Electricity stealing detection method and system, electric energy meter and storage medium

Abstract

The invention discloses an electricity stealing detection method and system, an electric energy meter and a storage medium, and the method comprises the following steps: respectively obtaining the current values of a live line and a zero line in a meter, calculating the current difference value of the zero line and the live line, and calculating a first confidence coefficient C1 according to the current difference value and a set difference value threshold value; disconnecting the live wire loop switch in the meter, obtaining a load side voltage value of the live wire loop switch in the meter, and calculating a second confidence coefficient C2 according to the voltage value and a set voltage threshold value; and fusing the first confidence coefficient C1 and the second confidence coefficient C2 to obtain a comprehensive confidence coefficient C, comparing the comprehensive confidence coefficient C with a set electricity stealing judgment threshold value, and outputting electricity stealing judgment according to a comparison result. According to the invention, normal current fluctuation or load imbalance can be ignored, the misjudgment and missed judgment probability of electricity larceny detection outside the electric energy meter is reduced, and the accuracy of electricity larceny detection is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electricity metering detection, and in particular relates to an electricity theft detection method, system, electric energy meter and storage medium. Background Art

[0002] With the rapid development of society, the demand for electricity resources is increasing, but it is also accompanied by waste and theft of electricity resources. Some unscrupulous residents steal electricity by privately modifying or damaging electricity meters, or use magnets, electromagnetic interference instruments, etc. to interfere with smart meters, making them unable to work normally and achieving the purpose of stealing electricity. However, current smart meters usually lack anti-theft design and cannot effectively prevent illegal theft of electricity.

[0003] The disclosure discloses an anti-electricity theft smart meter and a control method, the disclosure includes: a current sensor for detecting current information in a circuit in real time; a voltage sensor for detecting voltage information in a circuit in real time; a temperature sensor for detecting temperature information inside the smart meter in real time; an anti-interference module for detecting and suppressing electromagnetic interference signals; a wireless communication module for sending data information to a power grid management center and receiving instructions from the power grid center; a data storage module for storing operating data and operating instructions of the smart meter; and a controller capable of adjusting the working state of the smart meter according to abnormal current, voltage, temperature and electromagnetic interference information detected, and sending alarm information to the power grid management center, and capable of adjusting the working state of the smart meter according to the instructions of the power grid management center.

[0004] The above scheme solves the partial anti-electricity theft detection inside the meter, but there are deficiencies in the detection of electricity theft outside the meter. To this end, the Chinese invention patent application with publication number CN103558431A discloses a remote anti-electricity theft device and an anti-electricity theft method. The anti-electricity theft device of the scheme includes a phase current acquisition unit, a user current acquisition unit and an electricity theft judgment unit. The phase current acquisition unit and the user current acquisition unit are both communicatively connected to the electricity theft judgment unit; the phase current acquisition unit is used to collect the phase current value on the output phase line of the power supply unit, and output the collected phase current value to the electricity theft judgment unit; the user current acquisition unit is used to collect the sub-current values ​​of all users supplied by the output phase line of the power supply unit, and output the sub-current values ​​of each user to the electricity theft judgment unit; the electricity theft judgment unit judges whether electricity theft occurs on the power supply line of the power supply unit based on the comparison between the phase current value and the sum of the sub-current values.

[0005] The above solution solves the problem of electricity theft detection outside the electricity meter, but it also has some shortcomings: for example, relying solely on phase current and user current to judge electricity theft can be easily misled by normal current fluctuations or load imbalance, leading to misjudgment or missed judgment; it is sensitive to specific load types, especially nonlinear loads or intermittent loads; it is difficult to detect small-scale or intermittent electricity theft in a timely manner. Summary of the invention

[0006] The present invention provides a method and system for detecting electricity theft, an electric energy meter and a storage medium, aiming to solve the problems of misjudgment or missed judgment, sensitivity to nonlinear types and insensitivity to small-scale or intermittent electricity theft in the prior art when detecting electricity theft outside the electric energy meter.

[0007] In order to solve the above technical problems, a first aspect of the present invention provides a method for detecting electricity theft, comprising the following steps:

[0008] The current values ​​of the live wire and the neutral wire in the meter are respectively obtained, the current difference between the live wire and the neutral wire is calculated, and the first confidence level C1 is calculated according to the current difference and the set difference threshold;

[0009] Disconnect the live wire loop switch in the meter, obtain the voltage value on the load side of the live wire loop switch in the meter, and calculate the second confidence level C2 according to the voltage value and the set voltage threshold;

[0010] The first confidence level C1 and the second confidence level C2 are integrated to obtain a comprehensive confidence level, the comprehensive confidence level is compared with a set electricity theft judgment threshold, and an electricity theft judgment is output according to the comparison result.

[0011] Preferably, the first confidence level C1 is calculated as follows:

[0012]

[0013] In the formula, C1 is the calculated first confidence level, I L is the current value of the live wire in the meter, I N is the value of the neutral current in the meter, ΔI t is the set difference threshold.

[0014] Preferably, the second confidence level C2 is calculated as follows:

[0015]

[0016] Where C2 is the calculated second confidence level, V load is the voltage value of the load side of the live wire circuit switch in the meter, V t is the set voltage threshold.

[0017] Preferably, the fusion calculation method of the first confidence level C1 and the second confidence level C2 is:

[0018]

[0019] Wherein, P(y=1|C1, C2) represents the probability of electricity theft, β0, β1, β2 are calculation coefficients, which can be set in advance or obtained through historical data training.

[0020] Preferably, the calculation coefficient is obtained by training with historical data in the following manner:

[0021] Constructing a data set, the data set including output features and output labels, the input features are a first confidence level C1 and a second confidence level C2, and the output label is an actual mark of whether electricity theft occurs;

[0022] A model is constructed to predict the probability of electricity theft, and the model is expressed as:

[0023]

[0024] The log-likelihood function is used as the loss function to measure the gap between the model prediction results and the actual annotation results, and the gradient descent algorithm is used to obtain the coefficients β0, β1, and β2 that minimize the loss function.

[0025] Preferably, the fusion calculation method of the first confidence level C1 and the second confidence level C2 is:

[0026] f(C1, C2) = ω1C1 + ω2C2 + b

[0027] Where F(C1, C2) is the fusion result, f(C1, C2)>0 indicates that electricity theft exists, otherwise it indicates that electricity theft does not exist, ω1 and ω2 are weights, and b is a bias term. The weights and bias terms are obtained by training the support vector machine with historical data. The specific training method is:

[0028] Constructing a data set, the data set including output features and output labels, the input features are a first confidence level C1 and a second confidence level C2, and the output label is an actual mark of whether electricity theft occurs;

[0029] Constructing the objective function

[0030] Introduce constraint y i (ω1C 1，i +ω2C 2，i +b)≥1-ξ i ,ξ i ≥0, where y i is the true label of sample i, C 1，i and C 2，i are the first and second confidence levels of sample i, respectively, ξ i ≥0 is a slack variable, indicating the classification error of sample i;

[0031] The minimized objective function that satisfies the constraints is found through the SM0 algorithm or the gradient descent algorithm to obtain the optimal weights ω1, ω2 and bias term b.

[0032] Preferably, the fusion calculation method of the first confidence level C1 and the second confidence level C2 is:

[0033]

[0034] Where, P(S|C1, C2) is the posterior probability of the electricity theft event S occurring when the first confidence level C1 and the second confidence level C2 are given; P(C1|S) is the marginal probability of the first confidence level C1 under the condition of electricity theft; P(C2|S) is the marginal probability of the second confidence level C2 under the condition of electricity theft; P(S) is the prior probability of electricity theft; In the case of non-electricity theft, the marginal probability of the first confidence level C1; In the case of electricity theft, the marginal probability of the second confidence level C2; is the prior probability of non-electricity theft, Among them, P(C1|S), P(C2|S), P(S) is obtained through frequency estimation based on historical data or Gaussian distribution fitting calculation.

[0035] In a second aspect of the present invention, a power theft detection system is further provided, the system comprising a current detection module, a switch module, a voltage detection module, a data processing module and a fusion determination module, which cooperate to execute the power theft detection method provided in the first aspect of the present invention, wherein:

[0036] The current detection module is used to obtain the current value of the live wire in the meter and the current value of the neutral wire in the meter, and send them to the data processing module;

[0037] The switch module is used to switch on and off the live wire circuit in the meter;

[0038] The voltage detection module is used to obtain the voltage value of the load side of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected, and send it to the data processing module;

[0039] The data processing module calculates the current difference between the live wire and the neutral wire according to the current value of the live wire in the meter and the current value of the neutral wire in the meter, and calculates the first confidence C1 according to the current difference and the set difference threshold; calculates the second confidence C2 according to the load side voltage value of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected and the set voltage threshold; sends the first confidence C1 and the second confidence C2 to the fusion judgment module;

[0040] The fusion determination module fuses the first confidence C1 and the second confidence C2 to calculate a comprehensive confidence, compares the comprehensive confidence with a set electricity theft determination threshold, and outputs an electricity theft determination according to the comparison result.

[0041] According to a third aspect of the present invention, an electricity theft detection electric energy meter is provided, comprising an electric energy metering module and an electricity theft detection module, wherein the electricity theft detection module executes the electricity theft detection method as provided in the first aspect of the present invention.

[0042] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the electricity theft detection method according to the first aspect of the present invention is implemented.

[0043] Compared with the prior art, the present invention has the following technical effects:

[0044] The electricity theft detection method proposed in the present invention calculates the current difference between the neutral wire and the live wire, calculates the first confidence according to the current difference and the set difference threshold, and calculates the second confidence according to the load-side voltage value of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected and the set voltage threshold, and calculates the comprehensive confidence, which can ignore normal current fluctuations or load imbalance, reduce the probability of misjudgment and missed judgment of off-meter electricity theft detection, and improve the accuracy of electricity theft detection; reduce the sensitivity to harmonic distortion introduced by nonlinear loads or intermittent loads in different time periods; and improve the sensitivity to small-scale electricity theft or intermittent electricity theft. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flow chart of the electricity theft detection method of the present invention. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in combination with specific embodiments of the present application and with reference to the accompanying drawings.

[0047] Embodiment 1

[0048] A method for detecting electricity theft, such as Figure 1 As shown, the following steps are included:

[0049] Step 1, respectively obtain the current values ​​of the live wire and the neutral wire in the meter, calculate the current difference between the live wire and the neutral wire, and calculate the first confidence C1 according to the current difference and the set difference threshold. In this embodiment, the current values ​​of the live wire and the neutral wire in the meter can be obtained through a current transformer (Current Transformer, CT). Usually a current transformer is wound on the live wire (L line), and the secondary of the transformer outputs a small current signal, which is sent to the measurement circuit inside the electric energy meter. Similarly, there will be an independent current transformer on the neutral wire (N line). In this way, the current of the live wire and the neutral wire can be measured separately. This method can well isolate the high-voltage circuit from the low-voltage measurement circuit to ensure safety and measurement accuracy. In some other embodiments of the present invention, the current values ​​of the live wire and the neutral wire in the meter can also be measured by installing Hall current sensors on both the live wire and the neutral wire. The microcontroller unit (Microcontroller Unit, MCU) in the electric energy meter can also be used to obtain the current values ​​of the live wire and the neutral wire in the meter.

[0050] The calculation method of the first confidence level C1 is as follows:

[0051]

[0052] In the formula, C1 is the calculated first confidence level, I L is the current value of the live wire in the meter, I N is the value of the neutral current in the meter, ΔI t is the set difference threshold. L -I N The closer to ΔI t , the higher the confidence C1 (closer to 1), the greater the possibility of electricity theft; when |I L -I N |Less than ΔI t , the confidence level C1 will be lower, indicating that the system believes that the current difference is within the normal range.

[0053] Step 2: disconnect the live wire loop switch in the meter, obtain the voltage value on the load side of the live wire loop switch in the meter, and calculate the second confidence level C2 according to the voltage value and the set voltage threshold. In this embodiment, a digital voltage sensor (such as an isolated voltage sensor module) is used to directly convert the voltage signal on the load side into a digital signal, which is then transmitted to the microcontroller inside the electric energy meter.

[0054] The calculation method of the second confidence level C2 is:

[0055]

[0056] Where C2 is the calculated second confidence level, V loadis the voltage value of the load side of the live wire circuit switch in the meter, V t is the set voltage threshold. load Close to or exceeding V t , the higher the confidence C2 (closer to 1), the more current still flows after the live wire is disconnected, and the possibility of power theft is greater; if V load Less than V t , indicating that there is no other power supply in the circuit and the confidence level C2 is low.

[0057] Step three: integrate the first confidence level C1 and the second confidence level C2 to obtain a comprehensive confidence level, compare the comprehensive confidence level with a set electricity theft judgment threshold, and output an electricity theft judgment based on the comparison result.

[0058] In one embodiment of the present invention, the fusion calculation method of the first confidence level C1 and the second confidence level C2 is:

[0059]

[0060] Wherein, P(y=1|C1, C2) represents the probability of electricity theft, β0, β1, β2 are calculation coefficients, which can be set in advance or obtained through historical data training.

[0061] The method for obtaining the calculation coefficient through historical data training is:

[0062] Construct a data set, the data set includes output features and output labels, the input features are the first confidence C1 and the second confidence C2, and the output label is the actual mark of whether there is electricity theft. An example of a historical record data set of this embodiment is shown in Table 1:

[0063] First confidence Second confidence level Electricity theft mark 0.85 0.92 1 0.45 0.5 0 0.78 0.6 1 … … …

[0064] Table 1. Dataset examples

[0065] In Table 1, each row represents a record, wherein the first confidence level and the second confidence level are calculated by step 1 and step 2 respectively, the electricity theft mark of 1 indicates that there is electricity theft, and the electricity theft mark of 0 indicates that there is no electricity theft, which is marked by the grid operation record or manual inspection.

[0066] A model is constructed to predict the probability of electricity theft, and the model is expressed as:

[0067]

[0068] The log-likelihood function is used as the loss function to measure the gap between the model prediction results and the actual annotation results, and the gradient descent algorithm is used to obtain the coefficients β0, β1, and β2 that minimize the loss function.

[0069] Among them, the loss function is specifically:

[0070]

[0071] In the formula, y i is the actual label (0 or 1) of the ith sample, is the predicted value of the i-th sample, and n is the total number of samples. The loss function calculates the error between the model's prediction of all samples and the actual annotation, with the goal of minimizing the value of the loss function.

[0072] Gradient descent first initializes the calculation coefficients β0, β1, and β2 to random values, then calculates the gradient of the loss function relative to these coefficients, and then updates each coefficient to reduce the loss function value. The update method is:

[0073]

[0074] In the formula, β j is the jth update of the parameters, α is the learning rate, which is used to control the step size of each update, and L is the loss function. Repeat the update method until the loss function converges, that is, the error drops to a certain set threshold or reaches the maximum number of iterations.

[0075] During training, β0, β1, and β2 are first randomly initialized, and then the predicted value is calculated for each sample i

[0076]

[0077] Using the results of the forward calculation, calculate the loss function value of the entire data set; calculate the gradient of each parameter and update the model parameters; repeat the gradient update until the loss function converges or the maximum number of iterations is reached.

[0078] After training is completed, the performance of the model can be evaluated through cross-validation or by dividing the training set and the test set. Common evaluation indicators include: Accuracy, the proportion of samples predicted correctly; Precision, the proportion of samples predicted to be electricity theft that are actually electricity theft; Recall, the proportion of samples that are correctly detected in actual electricity theft; F1 Score: the harmonic mean of precision and recall.

[0079] After the above training process, the coefficients β0, β1, and β2 finally obtained are the optimal formula parameters for predicting electricity theft behavior.

[0080] In another embodiment of the present invention, the fusion calculation method of the first confidence level C1 and the second confidence level C2 is:

[0081] f(C1, C2) = ω1C1 + ω2C2 + b

[0082] Where, f(C1, C2) is the fusion result, f(C1, C2)>0 indicates the existence of electricity theft, otherwise it indicates the absence of electricity theft, ω1 and ω2 are the weights of the two confidences, indicating their contribution to the classification results, and b is the bias term. The weight and bias term are obtained by training the support vector machine with historical data. The specific training method is:

[0083] A data set is constructed, the data set including output features and output labels, the input features are the first confidence C1 and the second confidence C2, and the output label is the actual mark (0 or 1) of whether electricity theft exists. An example of a data set of this embodiment is also shown in Table 1.

[0084] Constructing the objective function in is the bi-norm of the weight vector.

[0085] Introduce constraint y i (ω1C 1，i +ω2C 2，i +b)≥1-ξ i ,ξ i ≥0, where y i is the true label of sample i (1 for electricity theft, 0 for non-electricity theft), C 1，i and C 2，i are the first and second confidence levels of sample i, respectively, ξ i ≥0 is a slack variable, indicating the classification error of sample i.

[0086] The minimized objective function that satisfies the constraints is found through the SMO (Sequential Minimal Optimization) algorithm or the gradient descent algorithm to obtain the optimal weights ω1, ω2 and bias term b.

[0087] Specifically, in this step, first initialize the parameters ω1, ω2, and b to random values, then calculate the objective function, and check whether the current weights and bias terms meet the constraints. Then use the optimization algorithm (such as SMO or gradient descent method) to gradually adjust ω1, ω2, and b so that the objective function value gradually decreases. When the objective function converges, or the number of iterations reaches the preset value, the training process ends and the optimal ω1, ω2, and b are obtained. Use the optimal ω1, ω2, and b to bring into the fusion algorithm and calculate f(C1, C2). If f(C1, C2) ≥ 0, it is predicted to be electricity theft (that is, the output label y = 1). If f(C1, C2) < 0, it is predicted to be non-electricity theft (that is, the output label y = 0).

[0088] In another embodiment of the present invention, the fusion calculation method of the first confidence level C1 and the second confidence level C2 is:

[0089]

[0090] Where, P(S|C1, C2) is the posterior probability of the electricity theft event S occurring when the first confidence level C1 and the second confidence level C2 are given; P(C1|S) is the marginal probability of the first confidence level C1 under the condition of electricity theft; P(C2|S) is the marginal probability of the second confidence level C2 under the condition of electricity theft; P(S) is the prior probability of electricity theft; In the case of non-electricity theft, the marginal probability of the first confidence level C1; In the case of electricity theft, the marginal probability of the second confidence level C2; is the prior probability of non-electricity theft, Among them, P(C1|S), P(C2|S), P(S) is obtained through frequency estimation based on historical data or Gaussian distribution fitting calculation.

[0091] The calculated posterior probability P(S|C1, C2) is compared with a preset threshold. If the posterior probability P(S|C1, C2) is greater than the preset threshold, it is determined that electricity theft has occurred; otherwise, it is determined that electricity theft has not occurred.

[0092] In some other embodiments of the present invention, the fusion calculation method of the first confidence level C1 and the second confidence level C2 also includes a weighted summation method with artificially assigned weights.

[0093] Embodiment 2

[0094] A power theft detection system, the system comprising a current detection module, a switch module, a voltage detection module, a data processing module and a fusion determination module, cooperates to execute the power theft detection method as described in the first embodiment, wherein:

[0095] The current detection module is used to obtain the current value of the live wire in the meter and the current value of the neutral wire in the meter, and send them to the data processing module;

[0096] The switch module is used to switch on and off the live wire circuit in the meter;

[0097] The voltage detection module is used to obtain the voltage value of the load side of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected, and send it to the data processing module;

[0098] The data processing module calculates the current difference between the live wire and the neutral wire according to the current value of the live wire in the meter and the current value of the neutral wire in the meter, and calculates the first confidence C1 according to the current difference and the set difference threshold; calculates the second confidence C2 according to the load side voltage value of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected and the set voltage threshold; sends the first confidence C1 and the second confidence C2 to the fusion judgment module;

[0099] The fusion determination module fuses the first confidence C1 and the second confidence C2 to calculate a comprehensive confidence, compares the comprehensive confidence with a set electricity theft determination threshold, and outputs an electricity theft determination according to the comparison result.

[0100] Embodiment 3

[0101] A power theft detection electric energy meter comprises an electric energy metering module and a power theft detection module, wherein the power theft detection module executes the power theft detection method as described in the first embodiment.

[0102] Embodiment 4

[0103] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the electricity theft detection method as described in the first embodiment is implemented.

[0104] The above is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for detecting electricity theft, characterized in that: The following steps are involved: The current values ​​of the live wire and the neutral wire in the meter are respectively obtained, the current difference between the live wire and the neutral wire is calculated, and the first confidence level C1 is calculated according to the current difference and the set difference threshold; Disconnect the live wire loop switch in the meter, obtain the voltage value on the load side of the live wire loop switch in the meter, and calculate the second confidence level C2 according to the voltage value and the set voltage threshold; The first confidence level C1 and the second confidence level C2 are integrated to obtain a comprehensive confidence level, the comprehensive confidence level is compared with a set electricity theft judgment threshold, and an electricity theft judgment is output according to the comparison result.

2. A method for detecting electricity theft according to claim 1, characterized in that: The calculation method of the first confidence level C1 is: In the formula, C1 is the calculated first confidence level, I L is the current value of the live wire in the meter, I N is the value of the neutral current in the meter, ΔI t is the set difference threshold.

3. The method for detecting electricity theft according to claim 1, characterized in that: The calculation method of the second confidence level C2 is: Where C2 is the calculated second confidence level, V load is the voltage value of the load side of the live wire circuit switch in the meter, V t is the set voltage threshold.

4. The method for detecting electricity theft according to claim 1, characterized in that: The fusion calculation method of the first confidence level C1 and the second confidence level C2 is: Where P(y=1|C1,C2) represents the probability of electricity theft, β0, β1, β2 are calculation coefficients, which can be set in advance or obtained through historical data training.

5. A method for detecting electricity theft according to claim 4, characterized in that: The method for obtaining the calculation coefficient through historical data training is: Constructing a data set, the data set including output features and output labels, the input features are a first confidence level C1 and a second confidence level C2, and the output label is an actual mark of whether electricity theft occurs; A model is constructed to predict the probability of electricity theft, and the model is expressed as: The log-likelihood function is used as the loss function to measure the gap between the model prediction results and the actual annotation results, and the gradient descent algorithm is used to obtain the coefficients β0, β1, and β2 that minimize the loss function.

6. The method for detecting electricity theft according to claim 1, characterized in that: The fusion calculation method of the first confidence level C1 and the second confidence level C2 is: f(C1,C2)=ω1C1+ω2C2+b In the formula, f(C1, C2) is the fusion result, f(C1, C2)>0 indicates the existence of electricity theft, otherwise it indicates the absence of electricity theft, ω1 and ω2 are weights, b is the bias term, and the weights and bias terms are obtained by training the support vector machine with historical data. The specific training method is: Constructing a data set, the data set including output features and output labels, the input features are a first confidence level C1 and a second confidence level C2, and the output label is an actual mark of whether electricity theft occurs; Constructing the objective function Introduce constraint y i (ω1C 1,i +ω2C 2,i +b)≥1-ξ i ,ξ i ≥0, where y i is the true label of sample i, C 1,i and C 2,i are the first and second confidence levels of sample i, respectively, ξ i ≥0 is a slack variable, indicating the classification error of sample i; The SMO algorithm or the gradient descent algorithm is used to find the minimized objective function that satisfies the constraints, and obtain the optimal weights ω1, ω2 and bias term b.

7. The method for detecting electricity theft according to claim 1, characterized in that: The fusion calculation method of the first confidence level C1 and the second confidence level C2 is: Where, P(S|C1,C2) is the posterior probability of the power theft event S occurring when the first confidence level C1 and the second confidence level C2 are given; P(C1|S) is the marginal probability of the first confidence level C1 under the power theft condition; P(C2|S) is the marginal probability of the second confidence level C2 under the power theft condition; P(S) is the prior probability of power theft; In the case of non-electricity theft, the marginal probability of the first confidence level C1; In the case of electricity theft, the marginal probability of the second confidence level C2; is the prior probability of non-electricity theft, Among them, P(C1|S), P(C2|S), P(S) is obtained through frequency estimation based on historical data or Gaussian distribution fitting calculation.

8. An electricity theft detection system, characterized in that: The system comprises a current detection module, a switch module, a voltage detection module, a data processing module and a fusion determination module, which cooperate to execute the electricity theft detection method according to any one of claims 1 to 7, wherein: The current detection module is used to obtain the current value of the live wire in the meter and the current value of the neutral wire in the meter, and send them to the data processing module; The switch module is used to switch on and off the live wire circuit in the meter; The voltage detection module is used to obtain the voltage value of the load side of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected, and send it to the data processing module; The data processing module calculates the current difference between the live wire and the neutral wire according to the current value of the live wire in the meter and the current value of the neutral wire in the meter, and calculates the first confidence C1 according to the current difference and the set difference threshold; calculates the second confidence C2 according to the load side voltage value of the live wire circuit switch in the meter when the live wire circuit in the meter is disconnected and the set voltage threshold; sends the first confidence C1 and the second confidence C2 to the fusion judgment module; The fusion determination module fuses the first confidence C1 and the second confidence C2 to calculate a comprehensive confidence, compares the comprehensive confidence with a set electricity theft determination threshold, and outputs an electricity theft determination according to the comparison result.

9. An electricity theft detection energy meter, comprising an energy metering module, characterized in that: It also includes an electricity theft detection module, which executes the electricity theft detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the electricity theft detection method according to any one of claims 1 to 7 is implemented.

Images