Remote anti-electricity-stealing monitoring alarm method based on artificial intelligence
By combining multi-source data acquisition and artificial intelligence analysis with time series and harmonic analysis, the inefficiency and misjudgment problems of traditional anti-electricity theft methods have been solved, achieving efficient and accurate remote electricity theft monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-03-31
AI Technical Summary
Traditional methods of combating electricity theft rely on manual inspections, which consume a lot of manpower and resources and make it difficult to achieve high-frequency, full-coverage inspections. Traditional electricity meter data collection has a low frequency and few dimensions, making it impossible to capture key characteristics of electricity theft. Furthermore, there is a lack of effective data cleaning and analysis methods, making it difficult to identify electricity theft.
By employing multi-source data acquisition, smart meters, leakage current sensors, and a distributed fiber optic temperature measurement system, combined with time series analysis and artificial intelligence, and through autoregressive integral moving average models and harmonic analysis, remote real-time monitoring is achieved to identify electricity theft.
It enables remote real-time monitoring without the need for manual inspections, improves the efficiency of anti-electricity theft, collects multi-dimensional data, enhances the accuracy and coverage of electricity theft identification, and ensures the reliability of analysis results.
Smart Images
Figure CN121762918A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power monitoring technology, and more specifically, to an artificial intelligence-based remote anti-electricity theft monitoring and alarm method. Background Technology
[0002] In the development of the power industry, electricity theft has always been a problem that plagues power companies. It not only causes huge economic losses, but also poses a serious threat to the safe and stable operation of the power system. Traditional anti-theft methods mainly rely on manual inspections. Staff regularly go to the site to check the status of electricity meters and lines, and at the same time combine simple electricity data analysis, such as comparing the differences in electricity consumption of users at different times, to determine whether there is any suspicion of electricity theft.
[0003] With the increase in the number of electricity users and the increasing complexity of electricity consumption scenarios, some anti-electricity theft technologies based on data monitoring have begun to be applied. These technologies collect electricity consumption data by installing ordinary electricity meters, perform simple threshold judgments, and issue warnings when the data exceeds the set range.
[0004] However, in practical use, it still has some drawbacks. For example, existing technologies rely on manual inspections, which require a lot of manpower, resources, and time. Especially in remote areas or densely populated areas, it is difficult to achieve high-frequency, full-coverage inspections, making it difficult to detect electricity theft in a timely manner. Traditional electricity meters collect data at low frequency and with few dimensions, only obtaining basic electricity data. They cannot capture key characteristics reflecting electricity theft, such as current harmonics and voltage distortion, making it difficult to identify covert electricity theft methods. Furthermore, traditional methods lack effective cleaning, standardization, and analysis methods, making it difficult to remove noise and outliers from the judgment results, thus affecting the accuracy of anti-electricity theft measures. Summary of the Invention
[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a remote anti-electricity theft monitoring and alarm method based on artificial intelligence, which solves the problems mentioned in the background art through the following scheme.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a remote anti-electricity theft monitoring and alarm method based on artificial intelligence, comprising: S1: Multi-source data acquisition: Acquire multi-source data based on a preset data acquisition frequency; the multi-source data includes user electricity consumption data and power grid operation data; S2: First-time data processing: The data collected in S1 is processed for outliers, noise, and missing data. S3: Secondary data processing: Standardize the numerical data after primary processing and encode the non-numerical data; S4: Time series analysis-based electricity consumption anomaly marking: Using time series analysis, an autoregressive integral moving average model is constructed based on the user's historical electricity consumption data. The constructed autoregressive integral moving average model is used to predict the user's electricity consumption in future periods. The real-time collected power is compared with the predicted power to determine whether the user has abnormal electricity consumption and to mark it. S5: Electricity Theft Detection: Using artificial intelligence, a theoretical calculation model of the user's power consumption is established for users marked as having abnormal electricity consumption. The power deviation is calculated, and features are extracted from the collected current harmonics and voltage harmonics data. The total harmonic distortion rate of the current and the total harmonic distortion rate of the voltage are calculated and compared with preset thresholds to determine whether the user has engaged in electricity theft.
[0007] Preferably, the method for collecting user electricity consumption data is as follows: High-precision smart meters are used to collect real-time data on current, voltage, power, energy consumption, current harmonics, and voltage harmonics from electricity users. Install a leakage current sensor in the user's distribution box to monitor for leakage current in real time. The threshold for leakage current is set to 30mA. Once the threshold is exceeded, the data is recorded and uploaded immediately. The method for collecting the power grid operation data is as follows: Through the power communication network, bus voltage, line current, and power factor data are obtained from substations and distribution stations. At the same time, zero-sequence current and negative-sequence current data of the power grid are collected. A distributed fiber optic temperature measurement system is used to monitor the temperature of high-voltage transmission lines and low-voltage distribution lines. A fiber optic temperature measurement point is set up every 100 meters to collect line temperature data in real time.
[0008] The preferred frequency for collecting user electricity consumption data is as follows: For ordinary residential users, current, voltage, power, and power consumption data are collected every 15 minutes, and current harmonics, voltage harmonics, and leakage current data are collected every hour. For commercial users, current, voltage, power, and energy data are collected every 5 minutes, and current harmonics, voltage harmonics, and leakage current data are collected every 30 minutes. For industrial users, current, voltage, power, and energy data are collected every 2 minutes, and current harmonics, voltage harmonics, and leakage current data are collected every 15 minutes. The power grid operation data is collected at the following frequencies: Bus voltage, line current, and power factor data of substations and distribution stations are collected at a high frequency once per second. Zero-sequence current and negative-sequence current data are collected every 10 seconds. The distributed fiber optic temperature measurement system collects line temperature data every 30 seconds.
[0009] Preferably, the outlier handling method is as follows: For current and voltage data, a reasonable threshold range is set based on the statistical 3σ principle. That is, the mean μ and standard deviation σ of the current or voltage data within 1 hour are calculated first. When the data exceeds μ+3σ or is lower than μ-3σ, it is judged as an outlier. For outliers, if it is a single outlier, it is corrected by linear interpolation between two adjacent points. If it is multiple consecutive outliers, it is filled by a time series prediction model based on historical data. For power data, anomaly detection is performed by combining user type and historical electricity consumption data; The method for processing the noise data is as follows: For high-frequency noise in current and voltage data, a low-pass filter is used for filtering.
[0010] For random noise in the power data, a median filtering algorithm is used to process it. Taking a certain moment as the center, five power data points before and after are taken to form a data window. The data in the data window are sorted and the median value is taken as the power value at that moment. The method for handling missing data is as follows: For data with no more than 3 consecutive missing data points, linear interpolation is used to fill the missing data. For data with more than 3 consecutive missing data points, the K-nearest neighbor algorithm is used. First, based on characteristics such as user type and electricity usage time, the K samples most similar to the time of the missing data are selected from historical data. Then, the missing data is filled in based on the average values of current, voltage, power and other data of these K samples.
[0011] Preferably, the standardization process is as follows: For physical quantities such as current, voltage, power, and electrical charge, a normalization method is used to map them to the [0,1] interval. The specific formula is: x norm = (x x min ) / (x max x min ), where x is the original data, x min and x max These represent the minimum and maximum values of this type of data within a certain time period; For environmental data such as temperature and humidity, the same normalization method is used to map them to the [0,1] interval; The data is encoded as follows: For user type data, one-hot encoding is used for conversion, and the user type is represented as a three-dimensional vector, with residential users as [1,0,0], commercial users as [0,1,0], and industrial users as [0,0,1]. For equipment operating status data, one-hot encoding is used to represent the equipment operating status as a three-dimensional vector, with power-on being [1,0,0], power-off being [0,1,0], and fault being [0,0,1].
[0012] Preferably, the method for constructing the autoregressive integral moving average model is as follows: Using a daily cycle, this study analyzes the variation patterns of a user's electricity consumption at different times of the day, constructing an autoregressive integral moving average model. First, the stationarity of the user's historical electricity consumption data is tested. Then, the model parameters, including the autoregressive order, differencing order, and moving average order, are determined using the autocorrelation function and partial autocorrelation function. By observing the autocorrelation and partial autocorrelation function graphs, the autoregressive order, differencing order, and moving average order of the model for this residential user are determined. Finally, the constructed model is used to predict the user's electricity consumption P in future periods. 预测 The collected power P 实际 With predicted power P 预测 When comparing, |P 实际 P 预测 When |>σ×k, it is marked as an abnormal power consumption situation, where σ is the standard deviation of the prediction error and k is a coefficient determined based on user type and historical data statistics.
[0013] Preferably, the theoretical calculation model is as follows: For three-phase four-wire users, the theoretical power P 理论 =√3UIcosφ, where U is the line voltage, I is the line current, and cosφ is the power factor. Based on real-time collected voltage, current, and power factor data, for a single-phase user, the theoretical power P... 理论 =UIcosφ.
[0014] Preferably, the power deviation is calculated as follows: A power deviation threshold ΔP is set, which is 15% for residential users, 10% for commercial users, and 8% for industrial users. When |P actually... When P-theory |> ΔP×P-theory, it is determined that the user has engaged in electricity theft.
[0015] Preferably, the calculation methods for the total harmonic distortion of the current and the total harmonic distortion of the voltage are as follows: ; ; Among them, In and U n Let I1 and U1 be the nth harmonic current and voltage, respectively, and THD be the fundamental current and voltage. I Total Harmonic Distortion (THD) refers to the total harmonic distortion of the current. U Total harmonic distortion (THD) of voltage; Preferably, the method for determining electricity theft is as follows: For residential users, the total harmonic distortion (THD) threshold is set to 8% for current and 5% for voltage; for commercial users, the THD threshold is set to 6% for current and 4% for voltage; for industrial users, the THD threshold is set to 5% for current and 3% for voltage. I or THD U If the threshold is exceeded, it is determined that the user has engaged in electricity theft.
[0016] The technical effects and advantages of this invention are as follows: 1. This invention achieves remote real-time monitoring through multi-source automated data acquisition combined with intelligent algorithms. It can continuously monitor the user's electricity status without manual on-site inspection, greatly improving the efficiency of anti-electricity theft work. It is especially suitable for large-scale user groups and complex geographical areas. 2. This invention uses high-precision smart meters, leakage current sensors, and distributed fiber optic temperature measurement systems to collect multi-dimensional data such as current, voltage, power, harmonics, and line temperature, covering various characteristic changes that may be caused by electricity theft, and providing data support for accurate identification. 3. This invention introduces time series analysis to capture the temporal characteristics of users' electricity consumption patterns, and combines power balance models and harmonic analysis models for multi-dimensional verification. It fully considers user types and sets differentiated thresholds, which significantly improves the accuracy of judgment. 4. This invention effectively improves data quality and ensures the reliability of analysis results by handling outliers, filtering noise, and standardizing data, thus avoiding misjudgments caused by data interference. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation
[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0019] As attached Figure 1The method for remote anti-electricity theft monitoring and alarm based on artificial intelligence, as shown, includes: S1: Multi-source data acquisition: Acquire multi-source data based on a preset data acquisition frequency; The multi-source data includes user electricity consumption data and power grid operation data; The method for collecting user electricity consumption data is as follows: High-precision smart meters are used to collect real-time data on current, voltage, power, electricity consumption, current harmonics, and voltage harmonics from electricity users. For example, the high-speed sampling chip inside the smart meter samples the current and voltage signals at a frequency of 1,000 times per second, and then uses the fast Fourier transform algorithm to calculate the content of each harmonic.
[0020] Install a leakage current sensor in the user's distribution box to monitor for leakage current in real time. The threshold for leakage current is set to 30mA. Once the threshold is exceeded, the data is immediately recorded and uploaded.
[0021] The method for collecting the power grid operation data is as follows: Through the power communication network, data on bus voltage, line current, and power factor are obtained from substations and distribution stations. At the same time, zero-sequence current and negative-sequence current data of the power grid are collected. These data are of great significance for determining whether there is an asymmetrical fault in the power grid caused by electricity theft. For example, zero-sequence current transformers are installed at the outgoing terminals of substations to accurately measure the zero-sequence current.
[0022] A distributed fiber optic temperature measurement system is used to monitor the temperature of high-voltage transmission lines and low-voltage distribution lines. Because electricity theft causes local resistance to increase, it leads to abnormal temperature rise. A fiber optic temperature measurement point is set up every 100 meters to collect line temperature data in real time.
[0023] The user electricity consumption data is collected at the following frequencies: For ordinary residential users, considering that their electricity consumption behavior is relatively regular and fluctuates little, current, voltage, power and power data are collected every 15 minutes, and current harmonics, voltage harmonics and leakage current data are collected every hour.
[0024] For commercial users, whose electricity consumption behavior is greatly affected by factors such as business hours and business activities, current, voltage, power and power data are collected every 5 minutes, and current harmonics, voltage harmonics and leakage current data are collected every 30 minutes.
[0025] For industrial users, their production processes have high requirements for power stability and the power load changes frequently. Current, voltage, power and power data are collected every 2 minutes, and current harmonics, voltage harmonics and leakage current data are collected every 15 minutes.
[0026] The power grid operation data is collected at the following frequencies: The bus voltage, line current, and power factor data of substations and distribution stations are collected at a high frequency once per second to ensure timely capture of dynamic changes in the power grid.
[0027] Zero-sequence current and negative-sequence current data are collected every 10 seconds, as they are mainly used to monitor power grid faults.
[0028] The distributed fiber optic temperature measurement system collects line temperature data every 30 seconds.
[0029] S2: First-time data processing: The data collected in S1 is processed for outliers, noise, and missing data. The method for handling outliers is as follows: For current and voltage data, a reasonable threshold range is set using the statistical 3σ principle. First, the mean μ and standard deviation σ of the current or voltage data within one hour are calculated. When the data exceeds μ+3σ or is lower than μ-3σ, it is judged as an outlier. For outliers, if it is a single outlier, it is corrected by linear interpolation between two adjacent points. If it is multiple consecutive outliers, it is filled using a time series prediction model based on historical data. For example, for the voltage data of a residential user, the mean of one hour is calculated to be 220V and the standard deviation is 2V. When a voltage value of 250V appears, it is judged as an outlier. It is corrected to 220V by linear interpolation between two adjacent points, 218V and 222V.
[0030] For power data, anomaly detection is performed by combining user type and historical electricity consumption data. For example, if a commercial user's historical average power is 50kW during normal business hours, and the power suddenly drops to 10kW at a certain moment and lasts for more than 10 minutes, it is judged as an anomaly. At this time, the power grid operation data of the user's area and the electricity consumption of other users are further checked. If the power grid is operating normally and the electricity consumption of other users in the surrounding area is normal, the power data of the user in the same period of the past week is used to correct the anomaly using a weighted average method.
[0031] The method for processing the noise data is as follows: For high-frequency noise in current and voltage data, a low-pass filter is used for filtering. For example, a Butterworth low-pass filter with a cutoff frequency of 50Hz can be designed to remove high-frequency noise interference in the power signal, making the acquired current and voltage signals smoother.
[0032] To address random noise in the power data, a median filtering algorithm is used. Taking a certain moment as the center, five power data points before and after the center are used to form a data window. The data within the data window are sorted, and the median value is taken as the power value at that moment, thereby effectively removing random noise.
[0033] The method for handling missing data is as follows: For data with no more than 3 consecutive missing data points, linear interpolation is used to fill the gaps. For example, if a resident's current data is missing at 10:00, 10:15, and 10:30, the current values at 9:45 and 10:45 are used for linear interpolation to calculate the current values at these three times.
[0034] For data with more than 3 consecutive missing data points, the K-nearest neighbor algorithm is used. First, based on characteristics such as user type and electricity usage time, the K samples most similar to the time of the missing data are selected from historical data. Then, the missing data is filled in based on the average values of current, voltage, power and other data of these K samples.
[0035] S3: Secondary data processing: Standardize the numerical data after primary processing and encode the non-numerical data; The standardization process is as follows: For physical quantities such as current, voltage, power, and electrical charge, a normalization method is used to map them to the [0,1] interval. The specific formula is: x norm = (x x min ) / (x max x min ), where x is the original data, x min and x max These are the minimum and maximum values of this type of data within a certain time period. For example, if a residential user's current has a minimum of 0A and a maximum of 50A within a month, and the current is 20A at a certain moment, the normalized value is 20 / 50 = 0.4.
[0036] For environmental data such as temperature and humidity, the normalization method is also used to map them to the [0,1] interval. For example, the minimum temperature in a certain region within a month is 10℃ and the maximum temperature is 35℃. When the temperature is 20℃ at a certain moment, the normalized value is 10 / 25=0.4.
[0037] The data is encoded as follows: For user type data, one-hot encoding is used for conversion, and the user type is represented as a three-dimensional vector, with residential users as [1,0,0], commercial users as [0,1,0], and industrial users as [0,0,1]. For equipment operating status data, one-hot encoding is used to represent the equipment operating status as a three-dimensional vector, with power-on being [1,0,0], power-off being [0,1,0], and fault being [0,0,1].
[0038] S4: Time series analysis-based electricity consumption anomaly marking: Using time series analysis, an autoregressive integral moving average model is constructed based on the user's historical electricity consumption data. The constructed autoregressive integral moving average model is used to predict the user's electricity consumption in future periods. The real-time collected power is compared with the predicted power to determine whether the user has abnormal electricity consumption and to mark it. It should be noted that the method for constructing the autoregressive integral moving average model is as follows: Using a daily cycle, this study analyzes the variation patterns of electricity consumption at different times of the day, constructing an autoregressive integral moving average (ARM) model. First, the historical electricity consumption data of the user is tested for stationarity. If the data is not stationary, it is stabilized through differencing. For example, after performing first-order differencing on the electricity consumption data of a residential user, if the ADF test value is less than the critical value, it indicates that the data has stabilized. Then, the autocorrelation function and partial autocorrelation function are used to determine the model parameters p (autoregressive order), d (difference order), and q (moving average order). By observing the autocorrelation and partial autocorrelation function graphs, the autoregressive order, differencing order, and moving average order of the model for this residential user are determined. The constructed model is then used to predict the user's electricity consumption P in future periods. 预测 For example, after training a model using the electricity consumption data of a residential user over the past week, predicting their electricity consumption between 7:00 and 8:00 AM the following morning, the collected power P... 实际 With predicted power P 预测 When comparing, |P 实际 P 预测 When |> σ×k, it is marked as an abnormal power consumption situation, where σ is the standard deviation of the prediction error, and k is a coefficient determined based on user type and historical data statistics (k=2 for residential users; k=1.5 for commercial users; k=1.2 for industrial users). For example, if the standard deviation of the prediction error for a commercial user is 0.5kW and k=1.5, when the deviation between the actual power and the predicted power exceeds 0.5×1.5=0.75kW, it is marked as an abnormal power consumption situation. S5: Electricity Theft Detection: Using artificial intelligence, a theoretical calculation model of the user's power consumption is established for users marked as having abnormal electricity consumption. The power deviation is calculated. At the same time, the features of the collected current harmonics and voltage harmonics data are extracted, and the total harmonic distortion rate of the current and the total harmonic distortion rate of the voltage are calculated. They are then compared with preset thresholds to determine whether the user has engaged in electricity theft. The theoretical calculation model is as follows: For three-phase four-wire users, the theoretical power P 理论=√3UIcosφ, where U is the line voltage, I is the line current, and cosφ is the power factor. The theoretical power is calculated using real-time collected voltage, current, and power factor data. For example, for a three-phase four-wire commercial user, if the real-time collected data shows a line voltage U = 380V, a line current I = 50A, and a power factor cosφ = 0.9, then the theoretical power P is... 理论 =√3×380×50×0.9≈28.5kW, for single-phase users, the theoretical power P 理论 =UIcosφ, for example, a residential user has a real-time voltage U=220V, current I=10A, and power factor cosφ=0.95, then the theoretical power P 理论 =220×10×0.95=2.09kW; The power deviation is calculated as follows: A power deviation threshold ΔP is set. This threshold varies depending on the type of user: 15% for residential users, 10% for commercial users, and 8% for industrial users. When |P 实际 P 理论 |>ΔP×P 理论 When determining that a user has engaged in electricity theft, for example, an industrial user with a theoretical power of 100kW but an actual power of 85kW has a power deviation of |85kW. 100 | / 100 = 15%, which exceeds the set threshold of 8%, so the user is judged to have engaged in electricity theft. The calculation methods for the total harmonic distortion of current and the total harmonic distortion of voltage are as follows: ; ; Among them, I n and U n Let I1 and U1 be the nth harmonic current and voltage, respectively, and THD be the fundamental current and voltage. I Total Harmonic Distortion (THD) refers to the total harmonic distortion of the current. U Total harmonic distortion (THD) of voltage; It should be further explained that the method for judging the aforementioned electricity theft is as follows: For residential users, the total harmonic distortion (THD) threshold is set to 8% for current and 5% for voltage; for commercial users, the THD threshold is set to 6% for current and 4% for voltage; for industrial users, the THD threshold is set to 5% for current and 3% for voltage. I or THD U If the threshold is exceeded, it is determined that the user has engaged in electricity theft.
[0039] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. An artificial intelligence-based remote anti-electricity-theft monitoring and alarm method, characterized in that, Comprise: S1: multi-source data acquisition: based on the preset data acquisition frequency acquisition multi-source data; the multi-source data includes user power consumption data and power grid operation data; S2: data processing: the data collected in S1 is processed for abnormal value, noise data and missing data; S3: data processing: the data processing is standardized, and the non numerical data is processed; S4: power consumption anomaly marking based on time series analysis: using time series analysis method, according to the historical power consumption data of user, the autoregressive integral moving average model is constructed, the power consumption of user in future period is predicted by the constructed autoregressive integral moving average model, the real-time collected power is compared with the predicted power, whether the user is abnormal power consumption is judged and marked; S5: electricity stealing identification: through artificial intelligence, the user power consumption power of the user marked as abnormal power consumption is calculated, the power deviation is calculated, the current harmonic and voltage harmonic data collected are extracted, the current total harmonic distortion rate and voltage total harmonic distortion rate are calculated, and whether the user exists electricity stealing behavior is judged.
2. The remote electricity stealing monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The collection method of the user power consumption data is as follows: Real-time current, voltage, power, power, current harmonic, voltage harmonic data of power users are collected by using high-precision intelligent electric meter; Install leakage sensor in user distribution box, monitor whether there is leakage current in real time, the threshold of leakage current is set to 30mA, once the threshold is exceeded, record and upload data immediately; The collection method of the power grid operation data is as follows: Through power communication network, bus voltage, line current, power factor data of transformer substation and distribution station are obtained, at the same time, zero sequence current, negative sequence current data of power grid are collected; The temperature of high-voltage transmission line and low-voltage distribution line is monitored by using distributed optical fiber temperature measurement system, an optical fiber temperature measurement point is set every 100 meters, and line temperature data is collected in real time.
3. The remote electricity stealing monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The collection frequency of the user power consumption data is as follows: For ordinary residential users, current, voltage, power, power data is collected every 15 minutes, current harmonic, voltage harmonic, leakage current data is collected every hour; For commercial users, current, voltage, power, power data is collected every 5 minutes, current harmonic, voltage harmonic, leakage current data is collected every 30 minutes; For industrial users, current, voltage, power, power data is collected every 2 minutes, current harmonic, voltage harmonic, leakage current data is collected every 15 minutes; The collection frequency of the power grid operation data is as follows: The bus voltage, line current, power factor data of transformer substation and distribution station are collected once every second; Zero sequence current, negative sequence current data is collected every 10 seconds; The line temperature data collected by the distributed optical fiber temperature measurement system is collected every 30 seconds.
4. The remote anti-electricity-stealing monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The construction method of the autoregressive integral moving average model is as follows: Using a daily cycle, this study analyzes the variation patterns of a user's electricity consumption at different times of the day, constructing an autoregressive integral moving average model. First, the stationarity of the user's historical electricity consumption data is tested. Then, the model parameters, including the autoregressive order, differencing order, and moving average order, are determined using the autocorrelation function and partial autocorrelation function. By observing the autocorrelation and partial autocorrelation function graphs, the autoregressive order, differencing order, and moving average order of the model for this residential user are determined. Finally, the constructed model is used to predict the user's electricity consumption P in future periods. 预测 The collected power P 实际 With predicted power P 预测 When comparing, |P 实际 P 预测 When |>σ×k, it is marked as an abnormal power consumption situation, where σ is the standard deviation of the prediction error and k is a coefficient determined based on user type and historical data statistics.
5. The remote electricity theft monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The theoretical calculation model is as follows: For three-phase four-wire users, the theoretical power P 理论 =√3UIcosφ, where U is the line voltage, I is the line current, and cosφ is the power factor. For single-phase users, the theoretical power P 理论 =UIcosφ.
6. The remote electricity theft monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The calculation method of the power deviation is as follows: setting a power deviation threshold ΔP, setting the power deviation threshold as 15% for residential users, 10% for commercial users, and 8% for industrial users, when ∣Pactual Ptheoretical > ΔP × Ptheoretical, it is judged that the user has electricity stealing behavior.
7. The remote electricity theft monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The calculation method of the current total harmonic distortion rate and voltage total harmonic distortion rate is as follows: ； ； wherein, wherein I n and U n are the nth harmonic current and voltage, respectively, and I1and U1are the fundamental current and voltage, respectively, and THD I denotes the total harmonic distortion of the current, THD U denotes the total harmonic distortion of the voltage.
8. The remote electricity theft monitoring and alarm method based on artificial intelligence according to claim 1, characterized in that: The judgment method of the electricity stealing behavior is as follows: for the resident user, the current total harmonic distortion rate threshold is set to 8%, and the voltage total harmonic distortion rate threshold is set to 5%; for the commercial user, the current total harmonic distortion rate threshold is set to 6%, and the voltage total harmonic distortion rate threshold is set to 4%; for the industrial user, the current total harmonic distortion rate threshold is set to 5%, and the voltage total harmonic distortion rate threshold is set to 3%, when THD I or THD U exceeds the corresponding threshold, it is judged that the user has the electricity stealing behavior.