An intelligent adjustment method for maintaining positive pressure in a large data center machine room based on deep learning

Abstract

A kind of based on deep learning's intelligent adjustment method of maintaining data center room positive pressure, comprising: obtaining training set F;Data cleaning is carried out to F, supplement missing value and handle abnormal value, continuous variable numerical value is discretely handled, and derived feature set F1 is formed;Feature data extraction is carried out to F1, and feature data set S and test data set C are extracted;Entropy and information gain ratio algorithm is used to process S, and decision tree data set TREE is formed;Pruning is carried out to TREE, and branch that contributes less to model generalization ability is deleted, and training set D is formed;The data of D is used to construct decision tree model, and model parameter is adjusted in training process;For the decision result obtained, the accuracy and generalization ability of decision tree model are evaluated using the data in C.The present application can realize the deep learning and intelligent adaptive adjustment of data center indoor positive pressure maintenance and different operating environment, and maintain the positive pressure operation of each functional room of data center.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer science, in particular to an intelligent adjustment method for maintaining positive pressure in a large data center room based on deep learning. TECHNICAL BACKGROUND

[0002] Dust management is crucial for large data center rooms, as it is directly related to the safety, economy and compliance of equipment operation. As an invisible threat, dust can penetrate the protective layer of electronic equipment, deposit on the circuit board and cause short circuits, overheating and even equipment failure, threatening the stability and safety of the data center. Maintaining positive pressure in the data center room not only effectively prevents dust accumulation, prolongs the service life of core equipment such as servers and switches, but also continuously maintains the cleanliness of cooling equipment fans or fins, avoids a decrease in cooling efficiency, maintains energy efficiency, helps to keep the data center power usage effectiveness (PUE) at a normal level, and promotes energy saving and emission reduction.

[0003] As the size of individual data centers continues to expand, data center buildings are modularly established in the park, the room area is growing in an open manner, and the functions of each room are becoming more independent, so the cleaning needs have changed. The current mainstream industrial fresh air equipment on the market, which is suitable for dust filtration and air renewal in a single room, cannot meet the needs of maintaining indoor positive pressure and keeping the room clean in large-scale cluster data center rooms, and there is an urgent need for a method for maintaining indoor positive pressure in data center rooms that can be deeply learned, intelligently adjusted and continuously self-adjusted for a building or several buildings, a floor or several floors of rooms. SUMMARY

[0004] To solve the problems in the prior art, the present application provides an intelligent adjustment method for maintaining positive pressure in a large data center room based on deep learning, which can maintain indoor positive pressure in a data center and intelligently adapt to different operating environments through deep learning, and maintain positive pressure operation in each functional room of the data center.

[0005] An intelligent adjustment method for maintaining positive pressure in a data center room based on deep learning, comprising the following steps:

[0006] Step 1: Data collection and extraction: obtain a training set F: collect various types of data inside and outside each room, including indoor pressure P in , outdoor pressure P out , inlet and outlet ventilation damper opening θ in and θ ex , fresh air system fan speed V, and fresh air equipment opening and closing time interval T. The collected data is preliminarily filtered according to the fresh air equipment opening and closing time interval T, and the maximum and minimum values are filtered out to form the training set F.

[0007] Second step, data preprocessing: clean the data in the training set F, supplement the missing values and handle the outliers, discretize the continuous variable values, and form the derived feature set F1 according to the influence relationship between the fresh air equipment opening and closing time interval T and the indoor and outdoor pressure difference P;

[0008] Third step, extracting feature data set S and test data set C: extracting feature data from the derived feature set F1, the indoor and outdoor pressure difference P, the fresh air equipment opening and closing time interval T, the fan speed V and the ventilation inlet and outlet opening degree θ in and θ ex、 As the main feature data, generate the time period feature data set S i ={P i , T i , V i , θ in , θ ex}, then extract the time period feature data set S i , generate the key feature data set S={S1+S2+S3+···+S n}; Then use the method of human intervention and mathematical method to extract a standard sample data to form the related data of the test data set C, C={δP1+δP2+δP3+···+δP n}, where i=0, 1, 2...n;

[0009] Fourth step, forming the decision tree data set TREE: processing the feature data set S using the entropy and information gain ratio algorithm, selecting the best feature and splitting point, creating the root node of the tree based on the selected best feature and splitting point, recursively executing the same process for each child node until the stopping condition is reached, forming the decision tree data set TREE, the stopping condition is that all samples in the node belong to the same class, or the maximum tree depth is reached, or the number of samples in the node is less than a preset threshold;

[0010] Fifth step, decision tree pruning: pruning the decision tree data set TREE from the bottom up to delete branches that contribute less to the model generalization ability, forming the training set D;

[0011] Sixth step, decision tree model construction and training: using the data of the training set D to construct the decision tree model, adjusting the model parameters during the training process, the model parameters include the maximum depth of the tree and the minimum leaf node sample number;

[0012] Seventh step, decision tree model evaluation: using the data in the test data set C to evaluate the accuracy and generalization ability of the decision tree model, re-discretizing the collected data for the inappropriate decision tree model, repeating the previous six steps until the decision tree model evaluation result meets the test requirements.

[0013] Further, in the first step, full-quantity sensors are deployed to collect various types of related data in and outside the room.

[0014] Further, in the second step, the missing values are supplemented, specifically, for the missing data of the fan speed V of the fresh air system and the opening degree θ of the inlet and outlet ventilation dampers in the case of non-starting of the fan. in and θ ex are supplemented by 0 processing, and the indoor pressure P in , the outdoor pressure P out that are negative in part are modified to 0.

[0015] The abnormal values are processed, specifically, the abnormal values are identified and removed by Z-score or IQR method.

[0016] The discrete processing is specifically: the regular pressure data P, T, V and θ in and θ ex in a time interval are discretely processed by using a discrete mathematical model to improve the data generalization ability.

[0017] Further, the decision tree model constructed in the sixth step is used to identify abnormal data and misjudgment data in the training set D classification, and the specific way of identifying abnormal data and misjudgment data is: 1) whether the opening degrees θ in and θ ex of the inlet and outlet ventilation dampers in the training data conform to the actual openable percentage, and the inappropriate data is removed; 2) whether the fan speed V of the fresh air system in part of the module in the training data exceeds the actual operating speed range, and the inappropriate data is removed.

[0018] Further, when the accuracy and generalization ability of the decision tree model are evaluated by using the data in the test set C, the confusion matrix, ROC curve and AUC value index are used to evaluate the performance of the model, and the evaluation result of the decision tree model meets the standard of 98% accuracy and 0.95 AUC value.

[0019] The present application uses deep learning and remote intelligent control technology to collect the running environment of one or more, one or more data center machine rooms, evaluate the running quality, deeply learn the historical data and autonomously adapt to the adjustment to realize the continuous maintenance of the positive pressure in the data center and the continuous cleaning of the air dust, and provide a set of intelligent adjustment method for maintaining the positive pressure of the large data center machine room based on deep learning. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of the data governance method for maintaining the positive pressure in the data center based on the decision tree model of the present application;

[0021] Figure 2Is a large data center interlayer data room positive pressure maintenance system plane logic diagram. Embodiments

[0022] To make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0023] Please refer to Figure 1 The embodiments of the present application provide a deep learning-based intelligent adjustment method for maintaining the positive pressure of a data center room, comprising the following steps:

[0024] First step, data acquisition and extraction

[0025] Data acquisition: as shown in the figure, high-precision sensors are installed in each room to collect data in real time; Figure 2 Indoor pressure P in (unit: Pa)

[0026] Outdoor pressure P out (unit: Pa)

[0027] Vent opening degree θ in and θ ex (unit: percentage, 0-100%)

[0028] Fresh air system fan speed V (unit: m / s)

[0029] Power consumption E (unit: kW·h)

[0030] Fresh air equipment opening and closing time interval T (unit: minutes)

[0031] Preliminary screening: the collected data is preliminarily screened with a time interval T=5 minutes as the dimension, and the maximum and minimum values are filtered out to form a training set F; a sliding window method (for example, 3 consecutive time points) is used to smooth the data to reduce the noise influence.

[0032] Specific case: data center A has 10 rooms, and data is collected every 5 minutes in each room, and a total of 7 days of data is collected. After preliminary screening, a training set F containing about 20,160 records is formed

[0033] Second step, data preprocessing

[0034]

[0035] ​Cleaning data: 1) Fill in missing values, fill in 0 for missing data θ of part of fan non-starting data V and damper closing, modify part of point pressure difference P to 0 for negative number; 2) Process abnormal values, identify and eliminate abnormal values by Z-score or IQR method.

[0036] Discretization processing: Discretization processing is performed on continuous variable values in data P, T, V, E and θ, for example, indoor pressure Pin is divided into low, medium and high three grades.

[0037] Derivative feature generation: According to the influence relationship of fresh air equipment opening and closing time interval T and indoor and outdoor pressure difference P = P in -P out , a derivative feature set F1 is formed;

[0038] Specific case:

[0039] Data center A collected data in the preprocessing process found that part of the time period of power consumption E exists obvious abnormal value, through IQR method eliminated these abnormal values, and supplemented the missing values. The data set F1 after preprocessing contains about 20,160 records, which is used for subsequent analysis

[0040] Third step, extract feature data set S and test data set C

[0041] Feature data extraction: Extract main feature data: indoor and outdoor pressure difference P, fresh air equipment opening and closing time interval T, fan speed V and ventilation inlet and outlet opening θ in and θ ex , generate each time period feature data set Si = {P i ,T i ,V i ,θ in ,θ ex}.

[0042] Generate key feature data set S: merge each time period feature data set Si to form key feature data set S = {S1+S2+S3+...+S n}.

[0043] Generate test data set C: calculate the pressure drop speed δp of each room by artificial intervention method and mathematical method, and arrange the δp values in ascending order, extract a standard sample data to form test set: C = {δP1+δP2+δP3+...+δP n}, where i = 0, 1, 2...n.

[0044] The specific algorithm of the artificial intervention method of test data set C is as follows:

[0045] ① Collect sensor data of each room, including fresh air fan speed V, inlet and outlet opening degree θ in and θ ex , indoor and outdoor pressure difference P.

[0046] ② Record the changes of these data and mark the corresponding pressure drop speed δp.

[0047] Establish a mathematical model based on the above formula to predict the pressure drop speed δp of each room, the specific formula is as follows:

[0048] δp=f(V,θ in ,θ ex ,P)

[0049] Here, f is a function that can be customized according to the actual application scenario. A simple typical model example used by the invention is as follows:

[0050] δp=C V *V+C in *θ in +C ex *θ ex +C P *P

[0051] Where:

[0052] C V is a coefficient related to fan speed;

[0053] C in is a coefficient related to fresh air inlet damper opening degree;

[0054] C ex is a coefficient related to fresh air outlet damper opening degree;

[0055] C P is a coefficient related to indoor and outdoor pressure difference;

[0056] Suppose that by using linear regression algorithm to statistically analyze the collected data, the following fitting coefficients are obtained:

[0057] C V =0.5, C in =0.3, C ex =0.4, C P =0.2

[0058] Then, for a certain room, if it is measured that:

[0059] Fresh air fan speed V=10m / s

[0060] Damper opening degree C in =0.8, C ex =1

[0061] The indoor and outdoor pressure difference P = 5 Pa

[0062] The pressure drop speed δp of the room in the time T can be calculated as:

[0063] δp = 0.5 * 10 + 0.3 * 0.8 + 0.4 * 1 + 0.2 * 5 = 5 + 0.24 + 0.4 + 1 = 6.64 Pa / s

[0064] Using the collected data, the change rule of δp under different conditions is analyzed.

[0065] The standard data of the sample can be calculated as the δP data set of each room in the continuous span of time T, as the test data set C,

[0066] C = {δP1+ δP2+ δP3+ ··· + δP n}, where i = 0, 1, 2...n;

[0067] Specific case:

[0068] Data center A selects 7 days of data as the training set, and randomly selects 1,440 records (i.e. 1 day of data) for calculation to form the test set C, which is used to evaluate the performance of the model.

[0069] The feature data set S contains the remaining 6 days of data, a total of 18,720 records.

[0070] Step 4, form the decision tree data set TREE

[0071] The entropy value and information gain ratio algorithm is used to process the feature data set S, select the best feature and split point, and create the root node of the tree based on the selected best feature and split point. The same process is recursively performed for each child node until the stopping condition is reached, forming the decision tree data set TREE;

[0072] (1) Entropy algorithm

[0073] The entropy value is used to measure the purity of the data set S. The higher the entropy, the greater the uncertainty of the data set; otherwise, the uncertainty is smaller, and its algorithm is

[0074]

[0075] Where p i represents the proportion of the i-th sample containing feature A in the total sample.

[0076] (2) Information gain algorithm

[0077] Information gain IG(A) is the amount of information gain obtained after using the feature A in the set S to divide the data set S, which is used to measure the reduction of uncertainty brought by an attribute to partition the data set, and its algorithm is as follows:

[0078]

[0079] Wherein:

[0080] Entropy(S) represents the entropy of the data set S.

[0081] Sv represents the data subset taking value v on attribute A.

[0082] |S| and |Sv| represent the sample number of the data set D and the subset Dv, respectively.

[0083] Values(A) represents the set of all possible values of attribute A.

[0084] (3) Information gain ratio algorithm

[0085] The information gain ratio algorithm is based on the information gain, and by dividing the inherent value of the attribute, it can better reflect the usefulness of the attribute. The specific algorithm is as follows:

[0086] Define IV(A) as the inherent value of the sample, then:

[0087]

[0088] The information gain ratio Gain_Ratio(A) is calculated as follows:

[0089]

[0090] IV(A) represents the inherent value of the sample in the feature A in the set S

[0091] IG(A) represents the information gain of the set S calculated according to the feature A.

[0092] Specific case:

[0093] Based on the feature data set S of 18,720 records, the entropy value and the information gain ratio algorithm are used to select the best feature and split point, to construct the decision tree root node, and recursively partition the data until the stopping condition is reached (all samples in the node belong to the same class, or the maximum tree depth is reached, or the number of samples in the node is less than the preset threshold), and finally the decision tree data set TREE contains 15 nodes, and the maximum depth is 5.

[0094] Step 5, decision tree pruning, forming the training set D

[0095] The cost complexity pruning algorithm is used to perform post-pruning on the data set D, removing branches that contribute less to the model's generalization ability from the bottom up, forming the training set D;

[0096] The specific steps of the cost complexity pruning algorithm include:

[0097] 1) Initialization, select a validation set, start from the root node, recursively traverse the entire decision tree;

[0098] 2) Calculate the error, for each internal node t, calculate the error err(Tt) when not pruned and the error err(Tt') when converted to a leaf node;

[0099] 3) Compare and prune, for each internal node t, calculate the cost complexity Rα(Tt) and Rα(Tt') before and after pruning. If Rα(Tt') < Rα(Tt), prune the node, i.e. convert it to a leaf node;

[0100] 4) Recursion, for the pruned tree, continue to recursively execute the above steps until no further pruning is possible;

[0101] 5) Select the optimal α, by evaluating the performance of the decision tree under different α values on the validation set, select a best α value that makes the model perform best on the validation set.

[0102] Specific case:

[0103] The decision tree is pruned, and branches with less contribution are removed, forming a simplified training set D, reducing the risk of overfitting. The simplified training set D contains 14,526 records, which are used for subsequent model training.

[0104] Step 6, decision tree model construction and training

[0105] Use the training set D data to build a decision tree model, adjust the model parameters (such as the maximum depth of the tree, the minimum number of leaf node samples) to optimize the model performance. Use grid search and cross-validation method to optimize the hyperparameters.

[0106] Specific case:

[0107] Data center A uses the training set D to build a decision tree model, and optimizes the model parameters through grid search and cross-validation, finally the maximum depth of the model is set to 6, and the minimum number of leaf node samples is set to 10, achieving good generalization performance. The training set D contains 14,526 records, ensuring that the model has enough data support

[0108] Step 7, decision tree model evaluation

[0109] Model evaluation: Evaluate the accuracy and generalization ability of the decision tree model using the data in test set C. Use confusion matrix, ROC curve and AUC value to evaluate the model performance. If the model is not suitable, re-discretize the data and repeat the first six steps until the model evaluation results meet the requirements.

[0110] Specific case:

[0111] Evaluate the accuracy and generalization ability of the decision tree model of data center A using test set C. It is found that the model performs slightly lower than expected in some time periods. By re-discretizing the data of the machine room in certain time periods and re-training the model, the accuracy of 98% and AUC value of 0.95 are finally achieved, meeting the needs of the data center.

[0112] Test set C contains 1,440 records, covering all-day time period data, ensuring the comprehensiveness and accuracy of the evaluation.

[0113] Simplify the evaluation algorithm:

[0114] Suppose we choose data Si in training set D, Si = {P i ,T i, V i ,θ in ,θ ex} contains five feature attributes P d、 θ′ in and θ′ ex as evaluation data, the sample data is as follows:

[0115] Table 1 sample data

[0116] P i (Pa)]]> T i (min)]]> V i (v)] in (%)]]> ​ ex (%)]]> ​ 3 5 1200 40 60 7 5 1330 60 40 5 5 1250 50 50 ... ... ... ... ...

[0117] (1) Input the data Si in training set D, Si = {P i ,T i, V i ,θ in ,θ ex} into the decision model to guide the operation of the fresh air system:

[0118] 1. Input T i, V i ,θ in ,θ ex into the decision tree model.

[0119] 2. Get the current indoor area pressure set P in and outdoor pressure P out every time Ti.

[0120] (2) Calculate the new pressure difference

[0121] 1. Calculate the current positive pressure difference set P d ′ = P in -P out ;

[0122] (3) Data acquisition

[0123] 1. Rewrite the record S i ′

[0124] S i ′ = {P i , T i , V i , θ in , θ ex}

[0125] 2. Collect data of each time period Si to form a new feature set S′,

[0126] S′ = {S1′ + S2′ + S3′ +... + S n ′}

[0127] (4) Calculate the new test data set C

[0128] 1. Calculate the pressure drop speed δp i ′

[0129] 2. As test data set C′, C′ = {δP1′ + δP2′ + δP3′ +... + δP n ′},

[0130] Where i = 0, 1, 2...n;

[0131] (5) Compare the test data set C with C′, and calculate whether the data entries with a positive pressure difference value deviation less than 2Pa reach more than 98% accuracy and 0.95 AUC value, if not, re-discretize the data of the specific time period of the machine room and re-train the model, finally reaching 98% accuracy and 0.95 AUC value, meeting the operation requirements of the data center

[0132] In this way, the intelligent adjustment and evaluation process of the decision tree model is completed. With the passage of time, the pressure in the room can be continuously maintained near the target positive pressure set by man through continuous intelligent adjustment. The algorithm case only considers the influence of internal and external pressure, air door inlet and outlet opening degree, fan speed, air inlet and outlet time on indoor pressure, and in a real case, factors such as indoor and outdoor temperature, humidity, natural wind pressure and power consumption can be considered comprehensively, and the decision tree model is more complex, but the principle is consistent; When we use internal and external pressure, air door inlet and outlet angle, fan speed and air inlet and outlet time for model training and decision, we can effectively adapt to the influence of external factors such as indoor and outdoor temperature, humidity, wind speed and weather, and intelligently adjust the equipment operating state to adapt to different conditions.

[0133] The present application uses a decision tree model to make decisions on indoor and outdoor temperature, humidity, pressure, fan speed, air door opening degree, air inlet and outlet time and other data in each room in the data center. In a few minutes, appropriate fresh air system adjustment input values can be given for the operation of different rooms, achieving the purpose of intelligent adjustment of maintaining positive pressure in each room, and maximizing the reduction of artificial intervention and cost waste caused by all-weather opening of the fresh air system. The innovation of the present application is to use big data and decision tree intelligent adjustment algorithm to effectively cope with the operation requirements of one or several buildings, one or several floors that need to maintain indoor positive pressure, and to provide an intelligent adjustment algorithm for maintaining a clean digital equipment operating environment for data center operators with the smallest investment. Compared with the traditional fresh air algorithm, the intelligent adjustment method has strong adaptability, flexible expansion, high intelligence, and can effectively meet the intelligent adjustment requirements of multiple target rooms with positive pressure environment, saving labor and energy consumption costs.

[0134] The above is only an example embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by any person skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A deep learning-based intelligent adjustment method for maintaining positive pressure in a data center room, characterized in that: The steps include: Step 1: Data collection and extraction: Obtain training set F: Collect various relevant data inside and outside the room, including indoor pressure P, for each room. in , outdoor pressure P out 、Inlet and outlet ventilation damper opening θ in and θ ex , the fan speed V of the fresh air system, the time interval T between the start and stop of the fresh air equipment, and the various types of collected data are preliminarily screened based on the time interval T between the start and stop of the fresh air equipment, filtering out the maximum and minimum values ​​to form a training set F; Step 2: Data preprocessing: Clean the data in the training set F, fill in missing values ​​and handle outliers, discretize the continuous variable values, and form a derived feature set F1 based on the influence relationship between the opening and closing time interval T of the fresh air equipment and the indoor and outdoor pressure difference P; The third step is to extract the feature data set S and the test data set C: extract the feature data of the derived feature set F1, and extract the indoor and outdoor pressure difference P, the time interval between opening and closing of the fresh air equipment T, the fan speed V and the ventilation inlet and outlet opening θ. in and θ ex、 Generate feature data set S for each time period as the main feature data i ={P i , T i , V i ,θ in ,θ ex }, then extract the feature data set S for each time period i , generate key feature data set S = {S1+S2+S3+···+S n }; Then, a standard sample data is extracted by human intervention and mathematical methods to form the relevant data of the test data set C, C={δP1+δP2+δP3+···+δP n }, where i = 0, 1, 2, ... n; Step 4: Form a decision tree data set TREE: Use entropy and information gain ratio algorithms to process the feature data set S, select the best features and splitting points, split the data based on the selected best features and splitting points to create the root node of the tree, and recursively perform the same process on each child node until the stopping condition is reached to form a decision tree data set TREE. The stopping condition is that all samples in the node belong to the same class, or the maximum tree depth is reached, or the number of samples in the node is less than a preset threshold; Step 5: Prune the decision tree to form the training set D: Perform post-pruning on the decision tree data set TREE, and delete the branches that contribute less to the generalization ability of the model from the bottom up to form the training set D; Step 6: Decision tree model construction and training: Use the data from the training set D to build a decision tree model. During the training process, adjust the model parameters, including the maximum depth of the tree and the minimum number of leaf node samples. Step 7: Decision tree model evaluation: Use the data in the test data set C to evaluate the accuracy and generalization ability of the decision tree model. For inappropriate decision tree models, re-discretize the collected data and repeat the first six steps until the decision tree model evaluation results meet the test requirements.

2. The deep learning-based intelligent adjustment method for maintaining positive pressure in a data center room according to claim 1, characterized in that: In the first step, a full range of sensors are deployed to collect various relevant data inside and outside the room.

3. The deep learning-based intelligent adjustment method for maintaining positive pressure in a data center computer room according to claim 1, characterized in that: The second step is to fill in the missing values, specifically, for the missing data of the fan speed V and the damper closing status of the fresh air system where some fans are not started, the inlet and outlet ventilation damper opening θ in and θ ex Perform zero filling processing and change the indoor pressure P of some points with negative pressure difference to in , outdoor pressure P out Modify to 0; The specific processing of abnormal values ​​is as follows: identifying and eliminating outliers through Z-score or IQR method; Discrete processing is specifically: for the regular pressure data P, T, V and θ within a period of time in and θ ex , using discrete mathematical models for discrete processing to improve data generalization capabilities.

4. The deep learning-based intelligent adjustment method for maintaining positive pressure in a data center computer room according to claim 1, characterized in that: The decision tree model constructed in the sixth step is used to identify abnormal data and misjudgment data in the training set D classification. The specific methods of identifying abnormal data and misjudgment data are as follows: 1) Identify the opening degree θ of the inlet and outlet ventilation dampers in the training data in and θ ex Whether it is consistent with the actual opening percentage, and eliminate inappropriate data; 2) Identify whether the fan speed V of the fresh air system of some modules in the training data exceeds the actual operating speed range, and eliminate inappropriate data.

5. The deep learning-based intelligent adjustment method for maintaining positive pressure in a data center computer room according to claim 1, characterized in that: In the seventh step, the accuracy and generalization ability of the decision tree model are evaluated using the data in the test set C. The confusion matrix, ROC curve and AUC value indicators are used to evaluate the model performance. The standard for the decision tree model evaluation results to meet the test requirements is to achieve an accuracy of 98% and an AUC value of 0.95.

Images