Data center circuit breaker abnormal heating automatic detection method based on voltage drop
By constructing an anomaly detection model based on voltage drop in data center circuit breakers, abnormal heating risks can be identified, solving the problems of adding new sensors and the difficulty of construction. This achieves efficient and accurate heating detection and improves the safety of the power distribution system.
Patent Information
- Application Number
- CN202410986718.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-12
AI Technical Summary
Existing technologies are insufficient to effectively monitor the surface temperature of data center circuit breakers, making it difficult to detect potential overheating issues in a timely manner. Furthermore, the addition and installation of new sensors present significant challenges.
By using a voltage drop-based method, three-phase voltage time series are obtained from voltage monitoring points, an anomaly detection model is constructed to identify the abnormal heating risk of circuit breakers without the need for additional sensors or power outages.
It enables automatic detection of abnormal overheating in circuit breakers, reducing detection costs and difficulty, improving identification accuracy, and ensuring the safety of data center power distribution systems.
Smart Images

Figure CN118937980A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data center infrastructure operation and maintenance, and more particularly to a method for automatically detecting abnormal heating of a data center circuit breaker based on voltage drop. Background Art
[0002] As a core component of the data center infrastructure, the reliable operation of the power distribution system is directly related to whether the cloud computing, artificial intelligence, large models and other businesses carried by the data center can continue uninterrupted. As a core component of the power distribution system, the safe operation of the circuit breaker is directly related to whether there is a power outage in the power distribution system. In daily operation, the circuit breaker will generate heat due to loose connections between the terminals and the conductors, or low contact spring pressure, poor contact and other reasons. Once the circuit breaker is seriously heated, it will cause power outages and business interruptions at the least, or even induce production safety accidents at the worst. Therefore, it is necessary to promptly detect the hidden dangers of circuit breaker heating and deal with them effectively.
[0003] In order to promptly discover the potential heating hazards of circuit breakers, the most direct method is to conduct online monitoring of the circuit breaker surface temperature. The dynamic environment monitoring system currently built in the data center can monitor the ambient temperature of the power distribution equipment, the voltage, current, and power of the power distribution equipment, but few can directly monitor the circuit breaker surface temperature. Moreover, the current use of this method requires the addition of new sensors, and for the systems that have been put into operation, power outages need to be coordinated during the implementation process, which has high implementation costs and construction difficulties.
[0004] Therefore, how to reduce the detection cost and construction difficulty of the circuit breaker surface temperature is a problem that technical personnel in this field need to solve urgently. Summary of the invention
[0005] In view of the above problems, the present invention provides a method for automatically detecting abnormal heating of a data center circuit breaker based on voltage drop, so as to at least solve some of the technical problems mentioned in the above background technology.
[0006] In order to achieve the above object, the present invention adopts the following technical solution:
[0007] A method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop comprises the following steps:
[0008] Determine the upper voltage monitoring point and the lower voltage monitoring point of the target circuit breaker;
[0009] Obtaining, within a preset time period, an upper three-phase voltage time series corresponding to the upper voltage monitoring point and a lower three-phase voltage time series corresponding to the lower voltage monitoring point;
[0010] Preprocessing the upper three-phase voltage time series and the lower three-phase voltage time series;
[0011] Based on the preprocessed upper end three-phase voltage time series and lower end three-phase voltage time series, obtaining the three-phase voltage drop time series of the target circuit breaker;
[0012] The three-phase voltage drop time series is input into an abnormality detection model to obtain an abnormal heating detection result of the target circuit breaker.
[0013] Further:
[0014] The upper three-phase voltage time series includes an upper A-phase voltage time series, an upper B-phase voltage time series and an upper C-phase voltage time series;
[0015] The lower end three-phase voltage time series includes a lower end A-phase voltage time series, a lower end B-phase voltage time series and a lower end C-phase voltage time series.
[0016] Further, the preprocessing of the upper three-phase voltage time series and the lower three-phase voltage time series specifically includes:
[0017] Respectively removing voltage data outside a preset voltage value range in the upper three-phase voltage time series and the lower three-phase voltage time series;
[0018] Resampling and data alignment processing are performed on the voltage data in the upper three-phase voltage time series and the lower three-phase voltage time series respectively.
[0019] Furthermore, the numbers of voltage data in the preprocessed upper three-phase voltage time series and the lower three-phase voltage time series are consistent.
[0020] Furthermore, the three-phase voltage drop time series includes a phase A voltage drop time series, a phase B voltage drop time series and a phase C voltage drop time series, which are expressed as:
[0021]
[0022]
[0023] Among them, ΔU a Represents the time series of the voltage drop of phase A; ΔU b represents the time series of the voltage drop of phase B; ΔU c represents the time series of the voltage drop of phase C; It represents the time series of the voltage of phase A at the upper end; represents the time series of the voltage of phase B at the upper end; represents the time series of the upper C phase voltage; It represents the time series of the voltage of phase A at the lower end; represents the time series of the voltage of phase B at the lower end; represents the time series of the voltage of phase C at the lower end; Δa m Indicates ΔU a The mth voltage drop data in; Δb m Indicates ΔU b The mth voltage drop data in; Δc m Indicates ΔU c The mth voltage drop data in .
[0024] Furthermore, the steps of constructing the anomaly detection model include:
[0025] S1. Construct a sliding time window: construct a left window and a right window for each voltage drop data in each phase voltage drop time series;
[0026] S2, respectively calculating the median of the voltage drop data in each of the left window and the right window; obtaining the median time series of the left window and the median time series of the right window;
[0027] S3, obtaining a median difference time series according to the left window median time series and the right window median time series;
[0028] S4, obtaining an absolute value sequence of the median difference time series, and calculating the first quartile and the third quartile in the absolute value sequence;
[0029] S5. Obtain a quantile difference according to the first quartile and the third quartile;
[0030] S6. Construct an abnormality judgment condition according to the first quartile, the third quartile and the quartile difference, and then obtain a voltage abnormality value sequence; if the voltage abnormality value sequence is empty, there is no abnormal heating; otherwise, there is an abnormal heating.
[0031] Furthermore, the abnormality judgment condition is expressed as:
[0032]
[0033] Where |Δx i | represents the absolute value of the i-th median difference in the absolute value sequence; Q1 represents the first quartile; Q3 represents the third quartile; IQR represents the quantile difference; a represents the abnormal parameter.
[0034] Furthermore, the step of constructing the anomaly detection model further includes:
[0035] S7. Each abnormal value in the voltage abnormal value sequence has a corresponding value in the median difference time series; the abnormal value whose corresponding value is greater than the preset value is taken as the final abnormal value to obtain the final voltage abnormal value sequence; if the final voltage abnormal value sequence is empty, there is no abnormal heating; otherwise, there is abnormal heating.
[0036] Furthermore, it also includes:
[0037] Obtaining the time of each abnormal value in the voltage abnormal value sequence to form a voltage abnormal time series;
[0038] Inputting the three-phase voltage drop time series into a large language model to obtain a large language abnormal heating detection result of the target circuit breaker;
[0039] The voltage abnormality time series is ANDed or ORed with the abnormal fever detection result of the loudspeaker to obtain a preliminary abnormality detection result.
[0040] Furthermore, it also includes:
[0041] Obtaining, within a preset time period, an upper three-phase load current time series corresponding to the upper voltage monitoring point and a lower three-phase load current time series corresponding to the lower voltage monitoring point;
[0042] Preprocessing the upper three-phase load current time series and the lower three-phase load current time series;
[0043] Based on the preprocessed upper three-phase load current time series and lower three-phase load current time series, obtaining a three-phase load current difference time series of the target circuit breaker;
[0044] Inputting the three-phase load current difference time series into the anomaly detection model to obtain a current anomaly value sequence;
[0045] Obtaining the time of each abnormal value in the current abnormal value sequence to form a current abnormal time series;
[0046] taking the same time value in the voltage abnormal time series and the current abnormal time series as a pseudo abnormal value;
[0047] removing pseudo abnormal values in the voltage abnormal time series to form a final voltage abnormal time series;
[0048] If the final voltage abnormal time series is empty, it is determined that the target circuit breaker does not have heating phenomenon.
[0049] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop, which has the following beneficial effects:
[0050] The present invention uses voltage drop as a key feature to identify the abnormal heating risk of the target circuit breaker. The entire identification process does not require the addition of new sensors or coordinated power outages, which effectively reduces the cost and difficulty of detection, helps to promptly discover circuit breaker heating hazards, and further improves the operating safety of the data center's power distribution system.
[0051] The present invention combines the three-phase load current time series, can avoid false detection caused by current increase, and improve the accuracy of the recognition result.
[0052] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the present invention.
[0053] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0055] Figure 1 A schematic flow chart of a method for automatically detecting abnormal heating of a data center circuit breaker based on voltage drop provided in an embodiment of the present invention.
[0056] Figure 2 A schematic diagram of the anomaly detection model construction process provided in an embodiment of the present invention.
[0057] Figure 3 A schematic diagram of abnormal fever detection results provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0059] Embodiment 1:
[0060] See also Figure 1 As shown, an embodiment of the present invention discloses a method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop, comprising the following steps:
[0061] Determine the upper voltage monitoring point and the lower voltage monitoring point of the target circuit breaker; obtain the upper three-phase voltage time series corresponding to the upper voltage monitoring point and the lower three-phase voltage time series corresponding to the lower voltage monitoring point within a preset time period; preprocess the upper three-phase voltage time series and the lower three-phase voltage time series; obtain the three-phase voltage drop time series of the target circuit breaker based on the preprocessed upper three-phase voltage time series and the lower three-phase voltage time series; input the three-phase voltage drop time series into the anomaly detection model to obtain the abnormal heating detection result of the target circuit breaker.
[0062] This method uses voltage drop as a key feature to identify the abnormal heating risk of the target circuit breaker, and the entire identification process does not require the addition of new sensors or coordinated power outages, effectively reducing the cost and difficulty of detection. The above part is explained in detail below.
[0063] 1. Voltage monitoring data acquisition:
[0064] First, the upper voltage monitoring point and the lower voltage monitoring point of the circuit breaker are determined according to the electrical connection relationship of the target circuit breaker; then, the historical voltage data of the upper voltage monitoring point and the lower voltage monitoring point in a preset time period (for example, nearly 3 hours, 12 hours, etc.) are obtained from the data center dynamic environment monitoring system to form a three-phase voltage time series; specifically, it includes the upper three-phase voltage time series corresponding to the upper voltage monitoring point, and the lower three-phase voltage time series corresponding to the lower voltage monitoring point;
[0065] The above upper three-phase voltage time series includes the upper A phase voltage time series, the upper B phase voltage time series and the upper C phase voltage time series, which are expressed as:
[0066] U a up ={a1',a'2,···,a' n1}
[0067]
[0068] in, It represents the time series of the voltage of phase A at the upper end; represents the time series of the voltage of phase B at the upper end; represents the time series of the upper C phase voltage; a' n1 express The n1th voltage data in b' n2 express The n2th voltage data in; c' n3 express The n3th voltage data in;
[0069] The above lower end three-phase voltage time series includes the lower end A phase voltage time series, the lower end B phase voltage time series and the lower end C phase voltage time series, which are expressed as:
[0070]
[0071] in, It represents the time series of the voltage of phase A at the lower end; represents the time series of the voltage of phase B at the lower end; represents the time series of the voltage of phase C at the lower end; a' n '4 means The n4th voltage data in b' n '5 means The n5th voltage data in; c' n '6 means The n6th voltage data in;
[0072] 2. Voltage monitoring data preprocessing:
[0073] Preprocessing the upper three-phase voltage time series and the lower three-phase voltage time series includes:
[0074] (1) Respectively remove the voltage data outside the preset voltage value range in the upper three-phase voltage time series and the lower three-phase voltage time series; for the preset voltage value range, the low-voltage cabinet can be set to 0-1000V, and the 10kV high-voltage cabinet can be set to 0-20kV;
[0075] (2) Resample and align the voltage data in the upper three-phase voltage time series and the lower three-phase voltage time series respectively; the resampling time interval can be set to 1 minute, 5 minutes, etc. according to the specific situation; when there is no collected value at the corresponding time, it can be filled by linear interpolation method;
[0076] The preprocessed upper three-phase voltage time series is expressed as:
[0077]
[0078] Among them, a' m Represents the upper end A phase voltage time series after preprocessing The mth voltage data in b' m Represents the upper end B phase voltage time series after preprocessing The mth voltage data in; c' m Represents the upper C phase voltage time series after preprocessing The mth voltage data in;
[0079] The preprocessed lower three-phase voltage time series is expressed as:
[0080]
[0081] Among them, a” m Represents the preprocessed lower end A phase voltage time series The mth voltage data in b' m ' represents the preprocessed lower end B phase voltage time series The mth voltage data in; c" m Represents the preprocessed lower end C phase voltage time series The mth voltage data in;
[0082] The number of voltage data in the preprocessed upper three-phase voltage time series and the lower three-phase voltage time series is consistent.
[0083] 3. Calculation of target circuit breaker voltage drop:
[0084] Based on the above preprocessed upper three-phase voltage time series and lower three-phase voltage time series, the three-phase voltage drop time series of the target circuit breaker is obtained; the three-phase voltage drop time series includes the A phase voltage drop time series, the B phase voltage drop time series and the C phase voltage drop time series; expressed as:
[0085]
[0086] Among them, ΔU a Represents the time series of the voltage drop of phase A; ΔU b represents the time series of the voltage drop of phase B; ΔU c represents the time series of the voltage drop of phase C; It represents the time series of the voltage of phase A at the upper end; represents the time series of the voltage of phase B at the upper end; represents the time series of the upper C phase voltage; It represents the time series of the voltage of phase A at the lower end; represents the time series of the voltage of phase B at the lower end; represents the time series of the voltage of phase C at the lower end; Δa m Indicates ΔU a The mth voltage drop data in; Δb m Indicates ΔU b The mth voltage drop data in; Δc m Indicates ΔU c The mth voltage drop data in ; Δa1=a1'-a1", and so on.
[0087] 4. Anomaly detection model construction, see Figure 2 As shown, the specific construction steps include the following:
[0088] S1. Construct a sliding time window: construct a left window and a right window for each voltage drop data in each phase voltage drop time series;
[0089] The current point is contained in the right window; the left window and the right window are of equal size, and the number of data points contained in each is c, and c is generally 5 to 10; the larger the resampling time interval in the preprocessing process, the smaller the value of c. Since c data points are required to construct a window, that is, for each phase voltage drop time series, the first left window The data points included in the voltage drop time series are the 1st to cth voltage drop data; the first right window The data points included in the voltage drop time series are the c+1 to c+5 voltage drop data; the second left window The data points included in the voltage drop time series are the 2nd to c+1th voltage drop data; the second right window The data points included in are the voltage drop data from c+2 to c+6 in the voltage drop time series; and so on;
[0090] Based on this, the c+1th voltage drop data in the voltage drop time series is the first point that can successfully construct the left window and the right window; the m-c+1th voltage drop data in the voltage drop time series is the last point that can successfully construct the left window and the right window; thus, the left window sequence and the right window sequence are obtained; the number of windows in the left window sequence and the right window sequence is m-2c+1; expressed as:
[0091]
[0092] Among them, W left represents the left window sequence; w L Represents the left window in the left window sequence; W right represents the right window sequence; w R Represents the right window in the right window sequence;
[0093] S2. Calculate the median of the voltage drop data in each left window and right window respectively, instead of the average value; obtain the median time series of the left window and the median time series of the right window; expressed as:
[0094]
[0095] Among them, X left represents the median time series of the left window; x Lrepresents the median of the voltage drop data in each left window, e.g. Represents the median of the voltage drop data in the first left window; X right represents the median time series of the right window; x R represents the median of the voltage drop data in each right window, e.g. Indicates the median of the voltage drop data in the first right window; it should be noted that in this step, the median is selected because the average value result will be disturbed by individual abnormal points, while the median will not, so selecting the median here can improve the subsequent detection accuracy;
[0096] S3. According to the median time series of the left window and the median time series of the right window, the median difference time series ΔX is obtained; it is expressed as:
[0097] ΔX=X right -X left ={Δx c+1 ,Δx c+2 ,···,Δx m-c ,Δx m-c+1}
[0098] Among them, And so on;
[0099] S4. Obtain the absolute value sequence of the median difference time series, and calculate the first quartile and the third quartile in the absolute value sequence; specifically:
[0100] The absolute value sequence |ΔX| is expressed as:
[0101] |ΔX|={|Δx c+1 |,|Δx c+2 |,···,|Δx m-c |,|Δx m-c+1 |}
[0102] Sort the absolute values of the median differences in the absolute value sequence |ΔX| from small to large, and calculate the first quartile Q1 (i.e., 25% quantile) and the third quartile Q3 (i.e., 75% quantile) respectively;
[0103] S5. According to the first quartile and the third quartile, obtain the quantile difference IQR; IQR = Q3-Q1;
[0104] S6. According to the first quartile, the third quartile and the quantile difference, an abnormality judgment condition is constructed to obtain a voltage abnormal value sequence; specifically:
[0105] The abnormal judgment condition is expressed as:
[0106]
[0107] Where |Δx i | represents the absolute value of the i-th median difference in the absolute value sequence; Q1 represents the first quartile; Q3 represents the third quartile; IQR represents the quantile difference; a represents the abnormal parameter.
[0108] According to the abnormal judgment conditions, the voltage abnormal value sequence Q is obtained temp ; expressed as:
[0109]
[0110] According to the voltage abnormal value sequence Q temp , the abnormal heating detection result of the circuit breaker can be obtained; that is, if the voltage abnormal value sequence Q temp If it is empty, there is no abnormal heating; otherwise, there is abnormal heating.
[0111] Furthermore, it also includes:
[0112] S7, each abnormal value in the voltage abnormal value sequence |Δx yi | is actually the absolute value of the median difference, which has a corresponding value Δx in the median difference time series ΔX i ; Take the abnormal value whose corresponding value is greater than the preset value (for example, 0.1) as the final abnormal value, and obtain the final voltage abnormal value sequence Q, which is expressed as:
[0113]
[0114] in, represents the median difference of voltage drops selected from the median difference time series ΔX.
[0115] According to the final voltage abnormal value sequence Q, the abnormal heating detection result of the circuit breaker can be obtained; that is, if the final voltage abnormal value sequence Q is empty, there is no abnormal heating; otherwise, there is abnormal heating.
[0116] After the above anomaly detection model is constructed, the three-phase voltage drop time series is used as input, and the final voltage anomaly value sequence Q is used as output. The constructed anomaly detection model is trained to obtain the final trained anomaly detection model. temp As for the constructed anomaly detection model, the detection accuracy of the anomaly detection model constructed based on the final voltage anomaly value sequence Q is more accurate;
[0117] 5. Obtain abnormal fever detection results:
[0118] After the anomaly detection model training is completed, the three-phase voltage drop time series obtained for the target circuit breaker can be input into the anomaly detection model to detect whether the target circuit breaker has abnormal heating.
[0119] Next, the first embodiment of the present invention is further described in detail by taking a specific example.
[0120] 1. Voltage monitoring data acquisition:
[0121] Select a low-voltage circuit breaker, determine the upper and lower voltage monitoring points of the circuit breaker according to the electrical connection relationship of the circuit breaker, obtain the upper and lower voltage historical data of the last 3 hours from the northbound interface of the data center dynamic environment monitoring system, and form a three-phase voltage time series. Each phase of the upper voltage forms a voltage time series, and each phase of the lower voltage forms a voltage time series. The upper and lower voltages have three phases, A, B, and C, for a total of 6 voltage time series.
[0122] 2. Voltage monitoring data preprocessing:
[0123] (1) Respectively remove the voltage data outside the preset voltage value range in the upper three-phase voltage time series and the lower three-phase voltage time series; for the preset voltage value range, the low-voltage cabinet can be set to 0-1000V, and the 10kV high-voltage cabinet can be set to 0-20kV;
[0124] (2) Resample and align the voltage data in the upper three-phase voltage time series and the lower three-phase voltage time series respectively; in this embodiment, the resampling time interval is set to 5 minutes; when there is no collected value at the corresponding time, the linear interpolation method can be used to fill it;
[0125] 3. Calculation of target circuit breaker voltage drop:
[0126] ΔU a ={Δa1,Δa2,···,Δa m}, ΔU b ={Δb1,Δb2,···,Δb m}, ΔU c ={Δc1,Δc2,···,Δc m}
[0127] The results are shown in the following table:
[0128]
[0129]
[0130] 4. Anomaly detection model construction, including the following construction steps:
[0131] S1. Construct a sliding time window: construct a left window and a right window for each voltage drop data in each phase voltage drop time series;
[0132] Assume that there are m = 30 voltage drop data in each phase voltage drop time series; the number of points contained in the left window or right window is c = 5; that is, for each phase voltage drop time series, the first left window The data points included are the 1st to 5th voltage drop data in the voltage drop time series; the first right window The data points included in the voltage drop time series are the 6th to 10th voltage drop data; the second left window The data points included in the second to sixth voltage drop data in the voltage drop time series; the second right window The data points included in are the 7th to 11th voltage drop data in the voltage drop time series; and so on;
[0133] Based on this, the 6th voltage drop data in the voltage drop time series is the first point that can successfully construct the left window and the right window; the 26th voltage drop data in the voltage drop time series is the last point that can successfully construct the left window and the right window; thus, the left window sequence and the right window sequence are obtained; the number of windows in the left window sequence and the right window sequence is 21; expressed as:
[0134]
[0135] S2. Calculate the median of the voltage drop data in each left window and right window respectively, and obtain the median time series of the left window and the median time series of the right window; expressed as:
[0136]
[0137] S3. According to the median time series of the left window and the median time series of the right window, the median difference time series is obtained, which is expressed as:
[0138] ΔX=X right -X left ={Δx6,Δx7,···,Δx 25 ,Δx 26}
[0139] In the formula And so on, the results are as follows:
[0140] ΔX a={nan,nan,nan,nan,nan,-0.1,0.0,0.0,0.5,0.5,0.6,0.5,0.5,0.0,0.0,0.0,0.1,-0.1 ,0.0,0.1,0.1,-0.1,0.1,0.0,-0.1,-0.1,0.0,0.1,0.1,0.1,0.1,0.1,nan,nan,nan,nan}
[0141] ΔX b ={nan,nan,nan,nan,nan,0.2,0.0,-0.3,-0.3,-0.3,-0.3,0.0,0.1,0.1,0.2,0.1,0.0,0.
[0142] 0,0.0,-0.1,-0.0,0.0,0.0,0.1,0.3,0.3,0.1,0.1,0.0,-0.3,-0.3,-0.1,nan,nan,nan,nan}
[0143] ΔX c ={nan,nan,nan,nan,nan,0.0,0.2,0.0,0.2,0.2,0.1,-0.2,0.1,-0.1,-0.1,-0.2,0.1,-0.2,
[0144] -0.1,-0.1,0.0,-0.1,0.0,0.0,0.0,0.2,0.2,0.7,0.7,0.7,0.5,0.3,nan,nan,nan,nan}
[0145] Where ΔX a represents the median difference time series of phase A; ΔX b represents the median difference time series of phase B; ΔX c Represents the median difference time series of phase C; nan means that the left window or right window cannot be constructed at the current point and no calculated value can be obtained.
[0146] S4. Obtain the absolute value sequence of the median difference time series, expressed as:
[0147] |ΔX|={|Δx6|,|Δx7|,···,|Δx 25 |,|Δx 26 |}
[0148] For the obtained time series, sort them from small to large according to the values, and calculate the first quartile Q1 (i.e. 25% quantile) and the third quartile Q3 (i.e. 75% quantile) respectively;
[0149] S5. According to the first quartile and the third quartile, obtain the quantile difference IQR; IQR = Q3-Q1;
[0150] S6. According to the first quartile, the third quartile and the quantile difference, an abnormality judgment condition is constructed. When a point in the sequence |ΔX| belongs to the abnormal range of the abnormality judgment condition, it is considered to be an abnormal value, and then a voltage abnormal value sequence is obtained; specifically:
[0151] The abnormal judgment condition is expressed as:
[0152]
[0153] In the formula, the value of parameter a is between 3 and 6. In the embodiment of the present invention, a is set to 4.5. Based on this, the voltage abnormal value sequence Q is obtained: temp , recorded as:
[0154]
[0155] For |ΔX a |, Q1 is equal to 0, Q3 is equal to 0.1, IQR is equal to 0.1, the range is [-0.45, 0.55], the value out of the range is 0.6, and the corresponding time value is 17:50:00.
[0156] For |ΔX b |, Q1 is equal to 0, Q3 is equal to 0.3, IQR is equal to 0.3, the range is [-1.35, 1.65], and there is no value out of range.
[0157] For |ΔX c |, Q1 equals 0.1, Q3 equals 0.2, IQR equals 0.1, the range is [-0.35, 0.65], the values out of range are 0.7, 0.7, 0.7, and the corresponding time values are 19:15:00, 19:20:00, 19:25:00.
[0158] S7, according to the voltage abnormal value sequence Q temp Each value in |Δx yi |, confirm whether the corresponding value in step S3 is greater than the preset value (for example, 0.1), if yes, keep it, if not, remove it. It can be found that the corresponding values are all greater than 0.1. Based on this, the final voltage abnormal value sequence Q is obtained, which is recorded as:
[0159]
[0160] 5. Obtain abnormal fever detection results:
[0161] After the training of the anomaly detection model is completed, the three-phase voltage drop time series obtained for the target circuit breaker can be input into the anomaly detection model to detect whether the target circuit breaker has abnormal heating;
[0162] The abnormal heating detection results of the circuit breaker can be found in Figure 3 As shown in the figure, the horizontal axis is in the format of "day-hour-minute", where hour-minute is the time when there is a risk of heating; voltagte_drop_A indicates the voltage drop of phase A; voltagte_drop_C indicates the voltage drop of phase C; and Anomaly indicates abnormality. Whether severe heating will eventually occur can be further judged in combination with the on-site temperature measurement results.
[0163] Embodiment 2:
[0164] The anomaly detection model used in the above-mentioned first embodiment is an anomaly detection model that is constructed and trained from scratch; in the second embodiment of the present invention, the anomaly detection model used can be an existing large language model, that is, the reasoning ability of the large language model is directly used to detect abnormal heating of the circuit breaker, and the three-phase voltage drop time series obtained for the target circuit breaker is input into the large language model to let it determine whether there is an abnormality; the prompt words constructed for the large language model are as follows:
[0165] "Please act as an assistant for detecting abnormal voltage drop.
[0166] Detection method: It is necessary to detect the change in the baseline value of the voltage drop, and this change is on an upward trend. The voltage drop is a time series, and the baseline value change is a type of time series anomaly, also known as Level Shift. We only focus on the increase in Level Shift.
[0167] Other notes 1: If a voltage drop has an abnormality at a single point (e.g., 1 point or 2 points), such as an increase in a single point, it is considered normal, that is, we do not care about the abnormality at the single point (PointAnomaly). For example, a voltage drop time series: 1, 1.2, 1, 1.1, 0.9, 1, 1, 1, 1, 1, 3, 0.9, 0.9, 1; the correct judgment result is no voltage drop abnormality. The output format is required to be "Judgment result: XXXX".
[0168] Based on the above prompts, the user can continue to input "The time series of phase A voltage drop is: 0.9, 1.0, 1.1, 0.9, 0.9, 0.8, 0.8, 0.9, 0.9, 0.8, 1.6, 1.4, 1.4, 1.3, 1.5, 1.4, 1.5, 1.3, 1.3, 1.5, 1.6, 1.2, 1.4, 1.6, 1.5, 1.2, 1.4, 1.4, 1.5, 1.4, 1.5, 1.5, 1.6, 1.6, 1.5, 1.3. Please output the judgment result";
[0169] The large language model outputs: "Judgment result: there is an abnormal voltage drop";
[0170] The user continues to input: "The time series of the voltage drop of phase B is: 0.5, 0.4, 0.1, 0.4, 0.6, 0.6, 0.6, 0.4, 0.6, 0.3, 0.3, 0.2, 0.1, 0.4, 0.5, 0.4, 0.7, 0.4, 0.5, 0.1, 0.4, 0.7, 0.4, 0.1, 0.2, 0.7, 0.7, 0.5, 0.7, 0.5, 0.4, 0.7, 0.2, 0.4, 0.4, 0.7. Please output the judgment result";
[0171] The large language model outputs: "Judgment result: no voltage drop abnormality";
[0172] The user continues to input: "The time series of the voltage drop of phase C is: 1.6, 1.5, 1.7, 1.6, 1.5, 1.4, 1.7, 1.5, 1.6, 1.7, 1.9, 1.4, 1.8, 1.3, 1.7, 1.5, 1.7, 1.6, 1.5, 1.5, 1.7, 1.5, 1.5, 1.6, 1.5, 1.7, 1.0, 1.7, 2.2, 2.3, 2.2, 2.3, 2.0, 2.0, 2.0. Please output the judgment result";
[0173] Large language model: "Judgment result: There is an abnormal voltage drop";
[0174] It can be seen that there are voltage drop anomalies in phases A and C, but there is no voltage drop anomaly in phase B. The judgment result of the large language model is consistent with the judgment result of the anomaly detection model constructed and trained in the above-mentioned embodiment 1.
[0175] Embodiment three:
[0176] On the basis of the above-mentioned embodiment 2, the embodiment 3 further includes:
[0177] Based on the above embodiment 1, the time of each abnormal value in the voltage abnormal value sequence is obtained to form a voltage abnormal time series, specifically including the A phase voltage abnormal time series Abnormal time series of phase B voltage and C phase voltage abnormal time series It is expressed as:
[0178]
[0179] in, Indicates the abnormal time series of phase A voltage The time data in; Indicates the abnormal time series of phase B voltage The time data in; Indicates the abnormal time series of phase C voltage The time data in;
[0180] The voltage anomaly time series obtained based on the constructed anomaly detection model is ANDed or ORed with the large language abnormal heating detection result obtained based on the large language model to obtain a preliminary anomaly detection result; based on this, the interference of a single abnormal point can be suppressed and the robustness of the detection method can be improved; specifically:
[0181] (1) The voltage abnormality time series and the large language abnormal heating detection result are ANDed. That is, when the voltage abnormality time series is not empty and the large language abnormal heating detection result is abnormal, it is determined that the target circuit breaker really has abnormal heating. This method can be applied to the use scenario where the recognition accuracy needs to be guaranteed;
[0182] (2) The voltage abnormality time series and the large language abnormal heating detection result are ORed, that is, when the voltage abnormality time series is not empty, or the large language abnormal heating detection result is abnormal, it is determined that the target circuit breaker really has abnormal heating. This method can be applied in scenarios where the recognition recall rate needs to be guaranteed;
[0183] For example, the abnormal heating detection result output by the abnormal detection model is that the voltage drop of phase A is abnormal, the voltage drop of phase B is not abnormal, and the voltage drop of phase C is abnormal; and the voltage abnormality time series obtained is: Then, no matter whether an AND operation or an OR operation is performed with the abnormal heating detection result of the large language in the above-mentioned embodiment 2, the result that the target circuit breaker has abnormal heating can be obtained.
[0184] Embodiment 4:
[0185] On the basis of the above-mentioned embodiment 3, the embodiment 4 further includes:
[0186] Obtaining an upper three-phase load current time series corresponding to an upper voltage monitoring point and a lower three-phase load current time series corresponding to a lower voltage monitoring point within a preset time period;
[0187] Preprocessing the upper three-phase load current time series and the lower three-phase load current time series; removing the current data outside the preset current value range (0 to infinity) in the upper three-phase load current time series and the lower three-phase load current time series; and resampling and data alignment processing are performed on the current data in the upper three-phase load current time series and the lower three-phase load current time series (refer to the preprocessing part of the voltage data in the first embodiment);
[0188] Based on the preprocessed upper three-phase load current time series and the lower three-phase load current time series, a three-phase load current difference time series of the target circuit breaker is obtained; the three-phase load current difference time series is input into the anomaly detection model to obtain a current anomaly value sequence;
[0189] It should be noted here that in the fourth embodiment of the present invention, the anomaly detection model has the function of obtaining a current anomaly value sequence according to the three-phase load current difference time series; the specific training process can refer to the relevant content of obtaining a voltage anomaly value sequence according to the three-phase voltage drop time series in the above-mentioned first embodiment;
[0190] Obtain the time of each current abnormal value in the current abnormal value sequence to form a current abnormal time series, including: A phase current abnormal time series Abnormal time series of phase B current and C phase current abnormal time series It is expressed as:
[0191]
[0192] in, Indicates the abnormal time series of phase A current The time data in; Indicates the abnormal time series of phase B current The time data in; Indicates the abnormal time series of phase C current The time data in;
[0193] The same time value in the voltage abnormal time series and the current abnormal time series is taken as the pseudo abnormal time value; the pseudo abnormal time value in the voltage abnormal time series is removed to form the final voltage abnormal time series; including the final voltage abnormal time series of phase A Phase B final voltage abnormal time series and the final voltage abnormal time series of phase C If the final voltage abnormal time series and If both are empty, there is no heating risk; otherwise, the circuit breaker has heating risk. For the circuit breaker with heating risk, the specific time when the heating risk occurs is further given.
[0194] In this embodiment, errors caused by increased current can be avoided, thereby improving the accuracy of the recognition result.
[0195] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0196] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop, characterized in that: The steps include: Determine the upper voltage monitoring point and the lower voltage monitoring point of the target circuit breaker; Obtaining, within a preset time period, an upper three-phase voltage time series corresponding to the upper voltage monitoring point and a lower three-phase voltage time series corresponding to the lower voltage monitoring point; Preprocessing the upper three-phase voltage time series and the lower three-phase voltage time series; Based on the preprocessed upper three-phase voltage time series and lower three-phase voltage time series, obtaining the three-phase voltage drop time series of the target circuit breaker; The three-phase voltage drop time series is input into an abnormality detection model to obtain an abnormal heating detection result of the target circuit breaker.
2. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 1, characterized in that: The upper three-phase voltage time series includes an upper A-phase voltage time series, an upper B-phase voltage time series and an upper C-phase voltage time series; The lower end three-phase voltage time series includes a lower end A-phase voltage time series, a lower end B-phase voltage time series and a lower end C-phase voltage time series.
3. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 1, characterized in that: The preprocessing of the upper three-phase voltage time series and the lower three-phase voltage time series specifically includes: Respectively removing voltage data outside a preset voltage value range in the upper three-phase voltage time series and the lower three-phase voltage time series; Resampling and data alignment processing are performed on the voltage data in the upper three-phase voltage time series and the lower three-phase voltage time series respectively.
4. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 1, characterized in that: The number of voltage data in the preprocessed upper three-phase voltage time series and the lower three-phase voltage time series is consistent.
5. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 4, characterized in that: The three-phase voltage drop time series includes a phase A voltage drop time series, a phase B voltage drop time series and a phase C voltage drop time series, which are expressed as: Among them, ΔU a Represents the time series of the voltage drop of phase A; ΔU b represents the time series of the voltage drop of phase B; ΔU c represents the time series of the voltage drop of phase C; It represents the time series of the voltage of phase A at the upper end; represents the time series of the voltage of phase B at the upper end; represents the time series of the upper C phase voltage; It represents the time series of the voltage of phase A at the lower end; represents the time series of the voltage of phase B at the lower end; represents the time series of the voltage of phase C at the lower end; Δa m Indicates ΔU a The mth voltage drop data in; Δb m Indicates ΔU b The mth voltage drop data in; Δc m Indicates ΔU c The mth voltage drop data in .
6. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 1, characterized in that: The steps of constructing the anomaly detection model include: S1. Construct a sliding time window: construct a left window and a right window for each voltage drop data in each phase voltage drop time series; S2, respectively calculating the median of the voltage drop data in each of the left window and the right window; obtaining the median time series of the left window and the median time series of the right window; S3, obtaining a median difference time series according to the left window median time series and the right window median time series; S4, obtaining an absolute value sequence of the median difference time series, and calculating the first quartile and the third quartile in the absolute value sequence; S5. Obtain a quantile difference according to the first quartile and the third quartile; S6. Construct an abnormality judgment condition according to the first quartile, the third quartile and the quartile difference, and then obtain a voltage abnormality value sequence; if the voltage abnormality value sequence is empty, there is no abnormal heating; otherwise, there is an abnormal heating.
7. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 6, characterized in that: The abnormal judgment condition is expressed as: Where |Δx i | represents the absolute value of the i-th median difference in the absolute value sequence; Q1 represents the first quartile; Q3 represents the third quartile; IQR represents the quantile difference; a represents the abnormal parameter.
8. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 7, characterized in that: The step of constructing the anomaly detection model further includes: S7. Each abnormal value in the voltage abnormal value sequence has a corresponding value in the median difference time series; the abnormal value whose corresponding value is greater than the preset value is taken as the final abnormal value to obtain the final voltage abnormal value sequence; if the final voltage abnormal value sequence is empty, there is no abnormal heating; otherwise, there is abnormal heating.
9. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 6, characterized in that: Also includes: Obtaining the time of each abnormal value in the voltage abnormal value sequence to form a voltage abnormal time series; Inputting the three-phase voltage drop time series into a large language model to obtain a large language abnormal heating detection result of the target circuit breaker; The voltage abnormality time series is ANDed or ORed with the abnormal fever detection result of the loudspeaker to obtain a preliminary abnormality detection result.
10. The method for automatically detecting abnormal heating of a circuit breaker in a data center based on voltage drop according to claim 9, characterized in that: Also includes: Obtaining, within a preset time period, an upper three-phase load current time series corresponding to the upper voltage monitoring point and a lower three-phase load current time series corresponding to the lower voltage monitoring point; Preprocessing the upper three-phase load current time series and the lower three-phase load current time series; Based on the preprocessed upper three-phase load current time series and lower three-phase load current time series, obtaining a three-phase load current difference time series of the target circuit breaker; Inputting the three-phase load current difference time series into the anomaly detection model to obtain a current anomaly value sequence; Obtaining the time of each abnormal value in the current abnormal value sequence to form a current abnormal time series; taking the same time value in the voltage abnormal time series and the current abnormal time series as a pseudo abnormal value; removing pseudo abnormal values in the voltage abnormal time series to form a final voltage abnormal time series; If the final voltage abnormal time series is empty, it is determined that the target circuit breaker does not have heating phenomenon.
