A public power utilization equipment detection system for intelligent buildings

Abstract

The application provides a kind of for intelligent building public power utilization equipment detection system, it is related to power utilization equipment detection technical field, including the parameter acquisition module for obtaining the operating parameter to be analyzed and the operating parameter standard value set under the same working condition;Interval division module for interval division;Fluctuation analysis module for quantifying fluctuation tolerance through fluctuation factor;Deviation feature group extraction module for constructing operating feature vector by combining fluctuation factor, penalty gain, concentration deviation, reasonable interval probability difference and distribution entropy gain;Device fault detection module for analyzing operating feature vector to output operating state, the application introduces penalty gain, concentration deviation, reasonable interval probability difference, distribution entropy gain and fluctuation factor to judge operating state on the basis of considering the factor of current working condition, significantly improves the accuracy of detecting and judging the operating state of public power utilization equipment.

Description

A detection system for public electrical equipment in intelligent buildings Technical Field

[0001] This invention relates to the field of electrical equipment testing technology, specifically to a testing system for public electrical equipment in intelligent buildings. Background Technology

[0002] In recent years, with the advent of the big data era and the development of smart grids, condition monitoring technology for public electrical equipment has gradually become a research hotspot in the power industry. By monitoring the condition of public electrical equipment, potential problems can be identified in a timely manner, facilitating timely maintenance by staff. Currently, condition monitoring technology for public electrical equipment relies on sensor technology and data analysis technology. First, various sensors are used to obtain data on public electrical equipment under different operating conditions. Then, data analysis technology is used to analyze the data, thereby achieving condition monitoring and fault diagnosis of public electrical equipment.

[0003] For example, in the prior art, a fault detection and processing method for electrical equipment based on key feature extraction (main classification number G06F) with the publication number "CN116956186A" includes: (1): collecting the working information of the electrical equipment; (2): making a preliminary judgment based on the working information of the electrical equipment. If there is an abnormality, an alarm is issued; if there is no abnormality, proceed to (3); (3): performing wavelet transform on the leakage current signal; (4): extracting the key features of the leakage current signal; (5): judging whether there is an abnormality based on the key features of the leakage current signal. If so, an alarm is issued and proceed to (6); (6): performing abnormality timing; (7): judging whether the abnormal timing is greater than the timing reference value. If so, controlling the electrical equipment to stop working; if not, returning to (6). This prior art can perform multi-faceted fault judgment and processing of electrical equipment, ensuring the safety and reliability of the electrical equipment.

[0004] However, existing technologies still have significant drawbacks. For example, when making preliminary judgments, these technologies use a reference threshold to determine whether electrical equipment is faulty, without considering the operating conditions of the equipment. The reference threshold can vary depending on the operating conditions, and the existing technologies using a fixed reference threshold suffer from poor accuracy. Furthermore, for faulty equipment, its operating parameters do not always deviate from the normal range. For instance, when a fault occurs due to poor contact, the current value will fluctuate drastically, but some current values ​​will still remain within the normal range. Therefore, if these existing technologies are applied to single-moment judgment scenarios, they may fail to identify faults even when the detected values ​​are within the normal range. Similarly, if applied to time-period judgment scenarios, they may fail to identify faults even when the average detected values ​​for that time period are within the normal range. Thus, the existing technologies have poor fault diagnosis accuracy in practical applications.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a detection system for public electrical equipment in intelligent buildings, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A detection system for public electrical equipment in intelligent buildings, comprising:

[0009] The parameter acquisition module is used to collect the operating condition identification parameters and operating parameters of various public electrical equipment in intelligent buildings during the latest detection time period when they are in stable operating conditions. Based on the operating condition identification parameters, it retrieves the operating parameters of the public electrical equipment during multiple similar detection time periods under normal operation to construct a set of standard values ​​for operating parameters.

[0010] The interval division module determines the distribution interval and standard probability value of each operating parameter based on the set of standard values ​​of operating parameters, and determines the penalty coefficient of each distribution interval based on the standard probability value. The distribution interval includes the reasonable interval and the penalty interval, and whether the doubtful interval is empty is determined by the distribution status of the elements in the set of standard values ​​of operating parameters.

[0011] The volatility analysis module statistically analyzes the current probability value and reference probability value of the same operating parameter in each distribution interval under the latest detection time period and each similar detection time period. Then, it determines the current distribution entropy value and the entropy value of each reference distribution. Combining the current distribution entropy value and the standard distribution entropy value determined based on the standard probability value, it determines the neighborhood search radius for recursively expanding clustering of the reference distribution entropy value. Based on the clustering results, it determines the volatility factor.

[0012] The deviation feature extraction module combines the distribution interval, current probability value, standard probability value and penalty coefficient to determine the penalty gain, concentration deviation, reasonable interval probability difference and distribution entropy value gain of each operating parameter in the latest detection period, and constructs the operating feature vector of each operating parameter by combining the volatility factor and statistical feature difference.

[0013] The equipment fault detection module is used to establish a mapping relationship between the operating feature vectors of each public electrical equipment and its operating status, and to process the operating feature vectors of each public electrical equipment in the latest detection time period based on the mapping relationship, so as to output the operating status of each public electrical equipment.

[0014] Furthermore, the definition of stable operation of public electrical equipment is as follows: if the status-type operating condition identification parameters of the public electrical equipment remain unchanged during a detection period, and the coefficient of variation of the numerical operating condition identification parameters is less than a preset threshold, then the operating state of the public electrical equipment during this detection period is defined as stable operation.

[0015] The operating condition identification parameters are used to identify the operating conditions of public electrical equipment during the latest detection period, including but not limited to environmental parameters that affect the operating performance of public electrical equipment, as well as setting parameters that control the operation of public electrical equipment to achieve the desired conditions.

[0016] The operating parameters are used to reflect the operating status of public electrical equipment during the latest detection period, including but not limited to one or more of the voltage value, current value, and power factor of the public electrical equipment during the latest detection period.

[0017] Furthermore, the logic for obtaining the set of standard operating parameters is as follows: based on the operating condition identification parameters of the public electrical equipment in the latest detection time period, the operating condition of the public electrical equipment in the latest detection time period is determined. Then, from the previous normal detection time periods of the public electrical equipment, similar detection time periods with similar operating conditions are selected. The operating parameters of the public electrical equipment in the similar detection time periods are used as the standard operating parameters. The same set of standard operating parameters is then summarized to form a set of standard operating parameters.

[0018] The logic for obtaining similar detection time periods is as follows: For any public electrical equipment, if there exists a normal detection time period that satisfies the following conditions: the relative error between the duration of this normal detection time period and the duration of the latest detection time period is less than a preset threshold, the status-type operating condition identification parameters of the public electrical equipment within this normal detection time period are consistent with the status-type operating condition identification parameters within the latest detection time period, and the relative error between the average value of the numerical operating condition identification parameters and the average value of the same numerical operating condition identification parameters within the latest detection time period is less than a preset threshold, then it is determined that the operating condition of the public electrical equipment within this normal detection time period is similar to that within the latest detection time period, and this normal detection time period is designated as the similar detection time period.

[0019] Furthermore, the standard probability value for each distribution interval is obtained as follows: for any distribution interval, calculate the ratio of the number of elements in the set of standard values ​​of operating parameters that fall into that distribution interval to the total number of elements in the set of standard values ​​of operating parameters, and use this ratio as the standard probability value for that distribution interval.

[0020] The logic for determining the distribution interval is as follows:

[0021] 1) For any operating parameter, sort the elements in the corresponding standard value set of operating parameters in ascending order of value to form a sequence of standard values ​​of operating parameters. Determine the window length based on the number of elements in the standard value set of operating parameters and the preset reasonableness ratio. Slide the window sequentially in the sequence of standard values ​​of operating parameters and calculate the difference between the maximum and minimum values ​​in each window as the interval width. Select the window corresponding to the minimum interval width as the optimal window. Use the maximum and minimum values ​​in the optimal window as the upper and lower limits of the reasonable interval, respectively.

[0022] The mathematical expression for the window length is as follows:

[0023]

[0024] In the formula, For window length, This indicates an upward fetch operation. The number of elements in the set of standard values ​​for the running parameters. The percentage of values ​​greater than 0 that are considered reasonable;

[0025] 2) With the standard probability value of the doubtful interval not being lower than a preset threshold as a constraint, expand to both sides of the reasonable interval to form a doubtful interval with the same width as the interval.

[0026] 3) Extract the minimum lower limit and maximum upper limit from the reasonable interval and all questionable intervals, and use them as the upper limit of the lowest penalty interval and the lower limit of the highest penalty interval, respectively. The lower limit of the lowest penalty interval is negative infinity, and the upper limit of the highest penalty interval is positive infinity.

[0027] Furthermore, the logic for determining the penalty coefficient for each distribution interval is as follows:

[0028] 1) Based on the standard probability value, assign a corresponding penalty value to each distribution interval, as follows:

[0029] 1.1) Assign a penalty value of 0 to a reasonable range;

[0030] 1.2) If the doubtful interval is not empty, for any doubtful interval, calculate the ratio of the reasonable interval to its standard probability value, and use it as the penalty value for the doubtful interval. If the doubtful interval is empty, no penalty value is calculated.

[0031] 1.3) Based on the penalty value of the doubtful interval, determine the penalty value of the penalty interval, and its mathematical expression is as follows:

[0032]

[0033] In the formula, This represents the penalty value within the penalty interval. This represents the sum of the penalty values ​​for all questionable intervals. This represents a magnification factor with a value greater than 1. Indicates the range of uncertainty. This indicates that the interval in question is not empty. This indicates that the questionable interval is empty;

[0034] 2) Using the cumulative value equal to 1 as the scaling target, the penalty value of each distribution interval is scaled proportionally, and the scaled penalty value is used as the penalty coefficient of each corresponding distribution interval.

[0035] Furthermore, the mathematical expression for the neighborhood search radius is as follows:

[0036]

[0037] In the formula, Indicates the neighborhood search radius. This represents the current distribution entropy value. This represents the entropy value of the standard distribution;

[0038] The logic for recursively expanding clustering based on the reference distribution entropy values ​​corresponding to each similar detection time period is as follows:

[0039] The reference distribution entropy values ​​corresponding to each similar detection time period are summarized to form a reference distribution entropy value set, and the initial state of each reference distribution entropy value in the reference distribution entropy value set is marked as unvisited.

[0040] Randomly select an unvisited reference distribution entropy value from the set of reference distribution entropy values. , This is the index of the reference distribution entropy value in the set of reference distribution entropy values, and m represents the number of elements in the set of reference distribution entropy values;

[0041] Based on neighborhood search radius Traverse the set of reference distribution entropy values ​​to determine the reference distribution entropy value. neighborhood Specifically, if there exists a reference distribution entropy value in the set of reference distribution entropy values ​​that is in an unvisited state, and this reference distribution entropy value is related to the reference distribution entropy value... If the absolute difference between the values ​​is less than the neighborhood search radius, then this reference distribution entropy value is included in the reference distribution entropy value. neighborhood middle;

[0042] If the reference distribution entropy value neighborhood If the number of internal elements is less than the number of minimum neighboring nodes, then a new unvisited reference distribution entropy value is selected from the set of reference distribution entropy values; otherwise, the reference distribution entropy value is... As the core point and to create clusters Clustering From the reference distribution entropy value and neighboring areas Composition, and based on clustering The status of the reference distribution entropy values ​​within the reference distribution entropy value set is updated in real time, specifically by including the reference distribution entropy value set into clusters. The reference distribution entropy value status has been changed to "visited".

[0043] For the neighborhood The reference distribution entropy value within a given area is used to determine its neighborhood. If the number of reference distribution entropy values ​​in that neighborhood is less than the minimum number of neighborhood nodes, then that neighborhood is not included in the cluster. Otherwise, include the neighborhood in the cluster. It also updates the status of the reference distribution entropy value in real time and continues to recursively expand until clustering occurs. Unable to expand further;

[0044] Randomly select another unvisited reference distribution entropy value from the set of reference distribution entropy values ​​again, and construct a new cluster in the same way until no cluster can be constructed. For reference distribution entropy values ​​that are not assigned to any cluster, mark them as outliers.

[0045] Furthermore, the clustering results include the number of clusters and the number of outliers. Based on the number of clusters and the proportion of outliers, a volatility factor is determined, the mathematical expression of which is as follows:

[0046]

[0047] In the formula, Indicates the number of cluster categories. Indicates the proportion of isolated points. As a volatility factor, and A growth rate adjustment factor greater than zero. The threshold for the proportion of isolated points. , and All of these are preset proportional coefficients, and their specific values ​​are determined based on the analytic hierarchy process (AHP).

[0048] Furthermore, the mathematical expression for the penalty gain is as follows:

[0049]

[0050] In the formula, The penalty gain for the running parameters, This represents the penalty coefficient for the k-th distribution interval. This represents the current probability value of the running parameters in the k-th distribution interval. Let k be the standard probability value of the same operating parameter in the k-th distribution interval, where k is the index of the distribution interval. K is the number of distribution intervals;

[0051] The concentration bias is a vector comprising the median bias and the width bias of the interval. The logic for obtaining the median bias and the width bias is as follows:

[0052] 1) For any operating parameter, sort all elements in the latest detection time period in ascending order of value to form the current value sequence of the operating parameter. Determine the window length based on the number of elements collected and the reasonableness ratio of the operating parameter in the latest detection time period. Slide the window sequentially in the current value sequence of the operating parameter and calculate the difference between the maximum and minimum values ​​in each window as the interval width. Select the window corresponding to the minimum interval width as the concentration window. Use the maximum and minimum values ​​in the concentration window as the upper and lower limits of the concentration distribution interval, respectively, to obtain the concentration distribution interval of the operating parameter in the latest detection time period.

[0053] 2) Determine the median value of the concentrated distribution interval and the median value of the reasonable interval, and take the absolute difference between the two as the median deviation. The median value is the average of the upper and lower limits of the interval.

[0054] 3) Determine the width of the concentrated distribution interval and the width of the reasonable interval, and take the difference between the two as the interval width deviation. The interval width is the difference between the upper and lower limits of the interval.

[0055] The probability difference within the reasonable interval is the difference between the current probability value and the standard probability value of the operating parameter within the reasonable interval.

[0056] The distribution entropy gain is the difference between the current distribution entropy value and the standard distribution entropy value;

[0057] The statistical feature differences include mean differences, amplitude differences, and variance differences, which are the differences between the mean, amplitude, and variance of the operating parameters in the latest detection time period and the mean, amplitude, and variance of the elements in the corresponding standard value set of operating parameters.

[0058] Furthermore, for any public electrical equipment, a deep neural network model is used to establish the mapping relationship between its operating feature vector and operating status;

[0059] The deep neural network model uses the TensorFlow deep learning framework, which includes an input layer for receiving the operating feature vectors of various operating parameters of the public electrical equipment, one or more hidden layers for processing the input operating feature vectors, with ReLU selected as the activation function in the hidden layers, and an output layer for outputting the operating status of the public electrical equipment, which is either normal or abnormal.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] The present invention provides a detection system for public electrical equipment in intelligent buildings. When judging the current operating status of public electrical equipment, it uses the operating parameters of the equipment in multiple similar detection time periods under normal operation as a reference benchmark, fully considering the current operating condition. Furthermore, when judging the operating status based on statistical feature differences, it introduces penalty gain, concentration bias, reasonable interval probability difference, and distribution entropy gain to characterize the distribution of operating parameters within a time period. It also introduces a volatility factor to characterize the tolerance for the degree of fluctuation of operating parameters, significantly improving the accuracy of detecting and judging the operating status of public electrical equipment. Attached Figure Description

[0062] Figure 1 is a module unit diagram of the detection system in this invention;

[0063] Figure 2 shows a comparison of the prediction accuracy of the proposed model and the reference model. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0065] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0066] Example 1:

[0067] Please refer to Figures 1-2. This invention provides a detection system for public electrical equipment in intelligent buildings, comprising:

[0068] The parameter acquisition module is used to collect the operating condition identification parameters and operating parameters of various public electrical equipment in intelligent buildings during the latest detection time period when they are in stable operating conditions. Based on the operating condition identification parameters, it retrieves the operating parameters of the public electrical equipment during multiple similar detection time periods under normal operation to construct a set of standard values ​​for operating parameters.

[0069] Among them, the operating condition identification parameters are used to identify the operating conditions of public electrical equipment in the latest detection period, including but not limited to environmental parameters that affect the operating performance of public electrical equipment, as well as setting parameters that control the operation of public electrical equipment to achieve the desired conditions. The specific operating condition identification parameters are customized and set according to different types of public electrical equipment, which will not be elaborated here.

[0070] The definition of stable operation of public electrical equipment is as follows: if the status-type operating condition identification parameters of the public electrical equipment remain unchanged during a detection period, and the coefficient of variation of the numerical operating condition identification parameters is less than a preset threshold, then the operating state of the public electrical equipment during this detection period is defined as stable operation.

[0071] It should be noted that the coefficient of variation is the ratio of the standard deviation to the mean, which is common knowledge to those skilled in the art and will not be elaborated here. The preset threshold corresponding to the coefficient of variation can be set to the usual 20% to ensure that the numerical working condition identification parameters are in the detection period with low variability.

[0072] It should be noted that state-based operating condition identification parameters are limited to a limited number of states, such as the air conditioning operation mode, which only includes a limited number of modes such as cooling and heating. On the other hand, numerical operating condition identification parameters can take any value, such as outdoor temperature, indoor temperature, indoor light intensity, etc.

[0073] It should be noted that the latest detection time period is the time period closest to the current moment when the public electrical equipment is under stable operating conditions. The operating conditions of the public electrical equipment are close to stable during the latest detection time period, which facilitates subsequent targeted comparison and analysis of the standard values ​​of operating parameters under similar operating conditions in the past to determine whether there are any abnormalities in the public electrical equipment during the latest detection time period. Compared with analyzing the operating parameters of public electrical equipment under variable operating conditions to determine whether there are abnormalities, this is more in line with the normal operating environment of public electrical equipment in intelligent buildings, and also greatly reduces the factors to be considered, thus simplifying the calculation difficulty. In addition, the latest detection time periods of each public electrical equipment may be different. In order to fully extract the operating parameter characteristics of public electrical equipment, a minimum duration constraint can be set for the latest detection time period, such as setting the latest detection time period to be no less than 5 minutes, etc. The sampling frequency of the operating condition identification parameters and operating parameters can be set to once per second, once every 5 seconds, once every 10 seconds, etc., which can be set by the staff according to the actual situation, and no restrictions are imposed here.

[0074] Among them, the operating parameters are used to reflect the operating status of public electrical equipment in the latest detection period, including but not limited to one or more of the voltage value, current value, and power factor of the public electrical equipment in the latest detection period. The specific operating parameters are customized and set according to different types of public electrical equipment, which will not be elaborated here.

[0075] As one implementation method, public electrical equipment in intelligent buildings includes, but is not limited to, air conditioning equipment, lighting equipment, etc. The specific selection of operating condition identification parameters and operating parameters for air conditioning equipment and lighting equipment is as follows:

[0076] The operating condition identification parameters of the air conditioning equipment may include the outdoor temperature, indoor temperature, indoor temperature change rate, air conditioning set temperature and air conditioning operation mode at each sampling time within the latest detection period. Among them, the outdoor temperature, indoor temperature, indoor temperature change rate and air conditioning set temperature are numerical operating condition identification parameters, while the air conditioning operation mode is a status-based operating condition identification parameter. Its operation parameters may include the voltage, current, air outlet temperature and power factor of the air conditioning equipment at each sampling time within the latest detection period.

[0077] It should be noted that the calculation logic for the rate of change of indoor temperature at any sampling time is as follows: calculate the difference between the indoor temperature at the current sampling time and the previous sampling time, and divide the difference by the time interval between the current sampling time and the previous sampling time to obtain the rate of change of indoor temperature at the current sampling time. The outdoor temperature reflects the influence of the outside temperature on the operation of the air conditioning equipment, the indoor temperature reflects the working effect of the air conditioning equipment, the air conditioning set temperature reflects the target working effect of the air conditioning equipment, the rate of change of indoor temperature reflects the working capacity of the air conditioning equipment, and the air conditioning operation mode reflects the working mode of the air conditioning equipment. Therefore, the above five parameters are used as the operating condition identification parameters of the air conditioning equipment to characterize the operating condition of the air conditioning equipment in the latest detection time period.

[0078] The operating condition identification parameters of lighting equipment may include outdoor light intensity, indoor light intensity and set brightness at each sampling time within the latest detection period. All three are numerical operating condition identification parameters. The operating parameters may include the lamp temperature, voltage, current and power factor of the lighting equipment at each sampling time within the latest detection period.

[0079] It should be noted that outdoor light intensity reflects the impact of external light and darkness on the operation of lighting equipment, indoor light intensity reflects the working effect of lighting equipment, and set brightness reflects the target working effect of lighting equipment. Therefore, the above three parameters are used as the operating condition identification parameters of lighting equipment to characterize the operating condition of lighting equipment in the latest detection period.

[0080] The logic for obtaining the standard value set of operating parameters is as follows: Based on the operating condition identification parameters of the public electrical equipment in the latest detection period, the operating condition of the public electrical equipment in the latest detection period is determined. Then, from the previous normal detection period of the public electrical equipment, similar detection period periods with similar operating conditions are selected. The operating parameters of the public electrical equipment in the similar detection period periods are used as the standard values ​​of operating parameters. The same type of standard values ​​of operating parameters are summarized to form the standard value set of operating parameters.

[0081] It should be noted that the normal testing period is the testing period when the public electrical equipment is in normal operating condition. Whether the public electrical equipment is in normal operating condition can be determined by relevant experts or by checking the maintenance records of the public electrical equipment. The testing period should be selected from the time of maintenance (e.g., at least one day away from the maintenance time) as the testing period when the public electrical equipment is in normal operating condition. This is existing technology and will not be elaborated here.

[0082] The logic for obtaining similar detection time periods is as follows: For any public electrical equipment, if there exists a normal detection time period that satisfies the following conditions: the relative error between the duration of the normal detection time period and the duration of the latest detection time period is less than a preset threshold, the status-type operating condition identification parameters of the public electrical equipment are consistent with the status-type operating condition identification parameters in the latest detection time period, and the relative error between the average value of the numerical operating condition identification parameters and the average value of the same numerical operating condition identification parameters in the latest detection time period is less than a preset threshold, then it is determined that the operating condition of the public electrical equipment in this normal detection time period is similar to that in the latest detection time period, and this normal detection time period is taken as the similar detection time period.

[0083] As one implementation method, the preset threshold for duration can be set between 10% and 20% to ensure the similarity between the duration of similar detection time periods and the duration of the latest detection time period. For numerical condition identification parameters, the preset threshold can be set between 3% and 8% to ensure that a large number of similar detection time periods are selected, providing a basis for retrieving a large number of standard values ​​of operating parameters. The consistency of state-related condition identification parameters ensures the rationality of the selection of similar detection time periods.

[0084] The interval division module determines the distribution interval and standard probability value of each operating parameter based on the set of standard values ​​of operating parameters, and determines the penalty coefficient of each distribution interval based on the standard probability value. The distribution interval includes the reasonable interval and the penalty interval, and whether the doubtful interval is empty is determined by the distribution status of the elements in the set of standard values ​​of operating parameters.

[0085] The method for obtaining the standard probability value of each distribution interval is as follows: For any distribution interval, calculate the ratio of the number of elements in the set of standard values ​​of operating parameters that fall into the distribution interval to the total number of elements in the set of standard values ​​of operating parameters, and use it as the standard probability value of the distribution interval. The larger the value, the more elements in the set of standard values ​​of operating parameters fall into the corresponding distribution interval.

[0086] The logic for determining the distribution interval is as follows:

[0087] 1) For any operating parameter, sort the elements in the corresponding standard value set of operating parameters in ascending order of value to form a sequence of standard values ​​of operating parameters. Determine the window length based on the number of elements in the standard value set of operating parameters and the preset reasonableness ratio. Slide the window sequentially in the sequence of standard values ​​of operating parameters and calculate the difference between the maximum and minimum values ​​in each window as the interval width. Select the window corresponding to the minimum interval width as the optimal window. Use the maximum and minimum values ​​in the optimal window as the upper and lower limits of the reasonable interval, respectively.

[0088] The mathematical expression for the window length is as follows:

[0089]

[0090] In the formula, For window length, This indicates an upward fetch operation. The number of elements in the set of standard values ​​for the running parameters. The percentage of reasonableness values ​​greater than 0 is a decimal not less than the percentage threshold, which is a decimal not less than 60%. The specific value is set by the staff according to the actual situation. The larger the percentage of reasonableness value, the more elements in the set of standard values ​​of operating parameters are included in the reasonable range, and the wider the reasonable range is.

[0091] It should be noted that while the operating parameters of public electrical equipment are concentrated in a certain range during normal operation, they also fluctuate within a certain range rather than remaining constant. Therefore, fluctuations in operating parameters within a certain range are normal. In addition, due to the influence of the external environment and the equipment itself, even during normal operation, the operating parameters of public electrical equipment may experience occasional sudden changes that deviate from the normal fluctuation range. Therefore, if the value of the reasonableness ratio is too high, it will excessively increase the reasonable range, leading to the inclusion of mutated element values ​​within the reasonable range. Conversely, if the value of the reasonableness ratio is too low, it will excessively reduce the reasonable range, leading to the exclusion of normally fluctuating element values ​​from the reasonable range. Therefore, it is necessary to set a suitable reasonableness ratio to determine the reasonable range. The optimal value range for the reasonableness ratio is generally between 60% and 80%.

[0092] 2) With the standard probability value of the doubtful interval not being lower than a preset threshold as a constraint, expand to both sides of the reasonable interval to form multiple doubtful intervals with the same width as the interval;

[0093] It should be noted that due to the subjectivity of the reasonableness ratio selection, the reasonable range screening can only filter out most of the normal data. It does not mean that the data outside the reasonable range is necessarily abnormal data. That is, the data outside the reasonable range is considered to be normal data with doubt. Therefore, it is necessary to set a doubtful range to comprehensively describe this part of the data. Since the vast majority of the standard values ​​of the running parameters are normal data and the proportion of abnormal data is extremely low, the value range of the preset threshold in the above constraints can generally be set between 5% and 10% to ensure that an interval with a large amount of data distribution outside the reasonable range is found as the doubtful range. The data in the doubtful range may be normal data or abnormal data.

[0094] It should be noted that if the distribution of elements in the set of standard values ​​of operating parameters is highly concentrated, then after determining that it is reasonable, there is a possibility that it is impossible to find a questionable interval that meets the above constraints. In this case, the distribution interval only includes the reasonable interval and the penalty interval, and there is no questionable interval, that is, the questionable interval is empty.

[0095] 3) Extract the minimum lower limit and maximum upper limit from the reasonable interval and all questionable intervals, and use them as the upper limit of the lowest penalty interval and the lower limit of the highest penalty interval, respectively. The lower limit of the lowest penalty interval is negative infinity, and the upper limit of the highest penalty interval is positive infinity.

[0096] As can be seen from the above, the reasonable range and the questionable range encompass the vast majority of elements in the set of standard values ​​of operating parameters. That is, elements in the set of standard values ​​of operating parameters that do not fall into the reasonable range and the questionable range can be considered as abnormal data by default. Therefore, the minimum penalty range and the maximum penalty range are set to comprehensively describe abnormal data. In addition, when the questionable range is empty, the lower limit and the upper limit of the reasonable range are the minimum lower limit and the maximum upper limit. Empty does not mean zero.

[0097] The logic for determining the penalty coefficient for each distribution interval is as follows:

[0098] 1) Based on the standard probability value, assign a corresponding penalty value to each distribution interval, as follows:

[0099] 1.1) Assign a penalty value of 0 to a reasonable range;

[0100] As can be seen from the above, element values ​​falling into the reasonable range are considered normal data by default. Therefore, the penalty value for the reasonable range is set to 0 to indicate that there is no penalty for data falling into the reasonable range.

[0101] 1.2) If the doubtful interval is not empty, for any doubtful interval, calculate the ratio of the reasonable interval to its standard probability value, and use it as the penalty value for the doubtful interval. If the doubtful interval is empty, no penalty value is calculated.

[0102] It should be noted that for the questionable interval, it is questionable whether the element values ​​inside it are normal data, and the element values ​​inside it deviate from the central tendency of the elements in the standard value set of the operating parameters. That is, even if the element values ​​inside the questionable interval are not abnormal data, they are inferior values ​​that deviate from the normal state. Therefore, it is necessary to assign a penalty value greater than 0 to it. The smaller the standard probability value of the questionable interval, the greater the possibility that the data falling into the questionable interval is abnormal data or inferior value compared with the reasonable interval. In order to measure the penalty value of each questionable interval on the same scale, the ratio of the standard probability value of the reasonable interval to each questionable interval is used as the penalty value of each questionable interval.

[0103] 1.3) Based on the penalty value of the doubtful interval, determine the penalty value of the penalty interval, and its mathematical expression is as follows:

[0104]

[0105] In the formula, This represents the penalty value within the penalty interval. This represents the sum of the penalty values ​​for all questionable intervals. This represents a magnification factor with a value greater than 1. Indicates the range of uncertainty. This indicates that the interval in question is not empty. This indicates that the questionable interval is empty;

[0106] As mentioned above, elements falling within the penalty interval can be considered as anomalous data by default. Therefore, the maximum penalty value for the first penalty interval needs to be assigned. To ensure that even if there is only one suspicious interval, the constraint of assigning the maximum penalty value to the first penalty interval is still satisfied, therefore, the following approach is adopted: The penalty value of the penalty interval when the suspected interval is not empty is used to characterize the penalty value of the penalty interval. The specific value of the amplification factor is set by the staff according to the actual situation. For example, it can be set between 1.5 and 3. There is no restriction here. Since the penalty value needs to be scaled proportionally to obtain the penalty coefficient, the penalty value of the penalty interval when the suspected interval is empty can be arbitrarily set to a positive number. For the sake of convenience, the penalty value of the penalty interval when the suspected interval is empty is set to 1. In addition, since the minimum penalty interval and the maximum penalty interval are essentially both characterizing the distribution of abnormal data, the penalty values ​​of the minimum penalty interval and the maximum penalty interval are made to be the same.

[0107] 2) Using the cumulative value equal to 1 as the scaling target, the penalty value of each distribution interval is scaled proportionally, and the scaled penalty value is used as the penalty coefficient of each corresponding distribution interval. The proportional scaling setting retains the relative size relationship of the penalty values ​​of each distribution interval, which facilitates subsequent data analysis and calculation.

[0108] The volatility analysis module statistically analyzes the current probability value and reference probability value of the same operating parameter in each distribution interval under the latest detection time period and each similar detection time period. Then, it determines the current distribution entropy value and the entropy value of each reference distribution. Combining the current distribution entropy value and the standard distribution entropy value determined based on the standard probability value, it determines the neighborhood search radius for recursively expanding clustering of the reference distribution entropy value. Based on the clustering results, it determines the volatility factor.

[0109] The mathematical expression for the distribution entropy is as follows:

[0110]

[0111] In the formula, Represents the distribution entropy value. Let represent the probability value corresponding to the k-th distribution interval, where k is the index of the distribution interval, and Let K be the number of distribution intervals, and let... ;

[0112] It should be noted that when calculating the current distribution entropy value, the probability value... This is the current probability value for the k-th distribution interval, specifically the ratio of the number of samples of the running parameters falling into the k-th distribution interval in the latest detection time period to the total number of samples of the running parameters in the latest detection time period. Similarly, when calculating the reference distribution entropy value, the probability value... This is the reference probability value for the k-th distribution interval. Specifically, it is the ratio of the number of samples of the operating parameters falling into the k-th distribution interval within the corresponding similar detection time period to the total number of samples of the operating parameters within the corresponding similar detection time period. When calculating the standard distribution entropy value, the probability value... That is, the standard probability value of the k-th distribution interval;

[0113] The mathematical expression for the neighborhood search radius is as follows:

[0114]

[0115] In the formula, Indicates the neighborhood search radius. This represents the current distribution entropy value. This represents the entropy value of the standard distribution;

[0116] The logic for recursively expanding clustering the reference distribution entropy values ​​corresponding to each similar detection time period is as follows:

[0117] The reference distribution entropy values ​​corresponding to each similar detection time period are summarized to form a reference distribution entropy value set, and the initial state of each reference distribution entropy value in the reference distribution entropy value set is marked as unvisited.

[0118] Randomly select an unvisited reference distribution entropy value from the set of reference distribution entropy values. , This is the index of the reference distribution entropy value in the set of reference distribution entropy values, and m represents the number of elements in the set of reference distribution entropy values;

[0119] Based on neighborhood search radius Traverse the set of reference distribution entropy values ​​to determine the reference distribution entropy value. neighborhood Specifically, if there exists a reference distribution entropy value in the set of reference distribution entropy values ​​that is in an unvisited state, and this reference distribution entropy value is related to the reference distribution entropy value... If the absolute difference between the values ​​is less than the neighborhood search radius, then this reference distribution entropy value is included in the reference distribution entropy value. neighborhood middle;

[0120] If the reference distribution entropy value neighborhood If the number of internal elements is less than the number of minimum neighboring nodes, then a new unvisited reference distribution entropy value is selected from the set of reference distribution entropy values; otherwise, the reference distribution entropy value is... As the core point and to create clusters Clustering From the reference distribution entropy value and neighboring areas Composition, and based on clustering The status of the reference distribution entropy values ​​within the reference distribution entropy value set is updated in real time, specifically by including the reference distribution entropy value set into clusters. The status of the reference distribution entropy value is changed to visited, where the minimum number of neighboring nodes is between 5% and 10% of the number of elements in the reference distribution entropy value set.

[0121] For the neighborhood The reference distribution entropy value within a given area is used to determine its neighborhood. If the number of reference distribution entropy values ​​in that neighborhood is less than the minimum number of neighborhood nodes, then that neighborhood is not included in the cluster. Otherwise, include the neighborhood in the cluster. It also updates the status of the reference distribution entropy value in real time and continues to recursively expand until clustering occurs. Unable to expand further;

[0122] Randomly select an unvisited reference distribution entropy value from the set of reference distribution entropy values ​​again, and construct a new cluster in the same way until no cluster can be constructed. For reference distribution entropy values ​​that are not assigned to any cluster, mark them as outliers.

[0123] It should be noted that if the reference distribution entropy values ​​in the reference distribution entropy value set are concentrated, it means that under the normal operating conditions corresponding to the latest detection time period, the operating parameters can only fluctuate within a small range, that is, the reasonable fluctuation range of the operating parameters is small. Conversely, if the reference distribution entropy values ​​in the reference distribution entropy value set are discrete, it means that under the normal operating conditions corresponding to the latest detection time period, the operating parameters can fluctuate within a large range, that is, the reasonable fluctuation range of the operating parameters is large. Based on this principle, the absolute difference between the current distribution entropy value and the standard distribution is used as the neighborhood search radius to recursively expand and cluster the reference distribution entropy values ​​in the reference distribution entropy value set. This facilitates the subsequent evaluation of the distribution concentration of each reference distribution entropy value in the reference distribution entropy value set based on the clustering results, and thus the evaluation of the reasonable fluctuation degree of the operating parameters under the normal operating conditions corresponding to the latest detection time period.

[0124] The clustering results include the number of clusters and the number of outliers. Based on the number of clusters and the proportion of outliers, the volatility factor is determined, and its mathematical expression is as follows:

[0125]

[0126] In the formula, Indicates the number of cluster categories. This represents the proportion of outliers, which is the ratio of the number of outliers to the number of elements in the reference distribution entropy set. As a volatility factor, a higher number of cluster types and a higher proportion of outliers indicate greater dispersion and poorer concentration of the reference distribution entropy values ​​within the reference distribution entropy value set. This means that under normal operating conditions corresponding to the latest detection period, the operating parameters can fluctuate within a larger range, i.e., the higher the tolerance for fluctuations in the operating parameters, the larger the volatility factor. The volatility factor is influenced by the number of cluster types. The impact of the proportion of isolated points and interaction effects It consists of three parts;

[0127] It should be noted that under the ideal centralized distribution, the number of clusters is 1. As the number of clusters increases, the entropy values ​​of the reference distributions in the reference distribution entropy value set become more dispersed. Therefore, the Sigmoid function form is used here. To characterize the influence of the number of clustering categories, in the formula This is a growth rate adjustment factor with a value greater than 0, used to adjust the growth rate of the factor affecting the number of cluster species. The specific value is determined based on the quantitative definition of the fluctuation caused by the number of cluster species. A larger value indicates greater sensitivity to the quantitative definition of the fluctuation caused by the number of cluster species. The specific value is not limited here. In the formula... The value of 1 is a preset threshold, indicating that the degree of fluctuation begins to increase rapidly as the number of cluster categories exceeds 1.

[0128] Similarly, as the proportion of isolated points increases, it also indicates that the entropy values ​​of each reference distribution in the set of reference distribution entropy values ​​become more discrete. Therefore, we continue to use the Sigmoid function form here. To characterize the influence of the proportion of isolated points, in the formula This is also a growth rate adjustment factor with a value greater than 0, used to adjust the growth rate of the outlier proportion influence term. The specific value is determined based on the quantitative definition of the fluctuation caused by the outlier proportion by the staff. A larger value indicates greater sensitivity to the quantitative definition of the fluctuation caused by the outlier proportion by the staff. The specific value is not limited here. In the formula... This is the outlier percentage threshold, representing the tolerance level of staff for the proportion of outliers when defining volatility. A higher value indicates a higher tolerance for outliers. The specific value is set by the staff based on the actual situation, generally between 3% and 5%. Setting it to 5% indicates that the volatility begins to increase rapidly once the proportion of outliers exceeds 5%. Furthermore, when determining the volatility based on the number of cluster types and the proportion of outliers, both factors often need to be considered together. Therefore, the conventional simple multiplication method is used here to construct the interaction term. ;

[0129] In the formula, , and These are all preset proportional coefficients, used to characterize the weights of the clustering category number influence term, the outlier proportion influence term, and the interaction influence term in the volatility factor calculation. The specific values ​​of the three can be determined based on the analytic hierarchy process, which will not be elaborated here.

[0130] The deviation feature extraction module combines the distribution interval, current probability value, standard probability value and penalty coefficient to determine the penalty gain, concentration deviation, reasonable interval probability difference and distribution entropy value gain of each operating parameter in the latest detection period, and constructs the operating feature vector of each operating parameter by combining the volatility factor and statistical feature difference.

[0131] The mathematical expression for the penalty gain is as follows:

[0132]

[0133] In the formula, The penalty gain for the running parameters, This represents the penalty coefficient for the k-th distribution interval. This represents the current probability value of the running parameters in the k-th distribution interval. This represents the standard probability value of the same operating parameter in the kth distribution interval.

[0134] It should be noted that when calculating the penalty gain for each operating parameter, firstly by... The current penalty level of the quantified operating parameter is used to characterize the degree of abnormality of that operating parameter in the latest detection time period. The larger the value, the higher the degree of abnormality. Then, through... The standard value for the degree of abnormality of a certain operating parameter of public electrical equipment under normal conditions is used. The larger the value, the higher the tolerance for the degree of abnormality of the operating parameter under normal conditions. Finally, the difference between the two is calculated as a penalty gain. The larger the value, the higher the probability that the operating parameter is in an abnormal state during the latest detection period;

[0135] The concentration bias is a vector comprising the median bias and the width bias of the interval. The logic for obtaining the median bias and the width bias is as follows:

[0136] 1) For any operating parameter, sort all elements in the latest detection time period in ascending order of value to form the current value sequence of the operating parameter. Determine the window length based on the number of elements collected and the reasonableness ratio of the operating parameter in the latest detection time period. Slide the window sequentially in the current value sequence of the operating parameter and calculate the difference between the maximum and minimum values ​​in each window as the interval width. Select the window corresponding to the minimum interval width as the concentration window. Use the maximum and minimum values ​​in the concentration window as the upper and lower limits of the concentration distribution interval, respectively, to obtain the concentration distribution interval of the operating parameter in the latest detection time period.

[0137] It should be noted that the methods for obtaining the concentrated distribution interval and the reasonable interval are the same, and will not be repeated here. Here, the reasonableness ratio is used to determine the setting of the concentrated distribution interval, which plays the role of controlling variables to ensure that the probability value of the operating parameter falling into the concentrated distribution interval in the latest detection period is close to the standard probability value corresponding to the reasonable interval. This provides a basis for the subsequent analysis of the concentrated distribution interval and the reasonable interval to determine the degree of deviation of the operating parameter from the normal condition in the latest detection period.

[0138] 2) Determine the median value of the concentrated distribution interval and the median value of the reasonable interval, and take the absolute difference between the two as the median deviation. The median value is the average of the upper and lower limits of the interval.

[0139] 3) Determine the width of the concentrated distribution interval and the width of the reasonable interval, and take the difference between the two as the interval width deviation. The interval width is the difference between the upper and lower limits of the interval.

[0140] It should be noted that the larger the median deviation of the interval, the greater the difference between the concentrated distribution position of this type of operating parameter in the latest detection period and the concentrated distribution position under normal conditions. In other words, the higher the probability that this type of operating parameter of the public electrical equipment is abnormal in the latest detection period. The larger the interval width deviation, the larger the width of the concentrated distribution interval compared to the area above the reasonable range. In other words, the more dispersed this type of operating parameter of the public electrical equipment is in the latest detection period. By combining the median deviation of the interval, it can be used to measure the probability that this type of operating parameter of the public electrical equipment is abnormal in the latest detection period.

[0141] For example, if the median deviation is large, a smaller interval width deviation indicates that the operating parameter of the public electrical equipment is concentrated within the abnormal range during the latest testing period, and the probability of the public electrical equipment being in an abnormal state is higher. Conversely, a larger interval width deviation indicates that the operating parameter of the public electrical equipment is not concentrated within the abnormal range during the latest testing period, and the probability of the public electrical equipment being in an abnormal state is relatively lower. On the other hand, if the median deviation is small, a smaller interval width deviation indicates that the operating parameter of the public electrical equipment is concentrated within the normal range during the latest testing period, and the probability of the public electrical equipment being in an abnormal state is lower. Conversely, a larger interval width deviation indicates that the operating parameter of the public electrical equipment is not concentrated within the normal range during the latest testing period, and the probability of the public electrical equipment being in an abnormal state is relatively higher.

[0142] Among them, the probability difference of the reasonable interval is the difference between the current probability value and the standard probability value of the operating parameter in the reasonable interval. The larger the value, the more reasonable values ​​the operating parameter has in the latest detection period, which means that the probability of the operating parameter being abnormal in the latest detection period is smaller.

[0143] Among them, the distribution entropy gain is the difference between the current distribution entropy value and the standard distribution entropy value. The larger the value, the more dispersed the distribution of this operating parameter is in the latest detection time period compared with the distribution under normal conditions.

[0144] Among them, the statistical characteristic differences include mean difference, amplitude difference and variance difference, which are the differences between the mean, amplitude and variance of the operating parameters in the latest detection period and the mean, amplitude and variance of the elements in the corresponding standard value set of operating parameters. The amplitude is the difference between the maximum value and the minimum value. The calculation methods of mean, amplitude and variance are common knowledge in the art and will not be elaborated here.

[0145] As an example, the runtime feature vector of the runtime parameters is represented as follows:

[0146]

[0147] In the formula, Indicates the running feature vector, These represent the penalty gain, interval median deviation, interval width deviation, reasonable interval probability difference, distribution entropy gain, volatility factor, mean difference, amplitude difference, and variance difference, respectively.

[0148] The equipment fault detection module is used to establish a mapping relationship between the operating feature vectors of each public electrical equipment and their operating status, and to process the operating feature vectors of each public electrical equipment in the latest detection time period based on the mapping relationship, so as to output the operating status of each public electrical equipment.

[0149] Specifically, for any public electrical equipment, a deep neural network model is used to establish the mapping relationship between its operating feature vector and operating status, as follows:

[0150] The deep neural network model uses the TensorFlow deep learning framework, which includes an input layer to receive the operating feature vectors of various operating parameters of the public electrical equipment, one or more hidden layers to process the input operating feature vectors, with ReLU selected as the activation function in the hidden layers, and an output layer to output the operating status of the public electrical equipment, which is either normal or abnormal. The training process of the deep neural network model is as follows:

[0151] Operating parameters of public electrical equipment are collected over historical time periods under multiple known operating states. Operating feature vectors for each operating parameter of the public electrical equipment in each historical time period are obtained using the same method. The operating feature vectors and operating states are aggregated to form a sample set. This sample set is divided into a training set, a test set, and a validation set in a 70:15:15 ratio. Batch size and training period are set, with the batch size between 32 and 256 and the training period between 10 and 100. The operating feature vectors of each operating parameter of the public electrical equipment in the training set are used as input, and the corresponding operating states are used as output labels to train a deep neural network model. The model is designed to perform training using binary cross-entropy loss as the loss function. During training, the model parameters are updated through backpropagation to minimize the loss function value. Specifically, the model parameters can be updated using optimization algorithms such as Adam and SGD. During training, the model hyperparameters, such as learning rate, batch size, and number of hidden layers, are adjusted using a validation set to optimize model performance. The initial value of the learning rate can be set between 0.00001 and 0.0001, and the initial number of hidden layers can be set to 6-8. After reaching the predetermined training period, the test set is input into the data simulation module for performance testing. If the accuracy is above 95%, the training is considered complete; otherwise, training is repeated.

[0152] It should be noted that the penalty gain, concentration bias, reasonable interval probability difference, and distribution entropy gain measure the differences of operating parameters between the latest detection period and previous normal detection periods from the perspective of probability distribution. The volatility factor quantifies the dispersion of such operating parameters in previous normal detection periods. Finally, the statistical feature difference measures the differences of operating parameters between the latest detection period and previous normal detection periods from a statistical perspective. Therefore, compared with using conventional operating parameters as model input, using the above parameters as model input simplifies the number of input parameters and can accurately characterize the relative differences of each operating parameter of public electrical equipment between the latest detection period and previous normal periods. This provides an accurate theoretical reference for the model to predict the operating status of public electrical equipment, significantly improving the efficiency of model learning and training and the accuracy of prediction.

[0153] As a comparative reference, for each public electrical device, a reference model based on conventional operating parameters was constructed and trained. Operating parameters of 20 public electrical devices were retrieved over 100 historical time periods with known operating states. These operating parameters were not used in the training of either the current model or the reference model. The operating parameters were directly input into the reference model for operating state prediction, and then converted into operating feature vectors and input into the current model for operating state prediction. The prediction accuracy of both models is shown in Table 1 below.

[0154] Table 1. Comparison of Prediction Accuracy

[0155]

[0156] As shown in Table 1 above, the average prediction accuracy based on the reference model is 83.75%, the maximum is 93.92%, and the minimum is 75.68%, with a variation of 18.25%. In contrast, the average prediction accuracy based on the proposed model is 95.68%, an improvement of 11.92%, the maximum is 97.77%, an improvement of 3.85%, and the minimum is 93.18%, an improvement of 17.5%, with a variation of 4.6%, and a decrease of 13.65%. For a clearer comparison, Figure 2 in the accompanying diagram below shows the prediction accuracy of the proposed model (red line) and the prediction accuracy of the reference model (blue line). This demonstrates that the proposed model has a higher prediction accuracy than the reference model. Furthermore, the framework of the proposed model is highly adaptable to various public electrical equipment. Each public electrical equipment can use the methods provided in this proposed model to build and train a prediction model suitable for itself. Therefore, this proposed model has wide system applicability and high prediction accuracy.

[0157] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0158] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0159] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0160] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A detection system for public electrical equipment in intelligent buildings, characterized in that, include: The parameter acquisition module is used to collect the operating condition identification parameters and operating parameters of various public electrical equipment in intelligent buildings during the latest detection time period when they are in stable operating conditions. Based on the operating condition identification parameters, it retrieves the operating parameters of the public electrical equipment during multiple similar detection time periods under normal operation to construct a set of standard values ​​for operating parameters. The interval partitioning module, based on the set of standard values ​​for operating parameters, determines the distribution interval and standard probability value of each operating parameter, and determines the penalty coefficient for each distribution interval based on the standard probability value. The distribution interval includes a reasonable interval and a penalty interval, and whether the doubtful interval is empty is determined by the distribution status of the elements in the set of standard values ​​for operating parameters. The volatility analysis module statistically analyzes the current probability value and reference probability value of the same operating parameter in each distribution interval under the latest detection time period and each similar detection time period, and then determines the current distribution entropy value and the entropy values ​​of each reference distribution. Combining the current distribution entropy value and the standard distribution entropy value determined based on the standard probability value, it determines to perform recursive expansion on the reference distribution entropy value. The clustering neighborhood search radius determines the volatility factor based on the clustering results; the deviation feature extraction module, combining the distribution interval, current probability value, standard probability value, and penalty coefficient, determines the penalty gain, concentration deviation, reasonable interval probability difference, and distribution entropy gain of each operating parameter in the latest detection period, and constructs the operating feature vector of each operating parameter by combining the volatility factor and statistical feature differences; the equipment fault detection module is used to establish the mapping relationship between the operating feature vector of each public electrical equipment and its operating status, and processes the operating feature vector of each public electrical equipment in the latest detection period based on the mapping relationship to output the operating status of each public electrical equipment.

2. The detection system for public electrical equipment in intelligent buildings according to claim 1, characterized in that, The definition of stable operation of public electrical equipment is as follows: if the state-type operating condition identification parameters of the public electrical equipment remain unchanged within a detection period, and the coefficient of variation of the numerical operating condition identification parameters is less than a preset threshold, then the operating state of the public electrical equipment within this detection period is defined as stable operation. The operating condition identification parameters are used to identify the operating conditions of the public electrical equipment in the latest detection period, including but not limited to environmental parameters that affect the operating performance of the public electrical equipment, and setting parameters that control the operation of the public electrical equipment to achieve the desired conditions. The operating parameters are used to reflect the operating state of the public electrical equipment in the latest detection period, including but not limited to one or more of the voltage value, current value, and power factor of the public electrical equipment in the latest detection period.

3. The detection system for public electrical equipment in intelligent buildings according to claim 2, characterized in that, The logic for obtaining the set of standard operating parameter values ​​is as follows: Based on the operating condition identification parameters of the public electrical equipment in the latest detection time period, the operating condition of the public electrical equipment in the latest detection time period is determined. Then, from the previous normal detection time periods of the public electrical equipment, similar detection time periods with similar operating conditions are selected. The operating parameters of the public electrical equipment in the similar detection time periods are used as the standard operating parameter values, and the same type of standard operating parameter values ​​are summarized to form a set of standard operating parameter values. The logic for obtaining similar detection time periods is as follows: For any public electrical equipment, if there exists a normal detection time period that satisfies the following conditions: the relative error between the duration of the normal detection time period and the duration of the latest detection time period is lower than a preset threshold, the state-type operating condition identification parameters of the public electrical equipment in this normal detection time period are consistent with the state-type operating condition identification parameters in the latest detection time period, and the relative error between the average value of the numerical operating condition identification parameters and the average value of the same type of numerical operating condition identification parameters in the latest detection time period is lower than a preset threshold, then it is determined that the operating condition of the public electrical equipment in this normal detection time period is similar to that in the latest detection time period, and this normal detection time period is taken as a similar detection time period.

4. The detection system for public electrical equipment in intelligent buildings according to claim 1, characterized in that, The standard probability value for each distribution interval is obtained as follows: For any distribution interval, the ratio of the number of elements falling into that distribution interval to the total number of elements in the standard value set of the operating parameters is calculated, and this ratio is used as the standard probability value for that distribution interval. The logic for determining the distribution interval is as follows: 1) For any operating parameter, all elements in its corresponding standard value set are sorted from smallest to largest to form a sequence of standard values. The window length is determined based on the number of elements in the standard value set and a preset reasonableness ratio. The window is then slid sequentially through the standard value sequence of the operating parameters, and the difference between the maximum and minimum values ​​in each window is calculated as the interval width. The window corresponding to the minimum interval width is selected as the optimal window, and the maximum and minimum values ​​in the optimal window are used as the upper and lower limits of the reasonable interval, respectively. The mathematical expression for the window length is as follows: In the formula, For window length, This indicates an upward fetch operation. The number of elements in the set of standard values ​​for the running parameters. 1) The reasonableness percentage of values ​​greater than 0; 2) With the standard probability value of the doubtful interval not lower than the preset threshold as a constraint, expand to both sides of the reasonable interval to form a doubtful interval with the same width as the interval; 3) Extract the minimum lower limit and maximum upper limit of the reasonable interval and all doubtful intervals, and use them as the upper limit of the lowest penalty interval and the lower limit of the highest penalty interval, respectively, with the lower limit of the lowest penalty interval being negative infinity and the upper limit of the highest penalty interval being positive infinity.

5. The detection system for public electrical equipment in intelligent buildings according to claim 4, characterized in that, The logic for determining the penalty coefficient for each distribution interval is as follows: 1) Based on the standard probability value, assign a corresponding penalty value to each distribution interval, specifically as follows: 1.1) Assign a penalty value of 0 to the reasonable interval; 1.2) If the doubtful interval is not empty, for any doubtful interval, calculate the ratio of the reasonable interval to its standard probability value, and use this ratio as the penalty value for the doubtful interval. If the doubtful interval is empty, do not calculate its penalty value; 1.3) Based on the penalty value of the doubtful interval, determine the penalty value for the penalty interval, the mathematical expression of which is as follows: In the formula, This represents the penalty value within the penalty interval. This represents the sum of the penalty values ​​for all questionable intervals. This represents a magnification factor with a value greater than 1. Indicates the range of uncertainty. This indicates that the interval in question is not empty. 1) Indicate that the doubtful interval is empty; 2) With the cumulative value equal to 1 as the scaling target, the penalty value of each distribution interval is scaled proportionally, and the scaled penalty value is used as the penalty coefficient of each corresponding distribution interval.

6. The detection system for public electrical equipment in intelligent buildings according to claim 1, characterized in that, The mathematical expression for the neighborhood search radius is as follows: In the formula, Indicates the neighborhood search radius. This represents the current distribution entropy value. The standard distribution entropy value is represented. The logic of recursive extended clustering for the reference distribution entropy values ​​corresponding to each similar detection time period is as follows: the reference distribution entropy values ​​corresponding to each similar detection time period are summarized to form a reference distribution entropy value set, and the initial state of each reference distribution entropy value in the reference distribution entropy value set is marked as unvisited. Randomly select an unvisited reference distribution entropy value from the set of reference distribution entropy values. ， This is the index of the reference distribution entropy value in the set of reference distribution entropy values, and , m represents the number of elements in the reference distribution entropy value set; based on the neighborhood search radius Traverse the set of reference distribution entropy values ​​to determine the reference distribution entropy value. neighborhood Specifically, if there exists a reference distribution entropy value in the set of reference distribution entropy values ​​that is in an unvisited state, and this reference distribution entropy value is related to the reference distribution entropy value... If the absolute difference between the values ​​is less than the neighborhood search radius, then this reference distribution entropy value is included in the reference distribution entropy value. neighborhood middle; If the reference distribution entropy value neighborhood If the number of internal elements is less than the number of minimum neighboring nodes, then a new unvisited reference distribution entropy value is selected from the set of reference distribution entropy values; otherwise, the reference distribution entropy value is... As the core point and to create clusters Clustering From the reference distribution entropy value and neighboring areas Composition, and based on clustering The status of the reference distribution entropy values ​​within the reference distribution entropy value set is updated in real time, specifically by including the reference distribution entropy value set into clusters. The reference distribution entropy value state is modified to "visited"; for the neighborhood... The reference distribution entropy value within a given area is used to determine its neighborhood. If the number of reference distribution entropy values ​​in that neighborhood is less than the minimum number of neighborhood nodes, then that neighborhood is not included in the cluster. Otherwise, include the neighborhood in the cluster. It also updates the status of the reference distribution entropy value in real time and continues to recursively expand until clustering occurs. Unable to expand further; Randomly select another unvisited reference distribution entropy value from the set of reference distribution entropy values ​​again, and construct a new cluster in the same way until no cluster can be constructed. For reference distribution entropy values ​​that are not assigned to any cluster, mark them as outliers.

7. The detection system for public electrical equipment in intelligent buildings according to claim 1, characterized in that, The clustering results include the number of clusters and the number of outliers. Based on the number of clusters and the proportion of outliers, the volatility factor is determined, and its mathematical expression is as follows: In the formula, Indicates the number of cluster categories. Indicates the proportion of isolated points. As a volatility factor, and A growth rate adjustment factor greater than zero. The threshold for the proportion of isolated points. 、 and All of these are preset proportional coefficients, and their specific values ​​are determined based on the analytic hierarchy process (AHP).

8. The detection system for public electrical equipment in intelligent buildings according to claim 1, characterized in that: The mathematical expression for the penalty gain is as follows: In the formula, The penalty gain for the running parameters, This represents the penalty coefficient for the k-th distribution interval. This represents the current probability value of the running parameters in the k-th distribution interval. Let k be the standard probability value of the same operating parameter in the k-th distribution interval, where k is the index of the distribution interval. K is the number of distribution intervals; the concentration deviation is a vector including the median deviation and the interval width deviation. The logic for obtaining the median deviation and the interval width deviation is as follows: 1) For any operating parameter, sort each element in the latest detection time period in ascending order of value to form a sequence of current values ​​of the operating parameter. Determine the window length based on the number of elements collected and the reasonableness ratio of the operating parameter in the latest detection time period. Slide the window sequentially in the sequence of current values ​​of the operating parameter and calculate the difference between the maximum and minimum values ​​in each window as the interval width. Select the window corresponding to the minimum interval width as the concentration window. Use the maximum and minimum values ​​in the concentration window as the upper and lower limits of the concentration distribution interval, respectively, to obtain the concentration of the operating parameter in the latest detection time period. 1) Determine the median of the concentrated distribution interval and the median of the reasonable interval, and take the absolute difference between the two as the median deviation. The median of the interval is the mean of the upper and lower limits of the interval. 2) Determine the width of the concentrated distribution interval and the width of the reasonable interval, and take the difference between the two as the width deviation. The width of the interval is the difference between the upper and lower limits of the interval. The probability difference of the reasonable interval is the difference between the current probability value and the standard probability value of the operating parameter at the reasonable interval. The distribution entropy gain is the difference between the current distribution entropy value and the standard distribution entropy value. The statistical feature difference includes mean difference, amplitude difference and variance difference, which are the differences between the mean, amplitude and variance of the operating parameter in the latest detection time period and the mean, amplitude and variance of the elements in the corresponding standard value set of the operating parameter.

9. The detection system for public electrical equipment in intelligent buildings according to claim 1, characterized in that: For any public electrical equipment, a deep neural network model is used to establish the mapping relationship between its operating feature vector and operating status. The deep neural network model uses the TensorFlow deep learning framework, which includes an input layer for receiving the operating feature vectors of various operating parameters of the public electrical equipment, one or more hidden layers for processing the input operating feature vectors, with ReLU selected as the activation function in the hidden layers, and an output layer for outputting the operating status of the public electrical equipment, which is normal or abnormal.

Images