A method for detecting a user's power load using an intelligent electric energy meter

Abstract

The present application discloses a method for detecting a user's power load identification intelligent electric energy meter, comprising: dividing all power lines in the power management area into several power line groups according to a preset power consumption behavior feature similarity algorithm; obtaining a power load fluctuation sequence of each power line based on each power line group, and generating a power load deviation sequence of each power line from other power lines in the power line group; if there is a time segment in a power load deviation sequence that reaches a deviation threshold greater than , the corresponding power line is determined as a target power line; generating an abnormal probability value sequence of each electric energy meter based on each target power line; judging whether the abnormal probability value sequence of each electric energy meter in the target power line meets the preset conditions, and obtaining the detection result of each electric energy meter in the target power line. Thus, more efficient and accurate detection of electric energy meters can be achieved in the power management area, saving power resources.

Description

Technical Field

[0001] The present invention relates to the technical field of electric energy meter detection, and in particular to a detection method of an intelligent electric energy meter for identifying user power load. Background Art

[0002] As a measuring instrument for electricity settlement, the accuracy of the electric energy meter will directly affect the fairness, accuracy and reliability of electricity settlement, and is directly related to the vital interests of power supply companies and users. Once a metering failure occurs in the electric energy meter, it will lead to inaccurate electricity settlement and electricity disputes are very likely to occur. In order to achieve self-checking of metering failures, the current conventional metering self-checking method is to use two independent metering circuits (metering module 1 and metering module 2) inside the electric energy meter to perform separate measurements and compare the metering results of the two metering modules. If the two are inconsistent, it means that there is a metering failure, and the metering module needs to be reset and other operations to troubleshoot the problem. The above-mentioned method of dual-channel metering self-checking requires an additional comparison metering module. This method is costly and is usually limited by the size of the structural space.

[0003] The invention patent with patent application number 202010475853.6 discloses a self-check method and system for metering faults of electric energy meters. The MCU of the electric energy meter to be self-checked intercepts the power load parameter-voltage data of multiple electric energy meters connected to the same power line and compares it with the voltage data on the same power line multiple times, so that the metering faults of the electric energy meter to be self-checked can be self-checked online. The self-check method is implemented by a pure software method and can realize self-checking without additional hardware. It has low cost and simple operation.

[0004] However, in the actual power management area, there are a large number of electricity meters. The workload of mining metering anomalies for the electricity meters on all power lines in the entire power management area through the above-mentioned self-inspection method is too large, and there is no targeted screening of self-inspections. Since all electricity meters in the area need to perform self-inspections at the same time during the self-inspection process, a large amount of data information needs to be transmitted, which may increase the network burden and waste collection resources. In addition, not only are there differences in the power load characteristics of electricity meters on different power lines, but different electricity meters on the same power line will also differ due to the users' power consumption patterns. Summary of the invention

[0005] The present application provides a method for detecting an intelligent electric energy meter for identifying a user's power load, which can achieve more efficient and accurate detection of electric energy meters in a power management area, thereby saving power resources.

[0006] The present application provides a method for detecting a smart electric energy meter for identifying a user's electric load, comprising:

[0007] S101, dividing all power lines in the power management area into a number of power line groups according to a preset power consumption behavior feature similarity algorithm;

[0008] S102, based on each power line group, obtain the power load extreme value of each power line in k time segments in the past day to form the power load fluctuation sequence of the power line , and generate the load deviation sequence of the power line and other power lines in the power line group ;

[0009] S103, based on all the power load deviation sequences of each power line in the power line group, if there is a power load deviation sequence in which the number of time segments reaching the deviation threshold is greater than , then the corresponding power line is determined as the target power line;

[0010] S104, based on each target power line, obtaining the voltage deviation set of each electric energy meter and its corresponding abnormal probability value every preset period until the preset monitoring time is reached, and generating an abnormal probability value sequence of each electric energy meter; the monitoring time is greater than the preset period, and the monitoring time is an integral multiple of the preset period;

[0011] S105, judging whether the abnormal probability value sequence of each electric energy meter in the target power line meets a preset condition, and obtaining a detection result of each electric energy meter in the target power line.

[0012] Preferably, the preset electricity consumption behavior feature similarity algorithm specifically includes:

[0013] A1. Based on each power line, obtain its power consumption in a preset historical time period and each time unit, and calculate the power consumption characteristic value of the power line according to the following formula:

[0014] .

[0015] Among them, F is the power consumption characteristic value of the corresponding power line in the preset historical time period, is the total power consumption of the ith electric energy meter on the power line in the preset historical time period, n is the number of electric energy meters, m is the total number of time units in the preset historical time period, is the total power consumption on the power line in the jth time unit, is the average power consumption of all time units, and are the preset weight coefficient values ​​respectively;

[0016] A2. Based on the rate of change of power consumption between adjacent time units , t=1,2,...,m-1, and obtain the time series of power consumption changes of the corresponding power line in the preset historical time period , generate the first statistical value of the power consumption change time series and the second statistic ;

[0017] A3. Construct a power consumption behavior characteristic vector v= of each power line according to the power consumption characteristic value, the first statistical value and the second statistical value. ;

[0018] A4. Calculate the similarity of the power consumption behavior feature vectors between all power lines, construct a power line similarity matrix, and obtain several power line groups according to a preset similarity threshold.

[0019] Preferably, in S104, obtaining the voltage deviation set of each electric energy meter and its corresponding abnormal probability value every preset period includes:

[0020] B1. Based on each electric energy meter, the voltage value at the current moment and the effective voltage value set of other electric energy meters on the power line are sampled every preset period; wherein the effective voltage value set is specifically: among the voltage values ​​sampled at the current moment by all electric energy meters in the power line except the electric energy meter, the maximum value and the minimum value are deleted, and the remaining voltage values ​​form an effective voltage value set; B2. The difference between the electric energy meter and each voltage value in the effective voltage value set is calculated respectively to form a voltage deviation set of the electric energy meter, each element in the set corresponds to a difference, and the ratio of the number of elements in the voltage deviation set that are greater than the preset reference difference threshold to the total number of elements is determined as the abnormal probability value;

[0021] Before sampling the voltage value of the next preset cycle, if the abnormal probability value of the previous cycle is greater than the first threshold, the metering chip of the electric energy meter to be detected is reset and the metering parameters are refreshed.

[0022] Preferably, the S105 specifically includes:

[0023] C1. When each abnormal probability value in the abnormal probability value sequence of the electric energy meter is less than the first threshold value, it is determined that the electric energy meter is measuring normally;

[0024] C2. When each abnormal probability value in the abnormal probability value sequence of the electric energy meter is greater than the first threshold value, the electric energy meter is determined to have a metering fault, a metering fault information flag is set, and the metering fault information flag is reported to the power management area;

[0025] C3. Otherwise, the difference in power consumption during the monitoring period between the electric energy meter and other electric energy meters on the power line that are judged to have normal metering is obtained as a metering error value, which is input into the corresponding pre-trained metering fault identification model, and the detection result of the electric energy meter is output, and the detection result includes normal metering and metering fault.

[0026] Preferably, each power line group corresponds to a metering fault identification model, and the corresponding metering fault identification model is determined according to the power line where the electric energy meter is located. The specific process of obtaining the pre-trained metering fault identification model is as follows:

[0027] In a certain power line group, a large amount of power consumption of all power meters within a historical monitoring period is obtained, and the power meters with stable and normal measurement on each power line in the power line group are selected as simulation objects. The difference between the power consumption of each power meter within all historical monitoring periods and the power consumption of the corresponding simulation object is obtained as a training data set, and each training data is labeled.

[0028] Preferably, the metering fault identification model includes several sub-models, and step C3 further includes:

[0029] S201, based on a preset abnormality ratio value interval [0, ),[ , ),[ , 1], each corresponding to a sub-model; wherein the abnormal proportion degree value is specifically: the ratio of the number of abnormal probability values ​​in the abnormal probability value sequence greater than the first threshold to the total number of abnormal probability values ​​in the abnormal probability value sequence;

[0030] S202, obtaining the abnormal proportion degree value corresponding to the abnormal probability value sequence of the electric energy meter, and locating the corresponding sub-model according to the abnormal proportion degree value interval in which the abnormal proportion degree value is located;

[0031] S203, inputting the measurement error value of the electric energy meter into the corresponding sub-model, and outputting the detection result of the electric energy meter;

[0032] The training method of the sub-model corresponding to each abnormal proportion value interval specifically includes:

[0033] D1. In the acquired training data set, obtain the abnormal proportion degree value corresponding to the abnormal probability value sequence obtained by the electric energy meter corresponding to each training data on the power line where it is located;

[0034] D2. According to the abnormality ratio value of each training data, divide it into corresponding abnormality ratio value intervals to obtain a sub-training set corresponding to each abnormality ratio value interval;

[0035] D3. Use the sub-training set to train and optimize the parameters of the preset network model to obtain a sub-model corresponding to each abnormal proportion value interval.

[0036] Preferably, in B1, the method for determining the effective voltage value set of each electric energy meter further includes:

[0037] S301, respectively calculating the power usage status similarity index between the electric energy meter and the remaining electric energy meters on the power line;

[0038] S302, using an electric energy meter whose power consumption status similarity index is greater than a preset status similarity threshold as a reference electric energy meter;

[0039] S303: Remove the maximum value and the minimum value from the voltage values ​​sampled by all reference electric energy meters at the current moment to obtain a valid voltage value set of the electric energy meter.

[0040] Preferably, the calculation method of the power consumption status similarity index between two electric energy meters is specifically as follows:

[0041] E1. Obtain the voltage value of each electric energy meter at each sampling time in the historical time window. The historical time window is set to nearly 6 hours, and obtain several data points (sampling time, voltage value) arranged in chronological order.

[0042] E2. Fit all data points of the electric energy meter using a preset fitting algorithm to obtain a sampling time-voltage value curve of each electric energy meter;

[0043] E3. Based on the sampling time-voltage value curve of each electric energy meter, the power consumption status similarity index between the two electric energy meters is calculated according to the following formula:

[0044] .

[0045] in, and are the voltage values ​​of the energy meters A and B at the sampling time t respectively, [t1, t2] is the time interval pre-set based on the historical time window, and a time interval is selected in the historical time window to calculate the similarity index of the power consumption status.

[0046] Preferably, in B2, the ratio of the number of elements in the voltage deviation set that are greater than a preset difference threshold to the total number of elements is determined as the abnormal probability value, and further includes:

[0047] S401, based on the reference electric energy meter corresponding to each element in the voltage deviation set of the electric energy meter, respectively obtain a power consumption status similarity index between the electric energy meter and each reference electric energy meter;

[0048] S402: adaptively adjusting a preset difference threshold of each reference electric energy meter according to a similarity index of the electric energy meter's power usage status with each reference electric energy meter.

[0049] Preferably, the S402 specifically includes:

[0050] According to the preset power consumption status similarity index interval-adjustment factor, determine the adjustment factor corresponding to the power consumption status similarity index interval of each reference electric energy meter, multiply the preset difference threshold by the corresponding adjustment factor, obtain the adjusted difference threshold of the reference electric energy meter, and replace the original preset difference threshold;

[0051] The preset power consumption status similarity index interval-adjustment factor is specifically: [a1, a2)-1.2, [a2, a3)-1, [a3, 1]-0.8, a1 is the preset status similarity threshold, a2 and a3 are adaptively set according to the preset status similarity threshold to satisfy a2 less than a3, a2 greater than a1, and a3 less than 1.

[0052] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0053] All power lines in the power management area are divided into several power line groups through the power consumption behavior feature similarity algorithm. The target power line with possible faults is determined according to the power load deviation sequence of each power line in each power line group, and mutual self-inspection between electric energy meters is carried out on the target power line, which reduces the workload of comprehensive self-inspection, improves the detection efficiency, avoids the data transmission peak caused by the simultaneous self-inspection of all electric energy meters, and reduces the network burden and collection resource waste in the power management area; combined with power load analysis and abnormal probability value monitoring of electric energy meters, electric energy meters with metering abnormalities can be located more accurately, and differentiated detection methods are carried out according to different abnormal probability value sequence distributions, thereby realizing efficient and accurate detection of electric energy meters.

[0054] By further subdividing the metering fault identification model into several sub-models, more refined detection is performed on electric energy meters with different abnormal proportion values ​​within the monitoring period. Different sub-models are used for detection, avoiding the problems of training errors and weak generalization ability that may be caused by using a unified model, and improving the detection accuracy and generalization ability of the model. The training of the sub-model takes into account the differences in the abnormal probability value sequences between the historically abnormal electric energy meters and other electric energy meters on the same power line, so that the model has better judgment ability and adaptability to different types of metering error values.

[0055] By calculating the similarity index of power consumption status between the power meters on a power line and selecting the power meters with similar power consumption status as reference, the comparison in the self-check process is made more standardized, and the accuracy and reliability of anomaly detection are improved; and the local errors that may be caused by the clustering of the overall power consumption behavior of the power line are compensated.

[0056] By adaptively adjusting the difference threshold of each reference electricity meter, the actual differences between the electricity meters can be reflected more accurately, reducing misjudgments and missed judgments caused by fixed difference thresholds; each reference electricity meter corresponds to a personalized difference threshold, which makes the calculation of the abnormal probability value more refined and helps to discover more hidden abnormal problems; the difference threshold can be automatically adjusted according to similar indicators of power consumption status between the electricity meters, thereby improving the adaptability and intelligence of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of a flow chart of a method for detecting a user's power load using a smart electric energy meter according to an embodiment of the present invention;

[0058] Figure 2 A schematic diagram of a preset electricity consumption behavior feature similarity algorithm according to an example of the present invention. DETAILED DESCRIPTION

[0059] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0061] Embodiment 1: Figure 1 The present invention is a flowchart of a method for detecting a user's power load using a smart energy meter according to an embodiment of the present invention.

[0062] like Figure 1 As shown, a method for detecting a user's power load by identifying an intelligent electric energy meter comprises the following steps:

[0063] S101, dividing all power lines in a power management area into a plurality of power line groups according to a preset power consumption behavior feature similarity algorithm.

[0064] Among them, the data output end of the electric energy meter on each power line is connected to a collection system for collecting metering data transmitted by the metering chip of each electric energy meter on the power line. The metering data includes but is not limited to voltage, current, power load, and power consumption.

[0065] Specifically, the preset electricity consumption behavior feature similarity algorithm includes:

[0066] A1. Based on each power line, obtain its power consumption in a preset historical time period and each time unit, and calculate the power consumption characteristic value of the power line according to the following formula:

[0067] .

[0068] Among them, F is the power consumption characteristic value of the corresponding power line in the preset historical time period, is the total power consumption of the ith electric energy meter on the power line in the preset historical time period, n is the number of electric energy meters, m is the total number of time units in the preset historical time period, is the total power consumption on the power line in the jth time unit, is the average power consumption of all time units, and They are respectively pre-set weight coefficient values, which can be adjusted according to actual conditions to reflect the influence of power consumption and time distribution of power consumption on the power consumption characteristic value of the power line.

[0069] Among them, the preset historical time period can be set to one month, which can accurately reflect the long-term electricity consumption behavior pattern of the power line.

[0070] A2. Based on the rate of change of power consumption between adjacent time units , t=1,2,...,m-1, and obtain the time series of power consumption changes of the corresponding power line in the preset historical time period , generate the first statistical value of the power consumption change time series and the second statistic .

[0071] It can be understood that the first statistical value can be set to the average value of the time series of power consumption changes, and the second statistical value can be set to the standard deviation of the time series of power consumption changes.

[0072] A3. Construct a power consumption behavior characteristic vector v= of each power line according to the power consumption characteristic value, the first statistical value and the second statistical value. .

[0073] A4. Calculate the similarity of the power consumption behavior feature vectors between all power lines, construct a power line similarity matrix, and obtain several power line groups according to a preset similarity threshold.

[0074] Specifically, please refer to Figure 2As an example shown, suppose there are 5 power lines in a power management area, and the preset similarity threshold is 0.6. The power lines reaching 0.6 are divided into a group to form a power line group. All power line pairs with similarity reaching 0.6 are found from the power line similarity matrix, and groups are constructed based on these power line pairs. The grouping results are:

[0075] Group 1: Powerline 1, Powerline 2, Powerline 5

[0076] Group 2: Powerline 3, Powerline 4

[0077] Among them, the similarity calculation of the power consumption behavior feature vectors between all power lines is performed as follows:

[0078] .

[0079] is the distance between the power consumption behavior feature vectors of the two power lines, is the similarity value between the power consumption behavior feature vectors of the two power lines.

[0080] S102, based on each power line group, obtain the power load extreme value of each power line in k time segments in the past day (the difference between the maximum and minimum power loads of all electric energy meters on the power line), and form the power load fluctuation sequence of the power line , and generate the load deviation sequence of the power line and other power lines in the power line group .

[0081] in, is the extreme value of power load in the kth time segment, It is the power load deviation value between the power line and other power lines in the power line group.

[0082] S103, based on all the power load deviation sequences of each power line in the power line group, if there is a power load deviation sequence in which the number of time segments reaching the deviation threshold is greater than , the corresponding power line is determined as the target power line (a large-scale preliminary screening is performed to screen out the power lines that may have abnormal electric energy meters as the target power lines).

[0083] The deviation threshold is preset according to actual conditions and is used to measure the degree of deviation of power loads between different power lines in the same power line group.

[0084] S104, based on each target power line, obtain the voltage deviation set of each electric energy meter and its corresponding abnormal probability value every preset period until the preset monitoring time is reached, and generate the abnormal probability value sequence of each electric energy meter. The monitoring time is greater than the preset period, and the monitoring time is an integer multiple of the preset period, for example, the integer multiple can be set to 5.

[0085] Specifically, the voltage deviation set of each electric energy meter and its corresponding abnormal probability value are obtained every preset period, including:

[0086] B1. Based on each electric energy meter, the voltage value at the current moment and the effective voltage value set of other electric energy meters on the power line are sampled every preset period.

[0087] Specifically, the effective voltage value set is: among the voltage values ​​sampled at the current moment by all electric energy meters in the power line except the electric energy meter, the maximum value and the minimum value are deleted, and the remaining voltage values ​​form the effective voltage value set.

[0088] B2. Calculate the difference between the electric energy meter and each voltage value in the effective voltage value set to form a voltage deviation set of the electric energy meter. Each element in the set corresponds to a difference. The ratio of the number of elements in the voltage deviation set that are greater than a preset reference difference threshold to the total number of elements is determined as the abnormal probability value. The preset reference difference threshold is set according to the actual situation and is used to measure the degree of deviation of the voltage values ​​of the two electric energy meters.

[0089] It should be noted that before sampling the voltage value of the next preset cycle, if the abnormal probability value of the previous cycle is greater than the first threshold, the metering chip of the electric energy meter to be detected is reset and the metering parameters are refreshed. The preset cycle can be set to 5 minutes.

[0090] S105, judging whether the abnormal probability value sequence of each electric energy meter in the target power line meets a preset condition, and obtaining a detection result of each electric energy meter in the target power line.

[0091] It can be understood that, based on each target power line, each electric energy meter therein is further self-checked using the abnormal probability value sequence to further accurately locate the position of the metering fault on the power line or to further accurately determine whether the power line has an abnormality.

[0092] Specifically, step S105 includes:

[0093] C1. When each abnormal probability value in the abnormal probability value sequence of the electric energy meter is less than the first threshold, the electric energy meter is judged to be measuring normally. The first threshold is set according to historical experience, for example, it can be set to 50%, which is used to measure the abnormal possibility of the electric energy meter to be detected on the corresponding power line.

[0094] C2. When each abnormal probability value in the abnormal probability value sequence of the electric energy meter is greater than the first threshold, the electric energy meter is determined to have a metering fault, a metering fault information flag is set and reported to the power management area.

[0095] C3. Otherwise, the difference in power consumption during the monitoring period between the electric energy meter and other electric energy meters on the power line that are judged to have normal metering is obtained as a metering error value, which is input into the corresponding pre-trained metering fault identification model, and the detection result of the electric energy meter is output, and the detection result includes normal metering and metering fault.

[0096] Among them, each power line group corresponds to a metering fault identification model. The corresponding metering fault identification model is determined according to the power line where the electric energy meter is located. The specific process of obtaining the pre-trained metering fault identification model is as follows:

[0097] In a certain power line group, obtain the power consumption of all power meters in a large amount of historical monitoring time, select the power meters with stable and normal measurement on each power line in the power line group as simulation objects, obtain the difference between the power consumption of each power meter in all historical monitoring time and the power consumption of the corresponding simulation object (the simulation object on the same power line) as the training data set, and label each training data (the power meter is measuring normally or has a measurement failure).

[0098] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0099] All power lines in the power management area are divided into several power line groups through the power consumption behavior feature similarity algorithm. The target power line with possible faults is determined according to the power load deviation sequence of each power line in each power line group, and mutual self-inspection between electric energy meters is carried out on the target power line, which reduces the workload of comprehensive self-inspection, improves the detection efficiency, avoids the data transmission peak caused by the simultaneous self-inspection of all electric energy meters, and reduces the network burden and collection resource waste in the power management area; combined with power load analysis and abnormal probability value monitoring of electric energy meters, electric energy meters with metering abnormalities can be located more accurately, and differentiated detection methods are performed according to different abnormal probability value sequence distributions. For abnormal probability value sequences that partially meet the first threshold, the metering error value of the abnormal probability value sequence and the normal electric energy meter within the monitoring period are input into the model to obtain the detection result of the electric energy meter, thereby realizing efficient and accurate detection of the electric energy meter.

[0100] Embodiment 2: In embodiment 1, when the abnormal probability value sequence of the electric energy meter does not completely meet the judgment condition of the first threshold, the corresponding metering error value is input into the metering fault identification model of the corresponding power line group (power lines of the same type) for detection and judgment. However, the proportion of the abnormal probability value sequence less than the first threshold in different situations varies greatly. The abnormal proportion is the comparison result of the electric energy meter and multiple other electric energy meters (electric energy meters of which it is unknown whether all are normal or not). In the face of metering error values ​​of electric energy meters with different proportions (the metering difference between the electric energy meter and the electric energy meter that has been detected as normal), using a unified model will inevitably result in training errors and weak generalization ability. Due to the differences in the abnormal probability value sequences between abnormal electric energy meters and other electric energy meters on the same power line in history, there are inevitably differences in the indicators required for judging the metering error value. In order to further refine and distinguish the training of the metering error value to obtain more accurate detection results, it is necessary to simultaneously consider the abnormal proportion and metering error value of the electric energy meter to be detected, and make further optimization and improvement.

[0101] In some embodiments, each power line group corresponds to a metering fault identification model, each metering fault identification model includes a plurality of sub-models, and step C3 further includes:

[0102] S201, based on a preset abnormality ratio value interval [0, ),[ , ),[ , 1], corresponding to a sub-model respectively.

[0103] The abnormal proportion degree value is specifically: the ratio of the number of abnormal probability values ​​in the abnormal probability value sequence that is greater than a first threshold (preset according to actual conditions) to the total number of abnormal probability values ​​in the abnormal probability value sequence.

[0104] S202, obtaining the abnormal proportion degree value corresponding to the abnormal probability value sequence of the electric energy meter, and locating the corresponding sub-model according to the abnormal proportion degree value interval in which the abnormal proportion degree value is located.

[0105] S203, inputting the measurement error value of the electric energy meter into the corresponding sub-model, and outputting the detection result of the electric energy meter.

[0106] In some embodiments, the training method of the sub-model corresponding to each abnormal proportion degree value interval specifically includes:

[0107] D1. In the acquired training data set, obtain the abnormal proportion degree value corresponding to the abnormal probability value sequence obtained by the electric energy meter corresponding to each training data on the power line where it is located.

[0108] D2. According to the abnormality ratio value of each training data, it is divided into corresponding abnormality ratio value intervals to obtain a sub-training set corresponding to each abnormality ratio value interval.

[0109] D3. Use the sub-training set to train and optimize the parameters of the preset network model to obtain a sub-model corresponding to each abnormal proportion value interval.

[0110] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0111] By further subdividing the metering fault identification model into several sub-models, more refined detection is performed on electric energy meters with different abnormal proportion values ​​within the monitoring period. Different sub-models are used for detection, avoiding the problems of training errors and weak generalization ability that may be caused by using a unified model, and improving the detection accuracy and generalization ability of the model. The training of the sub-model takes into account the differences in the abnormal probability value sequences between the historically abnormal electric energy meters and other electric energy meters on the same power line, so that the model has better judgment ability and adaptability to different types of metering error values.

[0112] Embodiment 3: In embodiment 1, the voltage value difference between different electric energy meters on the same power line is used to perform self-inspection to obtain the abnormal probability value sequence of the electric energy meter. However, although the power lines with similar overall power consumption patterns are divided into a group through power line grouping, different electric energy meters on the same power line may have significant differences due to the type of electricity consumption or electricity consumption habits, thereby reducing the accuracy due to inconsistent reference standards. For example, there are differences in the type and pattern of electricity consumption between different users in the same building. Then, when making a difference comparison with the voltage values ​​of other electric energy meters on the same power line, the standardization of the comparison object is extremely important. The mutual self-inspection comparison between electric energy meters with similar electricity consumption habits has higher credibility and reliability. Therefore, in order to differentiate the slight differences between different electric energy meters on the same power line (to make up for the local error of the clustering of the overall electricity consumption behavior of the power line), the selection of the reference electric energy meter in the self-inspection process is extremely critical.

[0113] In some embodiments, in step B1, the method for determining the effective voltage value set of each electric energy meter further includes:

[0114] S301, respectively calculating power usage status similarity indexes between the electric energy meter and the remaining electric energy meters on the power line.

[0115] Specifically, the calculation method of the power consumption status similarity index between two power meters is as follows:

[0116] E1. Obtain the voltage value of each electric energy meter at each sampling time in the historical time window. The historical time window is set to nearly 6 hours, and obtain a number of data points (sampling time, voltage value) arranged in chronological order.

[0117] E2. Fit all data points of the electric energy meter using a preset fitting algorithm to obtain a sampling time-voltage value curve of each electric energy meter.

[0118] Among them, the preset fitting algorithm is set to the polynomial regression method.

[0119] For example, the sampling time-voltage value curves of two electric energy meters A and B are respectively expressed as:

[0120] ;

[0121] .

[0122] Where t is the sampling time, and are the voltage values ​​of energy meters A and B at sampling time t, and are the coefficients of the polynomial regression.

[0123] E3. Calculate the power consumption status similarity index between the two energy meters according to the following formula:

[0124] .

[0125] in, and are the voltage values ​​of meters A and B at sampling time t respectively. [t1, t2] is a time interval pre-set based on the historical time window. A time interval is selected in the historical time window to calculate the similarity index of power consumption status. For example, it can be set to a time interval corresponding to nearly 2 hours.

[0126] S302: Use an electric energy meter whose power consumption status similarity index is greater than a preset status similarity threshold as a reference electric energy meter.

[0127] The preset state similarity threshold a1 is set according to actual conditions and experience, and is used to measure the similarity of power consumption states between different power meters.

[0128] S303: Remove the maximum value and the minimum value from the voltage values ​​sampled by all reference electric energy meters at the current moment to obtain a valid voltage value set of the electric energy meter.

[0129] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0130] By calculating the similarity index of power consumption status between the power meters on a power line and selecting the power meters with similar power consumption status as reference, the comparison in the self-check process is made more standardized, and the accuracy and reliability of anomaly detection are improved; by refining the selection of reference power meters, the local errors that may be caused by the clustering of the overall power consumption behavior of the power line are compensated, especially when the power consumption behaviors of different power meters on the same power line are quite different. This method is more suitable for scenarios with diverse power consumption behaviors and complex user types, such as commercial buildings, residential areas, etc., and can more accurately identify abnormal situations of power meters.

[0131] Embodiment 4: In embodiment 3, the comparison object of the self-checking electric energy meter is further limited, that is, a reference electric energy meter is selected from the remaining electric energy meters on the power line to compare the voltage values, thereby improving the accuracy and pertinence of the subsequent abnormal probability value calculation. Based on this, although the accuracy of self-check is improved by selecting electric energy meters with similar power consumption status as references, a unified preset difference threshold is still used to judge the voltage deviation. The fixed threshold method may not be able to fully reflect the actual differences between different electric energy meters, resulting in limited accuracy of abnormal detection. In order to make the detection results of the electric energy meters more reliable and accurate, the determination of the abnormal probability values ​​of the electric energy meters is further limited.

[0132] In some embodiments, in step B2, the ratio of the number of elements in the voltage deviation set that are greater than a preset difference threshold to the total number of elements is determined as the abnormal probability value, and further includes:

[0133] S401, based on the reference electric energy meter corresponding to each element in the voltage deviation set of the electric energy meter (i.e., the difference set of voltage values ​​currently sampled by the electric energy meter and each reference electric energy meter), respectively obtain the power consumption status similarity index between the electric energy meter and each reference electric energy meter.

[0134] S402: adaptively adjusting a preset difference threshold of each reference electric energy meter according to a similarity index of the electric energy meter's power usage status with each reference electric energy meter.

[0135] Specifically, step S402 includes:

[0136] According to the preset power consumption status similarity index interval-adjustment factor, the adjustment factor corresponding to the power consumption status similarity index interval of each reference electric energy meter is determined, and the preset difference threshold is multiplied by the corresponding adjustment factor to obtain the adjusted difference threshold of the reference electric energy meter, which replaces the original preset difference threshold. In this way, each reference electric energy meter corresponds to a personalized difference threshold, which can differentiate the comparison standards of different reference electric energy meters and realize the calculation of refined abnormal probability values. That is, the higher the power consumption status similarity index between electric energy meters, the greater the voltage value difference and the greater the possibility of abnormality, so the adaptability of reducing the difference threshold used for judgment.

[0137] Among them, the preset power consumption status similarity index interval-adjustment factor is specifically: [a1, a2)-1.2, [a2, a3)-1, [a3, 1]-0.8, a1 is the preset status similarity threshold, a2 and a3 are adaptively set according to the preset status similarity threshold, satisfying that a2 is less than a3, a2 is greater than a1, and a3 is less than 1.

[0138] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0139] By adaptively adjusting the difference threshold of each reference electricity meter, the actual differences between the electricity meters can be reflected more accurately, reducing misjudgments and missed judgments caused by fixed difference thresholds; each reference electricity meter corresponds to a personalized difference threshold, which makes the calculation of the abnormal probability value more refined and helps to discover more hidden abnormal problems; the difference threshold can be automatically adjusted according to similar indicators of power consumption status between the electricity meters, thereby improving the adaptability and intelligence of the system.

[0140] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting a user's power load by identifying an intelligent electric energy meter, characterized in that: include: S101, dividing all power lines in the power management area into a number of power line groups according to a preset power consumption behavior feature similarity algorithm; S102, based on each power line group, obtain the power load extreme value of each power line in k time segments in the past day to form the power load fluctuation sequence of the power line , and generate the load deviation sequence of the power line and other power lines in the power line group ; is the extreme value of power load in the kth time segment, is the power load deviation value between the power line and other power lines in the power line group, and k is the number of time segments in the past day; S103, based on all the power load deviation sequences of each power line in the power line group, if there is a power load deviation sequence in which the number of time segments reaching the deviation threshold is greater than , then the corresponding power line is determined as the target power line; S104, based on each target power line, obtaining the voltage deviation set of each electric energy meter and its corresponding abnormal probability value every preset period until the preset monitoring time is reached, and generating an abnormal probability value sequence of each electric energy meter; the monitoring time is greater than the preset period, and the monitoring time is an integral multiple of the preset period; S105, judging whether the abnormal probability value sequence of each electric energy meter in the target power line meets a preset condition, and obtaining a detection result of each electric energy meter in the target power line.

2. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 1, characterized in that: The preset electricity consumption behavior feature similarity algorithm specifically includes: A1. Based on each power line, obtain its power consumption in a preset historical time period and each time unit, and calculate the power consumption characteristic value of the power line according to the following formula: Among them, F is the power consumption characteristic value of the corresponding power line in the preset historical time period, is the total power consumption of the ith electric energy meter on the power line in the preset historical time period, n is the number of electric energy meters, m is the total number of time units in the preset historical time period, is the total power consumption on the power line in the jth time unit, is the average power consumption of all time units, and are the preset weight coefficient values ​​respectively; A2. Based on the rate of change of power consumption between adjacent time units , t=1,2,...,m-1, and obtain the time series of power consumption changes of the corresponding power line in the preset historical time period , generate the first statistical value of the power consumption change time series and the second statistic ; A3. Construct a power consumption behavior characteristic vector v= of each power line according to the power consumption characteristic value, the first statistical value and the second statistical value. ; A4. Calculate the similarity of the power consumption behavior feature vectors between all power lines, construct a power line similarity matrix, and obtain several power line groups according to a preset similarity threshold.

3. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 1, characterized in that: In S104, the voltage deviation set of each electric energy meter and its corresponding abnormal probability value are obtained every preset period, including: B1. Based on each electric energy meter, the voltage value at the current moment and the effective voltage value set of other electric energy meters on the power line are sampled every preset period; wherein the effective voltage value set is specifically: among the voltage values ​​sampled at the current moment by all electric energy meters in the power line except the electric energy meter, the maximum value and the minimum value are deleted, and the remaining voltage values ​​form an effective voltage value set; B2. The difference between the electric energy meter and each voltage value in the effective voltage value set is calculated respectively to form a voltage deviation set of the electric energy meter, each element in the set corresponds to a difference, and the ratio of the number of elements in the voltage deviation set that are greater than the preset reference difference threshold to the total number of elements is determined as the abnormal probability value; Before sampling the voltage value of the next preset cycle, if the abnormal probability value of the previous cycle is greater than the first threshold, the metering chip of the electric energy meter to be detected is reset and the metering parameters are refreshed.

4. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 3, characterized in that: The S105 specifically includes: C1. When each abnormal probability value in the abnormal probability value sequence of the electric energy meter is less than the first threshold value, it is determined that the electric energy meter is measuring normally; C2. When each abnormal probability value in the abnormal probability value sequence of the electric energy meter is greater than the first threshold value, the electric energy meter is determined to have a metering fault, a metering fault information flag is set, and the metering fault information flag is reported to the power management area; C3. Otherwise, the difference in power consumption during the monitoring period between the electric energy meter and other electric energy meters on the power line that are judged to have normal metering is obtained as a metering error value, which is input into the corresponding pre-trained metering fault identification model, and the detection result of the electric energy meter is output, and the detection result includes normal metering and metering fault.

5. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 4, characterized in that: Each power line group corresponds to a metering fault identification model. The corresponding metering fault identification model is determined according to the power line where the electric energy meter is located. The specific process of obtaining the pre-trained metering fault identification model is as follows: In a certain power line group, a large amount of power consumption of all power meters within a historical monitoring period is obtained, and the power meters with stable and normal measurement on each power line in the power line group are selected as simulation objects. The difference between the power consumption of each power meter within all historical monitoring periods and the power consumption of the corresponding simulation object is obtained as a training data set, and each training data is labeled.

6. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 5, characterized in that: The metering fault identification model includes several sub-models, and step C3 also includes: S201, based on a preset abnormality ratio value interval [0, ),[ , ),[ , 1], each corresponding to a sub-model; wherein the abnormal proportion degree value is specifically: the ratio of the number of abnormal probability values ​​in the abnormal probability value sequence greater than the first threshold to the total number of abnormal probability values ​​in the abnormal probability value sequence; S202, obtaining the abnormal proportion degree value corresponding to the abnormal probability value sequence of the electric energy meter, and locating the corresponding sub-model according to the abnormal proportion degree value interval in which the abnormal proportion degree value is located; S203, inputting the measurement error value of the electric energy meter into the corresponding sub-model, and outputting the detection result of the electric energy meter; The training method of the sub-model corresponding to each abnormal proportion value interval specifically includes: D1. In the acquired training data set, obtain the abnormal proportion degree value corresponding to the abnormal probability value sequence obtained by the electric energy meter corresponding to each training data on the power line where it is located; D2. According to the abnormality ratio value of each training data, divide it into corresponding abnormality ratio value intervals to obtain a sub-training set corresponding to each abnormality ratio value interval; D3. Use the sub-training set to train and optimize the parameters of the preset network model to obtain a sub-model corresponding to each abnormal proportion value interval.

7. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 3, characterized in that: In the above B1, the method for determining the effective voltage value set of each electric energy meter further includes: S301, respectively calculating the power usage status similarity index between the electric energy meter and the remaining electric energy meters on the power line; S302, using an electric energy meter whose power consumption status similarity index is greater than a preset status similarity threshold as a reference electric energy meter; S303: Remove the maximum value and the minimum value from the voltage values ​​sampled by all reference electric energy meters at the current moment to obtain a valid voltage value set of the electric energy meter.

8. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 7, characterized in that: The calculation method of the power consumption status similarity index between two electric energy meters is as follows: E1. Obtain the voltage value of each electric energy meter at each sampling time in the historical time window. The historical time window is set to nearly 6 hours, and obtain several data points (sampling time, voltage value) arranged in chronological order. E2. Fit all data points of the electric energy meter using a preset fitting algorithm to obtain a sampling time-voltage value curve of each electric energy meter; E3. Based on the sampling time-voltage value curve of each electric energy meter, the power consumption status similarity index between the two electric energy meters is calculated according to the following formula: in, is the similarity index of power consumption status between two electric energy meters, and are the voltage values ​​of the energy meters A and B at the sampling time t respectively, [t1, t2] is the time interval pre-set based on the historical time window, and a time interval is selected in the historical time window to calculate the similarity index of the power consumption status.

9. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 7, characterized in that: In B2, the ratio of the number of elements in the voltage deviation set that are greater than a preset difference threshold to the total number of elements is determined as the abnormal probability value, and further includes: S401, based on the reference electric energy meter corresponding to each element in the voltage deviation set of the electric energy meter, respectively obtain a power consumption status similarity index between the electric energy meter and each reference electric energy meter; S402: adaptively adjusting a preset difference threshold of each reference electric energy meter according to a similarity index of the electric energy meter's power usage status with each reference electric energy meter.

10. The method for detecting a user's power load by identifying an intelligent electric energy meter according to claim 9, characterized in that: The S402 specifically includes: According to the preset power consumption status similarity index interval-adjustment factor, determine the adjustment factor corresponding to the power consumption status similarity index interval of each reference electric energy meter, multiply the preset difference threshold by the corresponding adjustment factor, obtain the adjusted difference threshold of the reference electric energy meter, and replace the original preset difference threshold; The preset power consumption status similarity index interval-adjustment factor is specifically: [a1, a2)-1.2, [a2, a3)-1, [a3, 1]-0.8, a1 is the preset status similarity threshold, a2 and a3 are adaptively set according to the preset status similarity threshold to satisfy a2 less than a3, a2 greater than a1, and a3 less than 1.

Images