Electricity consumption abnormity identification method for intelligent electricity consumption

Abstract

The invention provides an electricity consumption abnormity identification method for intelligent electricity consumption, which analyzes the electricity consumption of common standby electric appliances in a user home through big data, monitors the total electricity consumption in combination with an intelligent electric meter, and draws and compares an electricity consumption curve. A model is established by using a training data set, and the model precision is improved through a verification data set and a cross validation method. The model can predict the electricity consumption in a time period when no one is at home of the user, and if the electricity consumption is abnormally high, an alarm is triggered and sent to the user through a mobile network. According to the method, abnormal electricity utilization is effectively identified, energy conservation and emission reduction are realized, the electricity utilization experience of a user is improved, and meanwhile, data security and privacy protection are ensured.

Description

Technical Field

[0001] The present invention relates to power monitoring technology, and in particular to a method for identifying power consumption anomalies in smart power consumption. Background Art

[0002] Smart electricity consumption refers to a new electricity consumption mode that uses advanced technologies such as modern information technology, the Internet of Things, big data, cloud computing, and artificial intelligence to intelligently manage, optimize, and dispatch all aspects of power resource production, transmission, distribution, and consumption. It not only focuses on the simple use of electric energy, but also emphasizes the efficient use of energy, energy conservation and emission reduction, and significant improvement of user electricity experience on the basis of ensuring stable and safe power supply.

[0003] Traditional electricity consumption monitoring methods rely on manual inspections or simple electricity statistics, lack the support of intelligence and big data analysis, and are difficult to accurately capture subtle changes and patterns in household electricity consumption. This makes it difficult to effectively distinguish between normal standby power consumption in an unmanned household and potential abnormal power consumption in practical applications, such as power surges caused by electrical leakage, illegal access, or other unknown factors. Therefore, the inability to promptly discover and deal with waste of power resources, electrical failures, or safety hazards not only increases the user's electricity bill burden, but may also pose a threat to family safety, while reducing the overall management efficiency of power resources and the user's electricity experience. The present invention effectively solves this problem by introducing a smart electricity consumption anomaly identification method and utilizing big data analysis and machine learning technology, thereby improving the accuracy and intelligence level of electricity consumption monitoring. Summary of the invention

[0004] The purpose of the present invention is to provide a method for identifying abnormal power consumption in smart power consumption to solve the above-mentioned deficiencies in the prior art.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a method for identifying abnormal power consumption of smart power consumption, comprising the following steps:

[0006] Step 1: Analyze the average power consumption data of commonly used standby appliances (including refrigerators and routers) in users' homes through big data, process the collected data, and draw a curve of the relationship between the power consumption data of standby appliances and time;

[0007] Step 2: Monitor the total electricity consumption of users in each time period through smart meters, standardize the data and draw a curve of the total electricity consumption of users over time;

[0008] Step 3: Compare the time relationship curve of the power consumption data of standby appliances with the time relationship curve of the total power consumption of users, calculate the difference between the actual power consumption in each time period and the average power consumption of standby appliances in the corresponding time period, and set a threshold. If the calculated difference is greater than the set threshold, it is judged that there is someone at home, otherwise, it is judged that there is no one at home, and this is used to form a training data set and a verification data set;

[0009] Step 4: Build a model using the training data set, and after the model is built, verify the model using the validation data set;

[0010] Step 5: After the model is established, the user's electricity consumption is monitored through the smart meter, and then the trained model is used to predict whether there is anyone at home. If the electricity consumption in the predicted time period when no one is at home is greater than the sum of the average standby appliance electricity consumption and the set threshold, the user is judged to have unreasonable electricity consumption in this time period, and then the unreasonable number threshold is set. If the number of unreasonable electricity consumption in each time period of the user exceeds the set unreasonable number threshold, an abnormal alarm is triggered, and an alarm is sent to the user with abnormal electricity consumption through the mobile network.

[0011] As a further description of the above technical solution: the specific method of drawing the relationship curve between the power consumption data of the standby electrical appliances and the time and the relationship curve between the total power consumption of the user and the time in step 1 and step 2 is divided into the following steps:

[0012] A: Remove outliers (extremely high or low data points) from the collected standby appliance power consumption data and actual power consumption data through a statistical-based method (box plot);

[0013] B: Convert the power consumption data of each time period and the power consumption data of standby appliances into a unified scale to eliminate the dimensional differences in power consumption in different time periods;

[0014] C: Use data analysis tools, and use the processed standby appliance power consumption data and actual power consumption data to draw the standby appliance power consumption data and time relationship curve and the user's total power consumption and time relationship curve, wherein the data analysis tools include but are not limited to Python's Matplotlib library.

[0015] As a further description of the above technical solution: 3. The specific method of establishing the model in step 4 is divided into the following steps:

[0016] D: Extract key features from the processed standby appliance power consumption data and user total power consumption data, including but not limited to power consumption in time period and power consumption change trend;

[0017] E: Divide the key features into a training data set and a validation data set, and adjust the model parameters to improve the model prediction accuracy, where the method of adjusting the model parameters includes but is not limited to grid search, random search or Bayesian optimization method;

[0018] F: Train the model using the training data set, and then use the validation data set to validate the established model. After the validation is completed, cross-validation is used to ensure the generalization ability of the model.

[0019] The model types established include but are not limited to random forests, gradient boosted trees (GBDT), and long short-term memory networks (LSTM, for time series data).

[0020] As a further description of the above technical solution: 4. The specific method of judging the abnormal power consumption of the user in step 5 is divided into the following steps:

[0021] G: The total electricity consumption of the user's home is continuously monitored through smart meters, and the data is transmitted to the system in real time. At the same time, the established model is used to predict the probability of someone being in the user's home in each time period:

[0022] H: In each time period predicted to be unmanned, the difference between actual power consumption and average standby power consumption is calculated. If the difference is greater than the set threshold (considering the average standby power consumption and a reasonable fluctuation range), it is considered unreasonable power consumption;

[0023] I: Record the number of unreasonable power usage in each time period and compare it with the set unreasonable number threshold. When the number of abnormal power usage exceeds the set unreasonable number threshold, the system automatically triggers an abnormal alarm;

[0024] J: Send alarm prompts to users through mobile networks (SMS, APP, 5G network), including information such as the time of abnormal power consumption, power consumption, recommended countermeasures, etc., and record detailed information of each alarm for subsequent analysis and tracking.

[0025] As a further description of the above technical solution: When analyzing the average power consumption data of relays (including refrigerators and routers) through big data in step one, data processing methods such as outlier elimination and trend analysis should also be used to improve the accuracy of data processing.

[0026] As a further description of the above technical solution: when building the model, the model parameters (learning rate, number of iterations, network structure) are adjusted to enable the model to gradually learn the rules and features in the data. At the same time, regularization is used to avoid overfitting, and the accuracy of the model is evaluated by accuracy, recall rate, F1 score, and area under the ROC curve.

[0027] In the above technical solution, the present invention provides a method for identifying abnormal power consumption in smart power consumption, which has the following beneficial effects:

[0028] Accurately identify abnormal electricity usage: Through big data analysis and smart meter monitoring, combined with machine learning models, it is possible to accurately identify abnormal electricity usage in users' homes, and promptly discover and deal with potential safety hazards or electrical failures.

[0029] Energy conservation and emission reduction: Through real-time monitoring and abnormality identification, it guides users to use electricity resources rationally, reduce unnecessary electricity consumption, achieve the goal of energy conservation and emission reduction, and is conducive to environmental protection and sustainable development.

[0030] Improve users' electricity experience: timely discover and solve abnormal electricity usage problems, avoid the inconvenience and losses caused by them, and improve users' electricity experience and quality of life. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the technical solution of the present invention, an embodiment of the present invention provides a technical solution: by installing a smart meter, the user's household electricity consumption is monitored, wherein the smart meter can transmit the monitored data to a cloud server in real time through a wireless network, such as WIFI, Zigbee or a communication interface, such as RS-485, Wi-Fi, NB-IoT, etc., and the cloud server uses a distributed database system to store the original electricity consumption data to ensure the security and scalability of the data, and the cloud server supports data search, which is convenient for the system to perform subsequent data comparison;

[0032] The method for identifying abnormal power consumption in smart power consumption includes the following specific identification steps:

[0033] Step 1: Analyze the average power consumption data of commonly used standby appliances in the user's home through big data, process the collected data, and draw a curve of the relationship between the power consumption data of standby appliances and time. This step is used to analyze the power consumption of some commonly used appliances in the user's home, such as a commonly used refrigerator, router, or a timed sweeping robot, etc. The average power consumed by various appliances in the standby state and when they switch from the standby state to the working state. This average value represents a basic power consumption in the user's home, that is, a basic power consumption in the unmanned state;

[0034] In order to explore this basic consumption and to improve the actual data of the power consumption of standby electrical appliances, a survey can also be used to explore the power consumption of standby electrical appliances in users' homes. If the number of users is large, the power consumption of standby electrical appliances is calculated by calculating the average value to meet the national standard energy consumption. The entire range is the energy consumption fluctuation range of standby electrical appliances. In some special cases, such as users with large power consumption at ordinary times, there are many standby electrical appliances in their homes. These users are specially marked and the power consumption of standby electrical appliances is additionally calculated. Among them, users who do not meet the national standard energy consumption, if their power consumption exceeds that of most users every month or in most cases, are directly judged as abnormal, and the user is surveyed to explore the reasons why their power consumption is greater than that of most users.

[0035] In addition, when recording the power consumption of standby appliances, data processing methods such as outlier elimination and trend analysis should be used to improve the accuracy of data processing;

[0036] Step 2: Monitor the total electricity consumption of users in each time period through smart meters, standardize the data and draw a curve of the total electricity consumption of users over time;

[0037] In step 1 and step 2, the power consumption of standby appliances and the total power consumption of users are monitored. In order to improve the accuracy of the data and facilitate subsequent exploration of whether there is abnormal power consumption, each day can be divided into 24 time periods, and each hour is a monitoring interval. The power consumption and actual power consumption of standby appliances in each time period are monitored, and the daily breakdown is made to facilitate exploration of whether there is anyone at home in each time period;

[0038] The specific method of drawing the relationship curve between the power consumption data of standby appliances and time and the relationship curve between the total power consumption of users and time in step 1 and step 2 is divided into the following steps:

[0039] A: Remove outliers (extremely high or low data points) from the collected standby appliance power consumption data and actual power consumption data through a statistical-based method (box plot);

[0040] B: Convert the power consumption data of each time period and the power consumption data of standby appliances into a unified scale to eliminate the dimensional differences in power consumption in different time periods;

[0041] C: Use data analysis tools, and use the processed standby appliance power consumption data and actual power consumption data to draw a curve of the relationship between the standby appliance power consumption data and time and a curve of the relationship between the user's total power consumption and time, wherein the data analysis tools include but are not limited to Python's Matplotlib library;

[0042] Step 3: Compare the time relationship curve of the power consumption data of standby appliances with the time relationship curve of the total power consumption of users, calculate the difference between the actual power consumption in each time period and the average power consumption of standby appliances in the corresponding time period, and set a threshold. If the calculated difference is greater than the set threshold, it is judged that there is someone at home. Otherwise, it is judged that there is no one at home. The training data set and the verification data set are formed by comparing the total power consumption curve of users with the power consumption curve of standby appliances, and the difference between the total power consumption of users in each time period and the average power consumption of standby appliances in the corresponding time period is calculated. This difference reflects the power consumption of other active appliances (such as lighting, television, air conditioning, etc.) in addition to standby appliances, and indirectly reflects the activities of family members, and sets a reasonable threshold. This threshold represents the power consumption fluctuation range that may be caused by accidental factors (such as appliance failure, external interference, etc.) in the absence of family members' activities. A reasonable threshold is set based on historical data. This threshold represents the power consumption fluctuation range that may be caused by accidental factors (such as appliance failure, external interference, etc.) in the absence of family members' activities;

[0043] For example, if 10% of the power consumption of standby appliances in the time period is the set threshold, then when the difference between the actual power consumption of the user and the average power consumption of the standby appliances is greater than the threshold, for each time period, if the calculated difference is greater than the set threshold, it is considered that there are people at home during the time period, because in addition to the standby appliances, there are other appliances in use; otherwise, it is considered that no one is at home; then, based on the above judgment results, the power consumption data, difference, judgment results (someone at home / no one) and other information of each time period are organized into a data set. This data set will be divided into two parts: training data set and verification data set, which will be used for the establishment and verification of subsequent models;

[0044] Step 4: Establish a model through the training data set, and after the model is established, verify the established model through the verification data set. D: Extract key features from the processed standby appliance power consumption data and user total power consumption data. The key features include but are not limited to power consumption in time period and power consumption change trend;

[0045] The specific steps of building the model are:

[0046] E: Divide the key features into a training data set and a validation data set, and adjust the model parameters to improve the model prediction accuracy, where the method of adjusting the model parameters includes but is not limited to grid search, random search or Bayesian optimization method;

[0047] F: Train the model using the training data set, and then use the validation data set to validate the established model. After the validation is completed, cross-validation is used to ensure the generalization ability of the model.

[0048] The model types established include but are not limited to random forest, gradient boosted tree (GBDT), long short-term memory network (LSTM, for time series data);

[0049] After the model is established, the model parameters (learning rate, number of iterations, network structure) are adjusted to allow the model to gradually learn the rules and features in the data. At the same time, regularization is used to avoid overfitting, and the accuracy of the model is evaluated by accuracy, recall rate, F1 score, and area under the ROC curve.

[0050] Step 5: After the model is established, the user's electricity consumption is monitored through the smart meter, and then the trained model is used to predict whether there is anyone at home. If the electricity consumption in the predicted unattended time period is greater than the sum of the average standby appliance electricity consumption and the set threshold, the user is judged to have unreasonable electricity consumption in this time period, and then the unreasonable number threshold is set. If the number of unreasonable electricity consumption in each time period of the user exceeds the set unreasonable number threshold, an abnormal alarm is triggered, and an alarm is sent to the user with abnormal electricity consumption through the mobile network;

[0051] It should be noted here that when the user's electricity consumption is monitored in the future, the user's electricity consumption in each time period is sent to the cloud server in real time through the smart meter, and the output result of the established model is used to determine whether there is someone at the user's home. One of the judgment methods is that when the output result of the model shows that there is a high probability that no one is at home at this time, for example, the output result shows that the probability of no one is greater than 80%, but in this case, the user's actual electricity consumption is too large (that is, greater than the sum of the power consumption of standby appliances and the set threshold), then it is judged that the user's electricity consumption is unreasonable. In order to prevent the system from being too sensitive, it is also necessary to set an unreasonable number threshold. When the number of unreasonable electricity consumption by the user exceeds the unreasonable number threshold, the user's electricity consumption is judged to be abnormal, an alarm is triggered, and an alarm prompt is sent to the user through the mobile network (SMS, APP, 5G network), including the time of abnormal electricity consumption, electricity consumption, recommended countermeasures and other information, and the detailed information of each alarm is recorded for subsequent analysis and tracking;

[0052] The alarm mechanism is not only triggered when the user's electricity consumption exceeds the set threshold, but also when the user's electricity consumption is lower than the set threshold for multiple times, and the number of times is greater than the set unreasonable number threshold, the alarm is triggered to warn that the meter is damaged or the appliance is damaged, etc., reminding the user to repair it in time; and when the user is warned, the system can also be used to feedback whether the prediction is accurate based on the user's feedback, thereby optimizing the entire system and improving the system's prediction accuracy;

[0053] It is worth noting that in the entire process, relevant laws and regulations on privacy protection and data security must be strictly observed to ensure the security and anonymity of user data and prevent data leakage and abuse.

[0054] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that, for those skilled in the art, the described embodiments can be modified in various ways without departing from the spirit and scope of the present invention. Therefore, the above description is illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for identifying abnormal power consumption in smart power consumption, characterized in that: The steps include: Step 1: Analyze the average power consumption data of commonly used standby appliances (including refrigerators and routers) in users' homes through big data, process the collected data, and draw a curve of the relationship between the power consumption data of standby appliances and time; Step 2: Monitor the total electricity consumption of users in each time period through smart meters, standardize the data and draw a curve of the total electricity consumption of users over time; Step 3: Compare the time relationship curve of the power consumption data of standby appliances with the time relationship curve of the total power consumption of users, calculate the difference between the actual power consumption in each time period and the average power consumption of standby appliances in the corresponding time period, and set a threshold. If the calculated difference is greater than the set threshold, it is judged that there is someone at home, otherwise, it is judged that there is no one at home, and this is used to form a training data set and a verification data set; Step 4: Build a model using the training data set, and after the model is built, verify the model using the validation data set; Step 5: After the model is established, the user's electricity consumption is monitored through the smart meter, and then the trained model is used to predict whether there is anyone at home. If the electricity consumption in the predicted time period when no one is at home is greater than the sum of the average standby appliance electricity consumption and the set threshold, the user is judged to have unreasonable electricity consumption in this time period, and then the unreasonable number threshold is set. If the number of unreasonable electricity consumption in each time period of the user exceeds the set unreasonable number threshold, an abnormal alarm is triggered, and an alarm is sent to the user with abnormal electricity consumption through the mobile network.

2. The method for identifying abnormal power consumption of smart power consumption according to claim 1, characterized in that: The specific method of drawing the relationship curve between the power consumption data of the standby electrical appliances and time and the relationship curve between the total power consumption of the user and time in step 1 and step 2 is divided into the following steps: A: Remove outliers (extremely high or low data points) from the collected standby appliance power consumption data and actual power consumption data through a statistical-based method (box plot); B: Convert the power consumption data of each time period and the power consumption data of standby appliances into a unified scale to eliminate the dimensional differences in power consumption in different time periods; C: Use data analysis tools, and use the processed standby appliance power consumption data and actual power consumption data to draw the standby appliance power consumption data and time relationship curve and the user's total power consumption and time relationship curve, wherein the data analysis tools include but are not limited to Python's Matplotlib library.

3. The method for identifying abnormal power consumption of smart power consumption according to claim 1, characterized in that: The specific method of establishing the model in step 4 is divided into the following steps: D: Extract key features from the processed standby appliance power consumption data and user total power consumption data, including but not limited to power consumption in time period and power consumption change trend; E: Divide the key features into a training data set and a validation data set, and adjust the model parameters to improve the model prediction accuracy, where the method of adjusting the model parameters includes but is not limited to grid search, random search or Bayesian optimization method; F: Train the model using the training data set, and then use the validation data set to validate the established model. After the validation is completed, cross-validation is used to ensure the generalization ability of the model. The model types established include but are not limited to random forests, gradient boosted trees (GBDT), and long short-term memory networks (LSTM, for time series data).

4. The method for identifying abnormal power consumption of smart power consumption according to claim 1, characterized in that: The specific method of judging abnormal power consumption of the user in step 5 is divided into the following steps: G: The total electricity consumption of the user's home is continuously monitored through smart meters, and the data is transmitted to the system in real time. At the same time, the established model is used to predict the probability of someone being in the user's home in each time period: H: In each time period predicted to be unmanned, the difference between actual power consumption and average standby power consumption is calculated. If the difference is greater than the set threshold (considering the average standby power consumption and a reasonable fluctuation range), it is considered unreasonable power consumption; I: Record the number of unreasonable power usage in each time period and compare it with the set unreasonable number threshold. When the number of abnormal power usage exceeds the set unreasonable number threshold, the system automatically triggers an abnormal alarm; J: Send alarm prompts to users through mobile networks (SMS, APP, 5G network), including information such as the time of abnormal power consumption, power consumption, recommended countermeasures, etc., and record detailed information of each alarm for subsequent analysis and tracking.

5. The method for identifying abnormal power consumption of smart power consumption according to claim 1, characterized in that: When analyzing the average power consumption data of relays (including refrigerators and routers) through big data in step 1, data processing methods such as outlier elimination and trend analysis are also used to improve the accuracy of data processing.

6. The method for identifying abnormal power consumption of smart power consumption according to claim 3, characterized in that: When building the model, the model parameters (learning rate, number of iterations, network structure) are adjusted to allow the model to gradually learn the patterns and features in the data. At the same time, regularization is used to avoid overfitting, and the accuracy of the model is evaluated through accuracy, recall rate, F1 score, and area under the ROC curve.