individual of each particle from the best individual historical recorder;
[0204] The particles of the second particle subgroup are updated according to the best historical individual of the particle and the particles in the first particle subgroup to obtain an updated second particle subgroup;
[0205] The particles in the first particle subgroup are updated according to the best particle in the population and the best historical individual of the particle to obtain an updated first particle subgroup;
[0206] Determine whether the current number of training times is greater than or equal to the maximum number of training times. If so, read the best particle of the population from the best particle recorder of the population, and fuse the best particle of the population with the CNN algorithm to obtain the power consumption anomaly detection rule. Otherwise, return to the step of executing the jump search.
[0207] like Figure 2 As shown, an embodiment of the present invention provides a multi-dimensional computer room management system, including: a multi-dimensional data acquisition module 201, a first management module 202, a second management module 203, a third management module 204 and a management result acquisition module 205;
[0208] The multi-dimensional data acquisition module 201 is used to acquire the multi-dimensional time series data corresponding to the managed computer room, and pre-process the multi-dimensional time series data to obtain the pre-processed multi-dimensional time series data;
[0209] Wherein, the multi-dimensional time series data includes video time series data, environmental information time series data and power consumption time series data;
[0210] The first management module 202 is used to perform personnel intrusion management on the managed computer room based on the video time series data to obtain a personnel intrusion management result;
[0211] The second management module 203 is used to perform operating environment management on the managed computer room based on the environmental information time series data to obtain an operating environment management result;
[0212] The third management module 204 is used to perform abnormal power consumption management on the managed computer room based on the power consumption time series data to obtain abnormal power consumption management results;
[0213] The management result acquisition module 205 is used to acquire multi-dimensional computer room management results according to the personnel intrusion management results, the operating environment management results and the power consumption abnormality management results.
[0214] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The schemes in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.
[0215] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0216] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0217] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0218] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0219] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A multi-dimensional computer room management method, characterized in that: include: Acquire multi-dimensional time series data corresponding to the managed computer room, and pre-process the multi-dimensional time series data to obtain multi-dimensional time series data after pre-processing; Wherein, the multi-dimensional time series data includes video time series data, environmental information time series data and power consumption time series data; Based on the video time series data, personnel intrusion management is performed on the managed computer room to obtain a personnel intrusion management result; Based on the environmental information time series data, the operating environment management of the managed computer room is performed to obtain the operating environment management result; Based on the power consumption time series data, abnormal power consumption management is performed on the managed computer room to obtain abnormal power consumption management results; According to the personnel intrusion management results, the operating environment management results and the abnormal power consumption management results, a multi-dimensional computer room management result is obtained.
2. The multi-dimensional computer room management method according to claim 1, characterized in that: Acquiring multi-dimensional time series data corresponding to the managed computer room, and preprocessing the multi-dimensional time series data to obtain the preprocessed multi-dimensional time series data, including: Obtain video time series data, environmental information time series data, and power consumption time series data corresponding to the managed computer room, and obtain multi-dimensional time series data corresponding to the managed computer room; The environmental information time series data includes environmental temperature time series data, equipment temperature time series data, environmental humidity time series data and environmental smoke concentration time series data; the power consumption time series data includes power consumption time series data, overvoltage record time series data, overcurrent record time series data, undervoltage record time series data and undercurrent record time series data; the data lengths of power consumption, overvoltage record, overcurrent record, undervoltage record and undercurrent record at each moment are the same; when there is no overvoltage, overcurrent, undervoltage or undercurrent at a certain time point, the data fields in the corresponding time series data are filled with zeros; Performing denoising processing on the video time series data in the multi-dimensional time series data to obtain the pre-processed video time series data; Processing abnormal values and missing values of the environmental information time series data and the power consumption time series data in the multi-dimensional time series data to obtain the pre-processed environmental information time series data and the power consumption time series data; The preprocessed multi-dimensional time series data is obtained according to the preprocessed video time series data, the environmental information time series data and the power consumption time series data.
3. The multi-dimensional computer room management method according to claim 2, characterized in that: Based on the video time series data, personnel intrusion management is performed on the managed computer room to obtain a personnel intrusion management result, including: Pre-training video frame recognition rules, wherein the video frame recognition rules are used to identify whether there is an illegal intruder in the video frame; Perform frame extraction on the video time series data to obtain the target video frame; The target video frame is identified by using the pre-trained video frame recognition rules to obtain the illegal intruder detection result corresponding to the target video frame; The illegal intruder detection result corresponding to the target video frame is attached with a timestamp and used as the intrusion management result.
4. The multi-dimensional computer room management method according to claim 3, characterized in that: Pre-trained video frame recognition rules, including: Obtaining historical video frames that are pre-stored or input by staff through human-computer interaction and illegal intruder detection results corresponding to the historical video frames; The improved CW-PSO-CNN algorithm is used to learn the relationship between historical video frames and the illegal intruder detection results corresponding to the historical video frames, and the video frame recognition rules are obtained.
5. The multi-dimensional computer room management method according to claim 4, characterized in that: The improved CW-PSO-CNN algorithm is used to learn the relationship between historical video frames and the illegal intruder detection results corresponding to the historical video frames, and the video frame recognition rules are obtained, including: Based on the CNN algorithm, the chaotic mapping rule is used to initialize the particle swarm, and the best particle recorder of the swarm is enabled to record and update the best particle of the swarm in real time, and the best individual history recorder is enabled to record and update the best individual history of the particle in real time; The historical video frames are used as the input data of the CNN algorithm, and the illegal intruder detection results corresponding to the historical video frames are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the particle swarm; According to the fitness of each particle, the search times of each particle are determined, and according to the search times of each particle, a jump search is performed on the particle to obtain the particle search value; If the fitness of the particle search value increases, the particle search value is used to replace the original particle, and the original particle is used as the first target particle; If the fitness of the particle search value decreases, the original particle is used as the second target particle and the corresponding particle search value is used as the third target particle; The number of particles to be updated is obtained by subtracting the first target number of particles from the total number of particles in the particle group; Determine the population discreteness corresponding to the first target particle, the second target particle, and the third target particle, and select the fourth target particles having the same number as the number of particles to be updated according to the population discreteness; The fourth target particle is used to randomly replace particles whose fitness of the particle search value decreases one by one to obtain an updated particle swarm; The historical video frames are used as the input data of the CNN algorithm, and the illegal intruder detection results corresponding to the historical video frames are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the updated particle swarm; According to the fitness of each particle in the updated particle swarm, a fast grouping value corresponding to each particle is determined, and the particle swarm is divided into a first particle subgroup and a second particle subgroup according to the fast grouping value corresponding to the particle; Pair the particles in the first particle subgroup with the particles in the second particle subgroup one by one to obtain particle pairs; Read the best particle of the population from the best particle recorder of the population and read the best historical individual of each particle from the best individual historical recorder; The particles of the second particle subgroup are updated according to the best historical individual of the particle and the particles in the first particle subgroup to obtain an updated second particle subgroup; The particles in the first particle subgroup are updated according to the best particle in the population and the best historical individual of the particle to obtain an updated first particle subgroup; Determine whether the current number of training times is greater than or equal to the maximum number of training times. If so, read the best particle of the population from the best particle recorder of the population, and fuse the best particle of the population with the CNN algorithm to obtain the video frame recognition rule. Otherwise, return to the step of executing the jump search.
6. The multi-dimensional computer room management method according to claim 4, characterized in that: The video timing data is subjected to frame extraction processing to obtain the target video frame, including: For any two consecutive video frames, obtain the difference in R value, G value and B value of the same pixel, and determine whether any one of the difference in R value, G value and B value is greater than the set difference threshold. If so, mark the pixel as a target pixel, otherwise mark the pixel as a non-target pixel; When the number of target pixel points between any two consecutive video frames is greater than a preset number threshold, it is determined that a dynamic transformation occurs in the next video frame, and the next video frame is subjected to frame extraction processing; Traverse all two consecutive video frames in the video time series data, and perform frame extraction to obtain the target video frame.
7. The multi-dimensional computer room management method according to claim 6, characterized in that: Based on the environmental information time series data, the operating environment of the managed computer room is managed to obtain an operating environment management result, including: Constructing a first regular expression corresponding to the ambient temperature, a second regular expression corresponding to the device temperature, a third regular expression corresponding to the ambient humidity, and a fourth regular expression corresponding to the ambient smoke concentration; Performing an over-limit detection on the ambient temperature time series data using a first regular expression corresponding to the ambient temperature to obtain an ambient temperature over-limit detection result; Perform over-limit detection on the device temperature time series data using the second regular expression corresponding to the device temperature to obtain the device temperature over-limit detection result; Perform over-limit detection on the ambient humidity time series data using the third regular expression corresponding to the ambient humidity to obtain an ambient humidity over-limit detection result; Perform over-limit detection on the time series data of ambient smoke concentration by using the fourth regular expression corresponding to the ambient smoke concentration to obtain an over-limit detection result of the ambient smoke concentration; The operating environment management results are obtained based on the ambient temperature over-limit detection results, equipment temperature over-limit detection results, ambient humidity over-limit detection results and ambient smoke concentration over-limit detection results.
8. The multi-dimensional computer room management method according to claim 5, characterized in that: Based on the power consumption time series data, abnormal power consumption management is performed on the managed computer room to obtain abnormal power consumption management results, including: Obtaining historical power consumption time series data stored in advance or input by staff through human-computer interaction and power consumption anomaly detection results corresponding to the historical power consumption time series data; The improved CW-PSO-CNN algorithm is used to learn the relationship between the historical power consumption time series data and the power consumption anomaly detection results corresponding to the historical power consumption time series data to obtain the power consumption anomaly detection rules; The power consumption time series data is identified through power consumption anomaly detection rules to obtain power consumption anomaly management results.
9. The multi-dimensional computer room management method according to claim 8, characterized in that: The improved CW-PSO-CNN algorithm is used to learn the relationship between the historical power consumption time series data and the power consumption anomaly detection results corresponding to the historical power consumption time series data, and the power consumption anomaly detection rules are obtained, including: Based on the CNN algorithm, the chaotic mapping rule is used to initialize the particle swarm, and the best particle recorder of the swarm is enabled to record and update the best particle of the swarm in real time, and the best individual history recorder is enabled to record and update the best individual history of the particle in real time; The historical electricity consumption time series data is used as the input data of the CNN algorithm, and the power consumption anomaly detection results corresponding to the historical electricity consumption time series data are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the particle swarm; According to the fitness of each particle, the search times of each particle are determined, and according to the search times of each particle, a jump search is performed on the particle to obtain the particle search value; If the fitness of the particle search value increases, the particle search value is used to replace the original particle, and the original particle is used as the fourth target particle; If the fitness of the particle search value decreases, the original particle is used as the fifth target particle and the corresponding particle search value is used as the sixth target particle; The number of particles to be updated is obtained by subtracting the fourth target number of particles from the total number of particles in the particle group; Determine the population dispersion corresponding to the fourth target particle, the fifth target particle, and the sixth target particle, and select the seventh target particle having the same number as the number of particles to be updated according to the population dispersion; The seventh target particle is used to randomly replace particles whose fitness of the particle search value decreases one by one to obtain an updated particle swarm; The historical electricity consumption time series data is used as the input data of the CNN algorithm, and the power consumption anomaly detection results corresponding to the historical electricity consumption time series data are used as the expected output data of the CNN algorithm to obtain the fitness of each particle in the updated particle swarm; According to the fitness of each particle in the updated particle swarm, a fast grouping value corresponding to each particle is determined, and the particle swarm is divided into a first particle subgroup and a second particle subgroup according to the fast grouping value corresponding to the particle; Pair the particles in the first particle subgroup with the particles in the second particle subgroup one by one to obtain particle pairs; Read the best particle of the population from the best particle recorder of the population and read the best historical individual of each particle from the best individual historical recorder; The particles of the second particle subgroup are updated according to the best historical individual of the particle and the particles in the first particle subgroup to obtain an updated second particle subgroup; The particles in the first particle subgroup are updated according to the best particle in the population and the best historical individual of the particle to obtain an updated first particle subgroup; Determine whether the current number of training times is greater than or equal to the maximum number of training times. If so, read the best particle of the population from the best particle recorder of the population, and fuse the best particle of the population with the CNN algorithm to obtain the power consumption anomaly detection rule, otherwise return to the step of executing the jump search.
10. A multidimensional computer room management system, the multidimensional computer room management system is used to execute the multidimensional computer room management method according to any of claims 1 to 9, characterized in that: include: A multi-dimensional data acquisition module, a first management module, a second management module, a third management module and a management result acquisition module; The multi-dimensional data acquisition module is used to acquire the multi-dimensional time series data corresponding to the managed computer room, and pre-process the multi-dimensional time series data to obtain the pre-processed multi-dimensional time series data; Wherein, the multi-dimensional time series data includes video time series data, environmental information time series data and power consumption time series data; The first management module is used to perform personnel intrusion management on the managed computer room based on the video time series data to obtain a personnel intrusion management result; The second management module is used to perform operating environment management on the managed computer room based on the environmental information time series data to obtain an operating environment management result; The third management module is used to perform abnormal power consumption management on the managed computer room based on the power consumption time series data to obtain abnormal power consumption management results; The management result acquisition module is used to obtain multi-dimensional computer room management results based on the personnel intrusion management results, the operating environment management results and the power consumption abnormality management results.
Citation Information
Cited By
Machine room state data processing method and device based on artificial intelligence and storage medium
CN119720111A
Artificial Intelligence-Based Computer Room Status Data Processing Method, Device, and Storage Medium
CN119720111B
