Non-intrusive single load identification method, micro circuit breaker equipment and computer equipment

Abstract

The invention provides a non-intrusive single load identification method, micro circuit breaker equipment and computer equipment, and the method comprises the steps: firstly obtaining an analog signal of a to-be-identified load, converting the analog signal into a digital signal, carrying out the time domain and frequency domain feature extraction of the digital signal, obtaining load feature data, and carrying out the recognition of the load. The load feature data is input into a first load identification model deployed in the micro-circuit breaker device, information of a to-be-identified load is identified, a first identification result is obtained, at least part of the load feature data and the first identification result are uploaded to a cloud for analysis, and a second identification result is obtained. Therefore, the comprehensiveness and accuracy of non-intrusive single load identification can be improved.

Description

Technical Field

[0001] This application relates to the technical field of photovoltaic power station operation and maintenance, and specifically relates to a non-intrusive single load identification method, a miniature circuit breaker device, and a computer device. Background Art

[0002] The existing NILM (Non-Intrusive Load Monitoring) technology realizes non-intrusive load identification by analyzing the total power consumption data on the household side.

[0003] However, the accuracy of the existing NILM technology significantly decreases when multiple loads are superimposed, and it has poor adaptability to low-power loads and fast-switching scenarios, and is prone to misjudgment and missed judgment. Summary of the Invention

[0004] This application provides a non-intrusive single load identification method, a miniature circuit breaker device, a computer device, and a readable storage medium, which can improve the accuracy of non-intrusive single load identification.

[0005] In a first aspect, this application provides a non-intrusive single load identification method, and the method includes:

[0006] Obtain an analog signal of the load to be identified, and convert the analog signal into a digital signal;

[0007] Extract time-domain and frequency-domain features from the digital signal to obtain load feature data;

[0008] Input the load feature data into a first load identification model deployed in the miniature circuit breaker device to identify information of the load to be identified, and obtain a first identification result;

[0009] Upload at least part of the load feature data and the first identification result to the cloud for analysis to obtain a second identification result.

[0010] Optionally, the miniature circuit breaker device includes a current detection device, a voltage detection device, and an analog-to-digital converter; the obtaining an analog signal of the load to be identified and converting the analog signal into a digital signal includes:

[0011] Collect a current signal of the load to be identified through the current detection device;

[0012] Collect a voltage signal of the load to be identified through the voltage detection device;

[0013] Convert the current signal and the voltage signal into the digital signal through the analog-to-digital converter.

[0014] Optionally, the extracting of time-domain and frequency-domain features from the digital signal to obtain load feature data includes:

[0015] Performing time-domain analysis on the characteristics of the voltage signal and the current signal changing with time to obtain a time-domain signal, and extracting time-domain features from the time-domain signal;

[0016] Converting the time-domain signal into a frequency-domain signal, and extracting frequency-domain features from the frequency-domain signal.

[0017] Optionally, the first recognition result and the second recognition result at least include the type and access status of the load to be recognized.

[0018] Optionally, the first load recognition model is a tree model based on machine learning, and the method further includes a training step for the tree model, specifically including:

[0019] Obtaining loads with known load types and access statuses, and extracting the load feature data;

[0020] Using the load feature data as samples and the load type and access status as labels to train an initial tree model, obtaining the trained tree model and deploying it on the miniature circuit breaker device.

[0021] Optionally, the uploading of at least part of the load feature data and the first recognition result to the cloud for analysis to obtain a second recognition result includes:

[0022] Transmitting the at least part of the load feature data and the first recognition result to an intelligent gateway, and uploading them to the cloud through the Ethernet protocol;

[0023] Analyzing and optimizing the uploaded data through a second load recognition model deployed in the cloud to obtain the second recognition result;

[0024] Wherein, the performance of the second load recognition model is better than that of the first load recognition model.

[0025] Optionally, the method further includes:

[0026] Performing a preliminary recognition on the first recognition result, and if the first recognition result indicates that the load to be recognized is using electricity illegally, triggering a local alarm mechanism;

[0027] Comparing the first recognition result with the second recognition result, and if the two are consistent, confirming to use the first recognition result as the final recognition result of the load;

[0028] If the two are inconsistent, updating the first load recognition model in the cloud.

[0029] Optionally, updating the first load recognition model in the cloud includes:

[0030] Using the load feature data in the cloud as samples and the second recognition result as labels to train the first load recognition model, obtaining an updated first load recognition model;

[0031] Issuing and deploying the updated first load recognition model to the miniature circuit breaker device.

[0032] In a second aspect, the present application further provides a non-invasive miniature circuit breaker device, including:

[0033] A metering chip, configured to obtain an analog signal of a load to be recognized and convert the analog signal into a digital signal;

[0034] A feature extraction module, configured to perform time-domain and frequency-domain feature extraction on the digital signal to obtain load feature data;

[0035] A load recognition module, configured to input the load feature data into the first load recognition model deployed in the miniature circuit breaker device to recognize information of the load to be recognized, obtaining a first recognition result;

[0036] A cloud-edge-end collaborative architecture, configured to upload at least part of the load feature data and the first recognition result to the cloud for analysis, obtaining a second recognition result.

[0037] In a third aspect, the present application further provides a computer device, including a processor and a memory; the memory is configured to store a computer program; the processor is configured to execute the computer program stored in the memory to implement the steps of the non-invasive single load recognition method as described above.

[0038] In a fourth aspect, the present application further provides a computer-readable storage medium, storing a computer program, where the computer program is loaded by a processor to implement the steps of the non-invasive single load recognition method as described above.

[0039] In the embodiment of the present application, first, an analog signal of the load to be recognized is acquired, and the analog signal is converted into a digital signal. Then, time-domain and frequency-domain feature extraction is performed on the digital signal to obtain load feature data. Next, the load feature data is input into the first load recognition model deployed in the miniature circuit breaker device to recognize the information of the load to be recognized, and a first recognition result is obtained. Thus, the accuracy of single-load recognition can be improved based on the time-frequency domain feature extraction technology and the machine learning model. Finally, at least part of the load feature data and the first recognition result are uploaded to the cloud for analysis to obtain a second recognition result. Therefore, the reliability and fault tolerance of scenario recognition such as single-load, multi-load superposition, low-power load, and rapid load mode switching can be improved through the collaborative work of edge computing and cloud computing. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is a flowchart of a non-intrusive single-load recognition method provided in the embodiment of the present application;

[0042] Figure 2 It is a schematic diagram of load recognition by the miniature circuit breaker device provided in the embodiment of the present application;

[0043] Figure 3 It is a schematic diagram of load recognition by the cloud provided in the embodiment of the present application;

[0044] Figure 4 It is a schematic diagram of cloud-edge-end integrated load recognition provided in the embodiment of the present application;

[0045] Figure 5 It is a schematic diagram of the recognition result provided in the embodiment of the present application;

[0046] Figure 6 It is a schematic diagram of the functional modules of the miniature circuit breaker device provided in the embodiment of the present application;

[0047] Figure 7 It is a schematic diagram of the structure of a computer device provided in the embodiment of the present application. Detailed Embodiments

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0049] In the description of the present application, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, "a plurality" means two or more, unless otherwise specifically defined.

[0050] In the present application, the term "exemplary" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "exemplary" in the present application is not necessarily construed as being more preferred or having more advantages than other embodiments. At the same time, it can be understood that in the specific implementation of the present application, when it comes to relevant data such as user information and user data, when the above embodiments of the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0051] In order for any person skilled in the art to implement and use the present application, the following description is provided. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present application can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid unnecessary details from obscuring the description of the present application. Therefore, the present application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present application.

[0052] In the prior art, the smart grid and smart building technologies have developed rapidly. By integrating advanced communication, information, and control technologies, the intelligent management of the power system and the efficient utilization of energy have been realized. In the field of safe electricity use monitoring, traditional intrusive monitoring methods obtain electricity use information by installing sensors or monitoring devices on the user side. Although they can provide accurate data, there are problems such as privacy leakage, high installation costs, and complex maintenance. In contrast, the non-intrusive load monitoring technology (NILM) has become a research hotspot because of its characteristic that it only needs to process and analyze the electricity use data on the user's incoming side to obtain the information of electrical appliances in use. The NILM technology can identify abnormal electricity use behaviors such as electric bicycles charging at home and high-power prohibited electrical appliances by analyzing the characteristics of voltage and current waveforms and combining machine learning algorithms, providing an effective means for the safe electricity use management of smart grids and smart buildings.

[0053] However, the existing NILM technologies rely on high-precision sampling devices and complex signal processing algorithms, resulting in high system costs and poor real-time performance, making it difficult to be deployed in resource-constrained edge devices (such as micro circuit breakers). Secondly, when identifying multi-load combinations, the mutual interference and noise between loads easily affect the accuracy of feature extraction, leading to a decrease in the recognition rate. In addition, the existing methods have insufficient adaptability to load changes. Especially when the load power is small or the switching is frequent, misjudgment and missed judgment are likely to occur, and the start and stop moments of the load cannot be accurately captured.

[0054] In view of the above, the present application provides a non-intrusive single-load identification method and a micro circuit breaker, which will be described in detail below.

[0055] Figure 1 It is a schematic flowchart of the non-intrusive single-load identification method in an embodiment of the present application. As Figure 1 shown, this non-intrusive single-load identification method is applied to a micro circuit breaker and may include the following steps S101 to S104.

[0056] Step S101, obtain the analog signal of the load to be identified and convert the analog signal into a digital signal.

[0057] Among them, the micro circuit breaker, whose full name is Micro Circuit Breaker (abbreviated as MCB), is a key component of a non-intrusive load identification system used in smart grids and smart buildings. In the scenario of the present application, the micro circuit breaker can perform edge computing work for single-load information identification as an edge device.

[0058] Specifically, the current detection device and voltage detection device in the micro circuit breaker can be used to capture the current and voltage passing through the circuit. These current transformers can sense the change in magnetic flux passing through them and convert these changes into corresponding analog voltage or current signals.

[0059] Among them, the collected analog signals may contain noise or be inaccurate within the required voltage or current range. Therefore, these signals need to be conditioned, including filtering, amplification, isolation, etc., to improve the quality and accuracy of the signals.

[0060] Specifically, the conditioned analog signals are then sent to the analog-to-digital converter ADC of the miniature circuit breaker device. The ADC is responsible for converting the continuously varying analog signals into discrete digital signals, which can be processed by the Arm core microcontroller MCU or other digital processing devices.

[0061] The conversion process of the ADC usually includes three steps: sampling, quantization, and encoding. First, the ADC samples the analog signal regularly at a certain sampling frequency; then, the sampled voltage or current value is quantized to the closest value; finally, the quantized value is encoded into a binary digital signal.

[0062] Finally, the converted digital signals will be sent to the microcontroller MCU in the miniature circuit breaker device for further processing, such as time-domain and frequency-domain feature extraction, to provide data support for subsequent load identification.

[0063] Through the above process, the miniature circuit breaker device can convert the analog signals of the load to be identified into digital signals, providing basic data for realizing non-intrusive load identification.

[0064] In some embodiments, the miniature circuit breaker device may include a current detection device, a voltage detection device, and an analog-to-digital converter, and step S101 may include:

[0065] First, the current detection device collects the current signal of the load to be identified;

[0066] Next, the voltage detection device collects the voltage signal of the load to be identified;

[0067] Finally, the current signal and the voltage signal are converted into digital signals by the analog-to-digital converter.

[0068] Among them, the current detection device is an electromagnetic induction device for measuring the alternating current on the conductor passing through it. When an alternating current flows through a conductor, a magnetic field will be generated around the conductor. The current detection device induces this magnetic field through its coil and generates a secondary current proportional to the original current. The current detection device is connected in series in the circuit to capture the current signal passing through the circuit in a non-intrusive manner. This signal is in analog form and represents the current consumption characteristics of the load.

[0069] Among them, the voltage detection device can be used to measure the voltage level in a circuit. Similar to the current detection device, the voltage detection device also works based on the principle of electromagnetic induction. It is usually connected in parallel in the circuit. By inducing the voltage through the internal coil, a secondary voltage proportional to the original voltage is generated. The voltage detection device is connected in parallel in the circuit to capture the voltage signal in the circuit in a non-invasive manner. This signal is also in analog form and represents the voltage characteristics of the load.

[0070] Among them, the analog-to-digital converter ADC is an electronic device that converts analog signals into digital signals. The ADC converts analog signals into digital signals through three steps: sampling, quantization, and encoding. Sampling is the process of periodically measuring the analog signal; quantization is the process of mapping the sampled value to the closest digital value; encoding is the process of converting the quantized value into a binary number.

[0071] Specifically, the ADC can sample the current and voltage signals periodically at a specific sampling frequency (such as 3.2 kHz). The ADC converts the sampled analog values into a finite number of discrete levels, and each level corresponds to a specific digital value. The ADC encodes the quantized values into binary digital signals, and these signals can be processed by a microcontroller (MCU) or other digital processing devices.

[0072] In the above way, the miniature circuit breaker device of the present application can obtain digital signals representing the current and voltage characteristics of the load to be identified. These digital signals contain the time-domain and frequency-domain characteristics of the load and are the basis for realizing non-invasive load identification. The digital signals can then be used in subsequent processing steps such as feature extraction, model training, and load identification to achieve intelligent identification and management of the load.

[0073] The miniature circuit breaker device integrated with the current detection device, voltage detection device, and ADC in the present application provides an efficient and accurate means for power consumption monitoring and analysis for smart grids and intelligent buildings, which helps to improve energy utilization efficiency and power consumption safety.

[0074] Step S102: Extract time-domain and frequency-domain features from the digital signal to obtain load feature data.

[0075] In some embodiments, before feature extraction, the digital signal can be preprocessed first.

[0076] Specifically, as Figure 2 shown, the current signal and voltage signal can be preliminarily processed. Removing the DC component is the first step of preprocessing, and the purpose is to eliminate the DC bias in the signal and make the mean value of the processed signal zero. The DC component may cause baseline drift and affect the accuracy of signal analysis. The mathematical expression of this step is as follows:

[0077]

[0078] Among them, x[n] is the original signal, and x clean [n] is the signal after removing the DC component, N is the signal length, is the sum of all sampling points of the signal.

[0079] Specifically, a FIR filter can be used to filter out specific frequency components in the signal, such as high-frequency noise or power frequency interference. A FIR filter is a commonly used digital filter for filtering out specific frequency components in the signal, such as high-frequency noise or power frequency interference. One of the main advantages of a FIR filter is its linear phase characteristic, which does not change the phase relationship of the signal, and this is very important for maintaining the integrity of the signal characteristics. The difference equation of the FIR filter is as follows:

[0080]

[0081] Among them, y[n] is the filtered output signal, h[k] is the filter coefficient, which determines the characteristics of the filter (such as cut-off frequency, passband, stopband, etc.), M is the order of the filter, that is, the number of filter coefficients minus one, and x[n - k] is the delayed version of the input signal.

[0082] By denoising and filtering, the signal-to-noise ratio of the signal can be improved, making the useful signal more prominent. Removing the DC component and unnecessary frequency components can avoid introducing errors in subsequent analysis. The preprocessed signals are more suitable for feature extraction because they are cleaner and closer to the true load characteristics. Preprocessing can improve the accuracy of model training and prediction because the quality of the input data directly affects the performance of the model.

[0083] Through the above preprocessing steps, it can be ensured that the quality of the signal input to the feature extraction stage is as high as possible, thereby improving the accuracy and reliability of the entire non-intrusive load monitoring system.

[0084] In some embodiments, step S102 may include:

[0085] Performing time-domain analysis on the characteristics of the voltage signal and current signal changing with time to obtain a time-domain signal, and extracting time-domain features from the time-domain signal;

[0086] Converting the time-domain signal into a frequency-domain signal, and extracting frequency-domain features from the frequency-domain signal.

[0087] Among them, time-domain features can include average value, variance, peak-to-peak value, root mean square (RMS) value, and energy, etc. The average value is to calculate the average magnitude of voltage or current over a period of time, reflecting the average power consumption of the electrical appliance. The variance can measure the degree of fluctuation of the voltage or current signal. The larger the variance, the more intense the signal fluctuation. The peak-to-peak value is the difference between the maximum and minimum values of the voltage or current signal within one cycle, reflecting the amplitude range of the signal. The root mean square (RMS) value is defined according to the thermal effect of the signal and is the value of the AC signal equivalent to the DC signal, which is more in line with people's intuitive perception of the power of the electrical appliance. Energy is the area enclosed by the voltage and current waveforms within a certain time interval, reflecting the total electrical energy consumed by the electrical appliance during this time period.

[0088] Specifically, the DC component can be removed to make the signal mean zero, avoiding the interference of baseline drift on subsequent analysis. An FIR filter can be applied to remove high-frequency noise or power frequency interference, retaining the useful frequency components in the signal.

[0089] It can be understood that frequency-domain analysis focuses on the frequency components of voltage and current signals. By converting the time-domain signal into a frequency-domain signal, the frequency characteristics of the electrical appliance can be analyzed.

[0090] Among them, frequency-domain features can include power spectrum, band energy, fundamental frequency, and harmonic content. The power spectrum describes the power distribution of each frequency component in the signal, reflecting the contribution of different frequency components to the total power. The band energy is to calculate the signal energy within a specific frequency band, which is used to identify the specific frequency characteristics of the electrical appliance, such as the rotation speed of the motor or the flashing frequency of the fluorescent lamp. The fundamental frequency is to identify the main frequency component in the signal, which is very important for identifying periodic loads (such as air conditioners, refrigerators). The harmonic content is to analyze the harmonic components in the signal. Harmonics are common interference sources in the power system and have an impact on the stability and safety of the power system.

[0091] In some embodiments, unnecessary frequency components can be removed through a filter, retaining the frequency components related to the load characteristics. Further analysis can be performed on the frequency-domain signal, such as calculating features such as the peak value and bandwidth of the spectrum.

[0092] In some embodiments, the present application also involves a feature selection process during feature extraction. The random forest algorithm is used for feature selection, and the specific steps are as follows:

[0093] When constructing each decision tree, the data not selected into the training samples (out-of-bag data) is used to calculate the error. The out-of-bag (OOB) error is used as an unbiased estimate of the model performance without the need for a separate validation set.

[0094] For each feature f i , the random forest calculates its importance through the following steps:

[0095] On the out-of-bag data of each tree, randomly permute the values of feature f i Calculate the out-of-bag error ErrorOOBm(i) after permutation.

[0096] The importance of feature f

[0097] is represented by the average difference in error before and after permutation as: i The index reflects the contribution of feature f

[0098]

[0099] to the prediction accuracy of the model; the larger the value, the more important the feature. According to the importance scores of each feature obtained from the above calculations, all features can be sorted from high to low according to the scores. The features ranked in the front contribute more to the model, and vice versa. i

[0100] Through feature extraction and feature selection, this application can extract the features that best reflect the load characteristics from the original signal and select the features that are most helpful for identification, thereby improving the accuracy and efficiency of load identification.

[0101] In summary, through time-domain and frequency-domain analysis of voltage and current signals, this application can extract a series of features that can reflect the characteristics of electrical appliances. These features are crucial for realizing non-intrusive load monitoring and can help identify and classify different types of electrical load. By combining time-domain and frequency-domain features, the usage of electrical appliances can be more accurately identified and monitored, thereby improving energy utilization efficiency and electrical safety.

[0102] Step S103: Input the load feature data into the first load identification model deployed on the miniature circuit breaker device to identify the information of the load to be identified and obtain the first identification result.

[0103] Figure 3 Among them, the first identification result can at least include the type and access status of the load to be identified. As shown, before load identification, it is first necessary to extract useful feature data from the collected current and voltage signals, including time-domain features (such as average value, variance, peak-to-peak value, etc.) and frequency-domain features (such as fundamental frequency, harmonic components, power spectrum, etc.).

[0104] Then, the extracted load feature data can be input into the first load identification model deployed on the miniature circuit breaker device. For example, the first load identification model can be an algorithm based on machine learning, such as random forest, support vector machine (SVM), neural network, or gradient boosting decision tree (GBDT), etc.

[0105] ​The first load recognition model can analyze and identify the information of the load to be recognized by using the input feature data. Involving pattern matching and classification, the model can classify the input feature data according to the mapping relationship between the features learned during training and the load types.

[0106] In some embodiments, the first recognition result output by the model can at least include:

[0107] The type of the load to be recognized, such as air conditioner, refrigerator, washing machine, electric bicycle charger, etc.

[0108] The access status of the load to be recognized, for example, whether the load is connected to the power grid, and its working status, such as running, standby, off, etc. As Figure 5 shown in the schematic diagram, in this diagram, the number "1" indicates that the load is in the access state, and the number "0" indicates that the load is in the disconnected state.

[0109] In some embodiments, the first recognition result can be uploaded to the cloud through a cloud-edge-end collaborative architecture for further analysis to improve the accuracy and reliability of recognition. The cloud server can use more complex models or algorithms to deeply analyze the uploaded data and provide more accurate recognition results.

[0110] In summary, by inputting the load feature data into the first load recognition model deployed on the miniature circuit breaker device, the present application can achieve non-invasive recognition of electrical devices in homes or buildings and obtain the first recognition result including the load type and access status. This process not only improves the electrical safety but also provides important information for energy management and optimization of smart grids and smart buildings.

[0111] Among them, the first load recognition model is a tree model based on machine learning, and the training steps of the tree model specifically include:

[0112] First, obtain the loads with known load types and access statuses, and extract the load feature data;

[0113] Next, use the load feature data as samples and the load types and access statuses as labels to train the initial tree model, and obtain the trained tree model and deploy it on the miniature circuit breaker device.

[0114] In specific implementation, the first load recognition model is a tree model based on machine learning, which is used to identify the type and access status of the load to be recognized. The tree model is a common classification algorithm that constructs a decision tree for data classification.

[0115] Specifically, a dataset for training the model can be collected first, which includes various loads with known load types and access states. The current and voltage signals of different types of loads in different states can be collected through the miniature circuit breaker device, including normal working state, standby state, off state, etc.

[0116] Next, features that help identify the load type and access state can be extracted from the collected current and voltage signals. The current and voltage signals can be preprocessed, including denoising, filtering, removing DC components, etc. Time-domain features (such as average value, variance, peak-to-peak value, etc.) and frequency-domain features (such as fundamental frequency, harmonic components, power spectrum, etc.) can be extracted from the preprocessed signals.

[0117] Next, the extracted feature data can be corresponded to the known load types and access states to construct training samples and labels. The extracted feature data can be used as samples, and each sample corresponds to a set of feature values. The known load types and access states are used as labels, and each sample corresponds to a label, which represents the load type and access state to which the sample belongs.

[0118] Specifically, the training samples and labels can be used to train a tree model so that the model can learn the mapping relationship between the features and the load type and access state. Appropriate machine learning algorithms can be selected, such as decision tree, random forest, gradient boosting decision tree (GBDT), etc., and the model can be trained using the training samples and labels, and the model parameters can be adjusted to minimize the prediction error. During the training process, methods such as cross-validation can be used to evaluate the performance of the model to avoid overfitting.

[0119] Finally, the trained model can be deployed to the miniature circuit breaker device so that it can identify the types and access states of the loads to be identified in actual applications. The parameters of the trained model can be saved and integrated into the firmware or software of the miniature circuit breaker device. Ensure that the miniature circuit breaker device has sufficient computing power and storage space to run the model. The model can be tested on the miniature circuit breaker device to verify the recognition accuracy and real-time performance of the model.

[0120] It should also be noted that the edge computing of this application uses a lightweight LightGBM (LGBM) model to quickly classify local data on the edge side (i.e., the miniature circuit breaker device) and upload key information to the cloud. LGBM is an efficient gradient boosting decision tree (GBDT) algorithm designed to provide fast and efficient model training and prediction.

[0121] LGBM is based on the gradient boosting decision tree (GBDT) framework and gradually reduces the prediction error by iteratively training multiple decision trees. The prediction value of each tree is the sum of the prediction results of all trees:

[0122]

[0123] Among them, K is the total number of trees, and f k (x i ) is the predicted value of the k-th tree for the sample x i , and F is the decision tree function space.

[0124] The objective function of LGBM consists of a loss function and a regularization term to balance the prediction accuracy and complexity of the model:

[0125]

[0126] The loss function measures the difference between the predicted value and the true value of the model:

[0127]

[0128] The regularization term (L1 / L2) prevents overfitting:

[0129]

[0130] Among them, w j is the leaf node weight, m is the number of leaf nodes, and λ1 and λ2 are regularization coefficients. In each iteration, LGBM uses the gradient descent method to fit the residuals and update the model:

[0131]

[0132] First-order gradient:

[0133]

[0134] Second-order gradient:

[0135]

[0136] The weight of the leaf node is calculated by the following formula:

[0137]

[0138] Among them, I j is the sample set of leaf node j, and λ2 is the L2 regularization coefficient.

[0139] The split gain is used to select the best split point, and the formula for maximizing the split gain is:

[0140]

[0141] Among them, I L and I R are the sample sets of the left and right child nodes, and γ is the split threshold.

[0142] In the above manner, the LGBM model can quickly and efficiently process and classify data on the edge side of this application, and upload key information to the cloud for further analysis, thereby realizing power consumption monitoring and management in smart grids and smart buildings.

[0143] In summary, this application can obtain a trained first load recognition model based on a machine learning tree model and deploy it on the miniature circuit breaker device. This model can accurately identify the type and access status of the load to be recognized, providing support for power consumption management and security monitoring in smart grids and smart buildings.

[0144] Step S104: Upload at least part of the load feature data and the first recognition result to the cloud for analysis to obtain a second recognition result.

[0145] As Figure 4 shown, the extracted load feature data and the first recognition result can be uploaded to the cloud through a communication interface (such as Ethernet). The uploaded data may include key feature values and a preliminary judgment of the load type.

[0146] The cloud server receives the uploaded data and performs in-depth analysis using a more complex and powerful model. Cloud analysis may include more advanced feature processing, model training and optimization, and global data analysis.

[0147] After the cloud analysis is completed, a second recognition result can be generated, which is more accurate and detailed than the first recognition result. The second recognition result may include a more accurate load type, access status, and other relevant information, such as energy consumption assessment or power consumption suggestions.

[0148] The second recognition result can be used for energy management and optimization in smart grids and smart buildings, improving power consumption efficiency and safety. It can also be used to provide power consumption reports and suggestions to users or managers to help them better manage power consumption.

[0149] In the above manner, this application utilizes a cloud-edge-end collaborative architecture, combines the real-time processing capabilities of edge devices and the powerful computing resources of the cloud, and realizes more efficient and accurate load recognition and management.

[0150] In some embodiments, step S104 may include:

[0151] First, transmit at least part of the load feature data and the first recognition result to the smart gateway and upload them to the cloud through the Ethernet protocol;

[0152] Next, analyze and optimize the uploaded data through a second load recognition model deployed in the cloud to obtain a second recognition result;

[0153] Among them, the performance of the second load recognition model is better than that of the first load recognition model.

[0154] Specifically, at least part of the load feature data and the first recognition result can be transmitted to the smart gateway. The transmission can be carried out wirelessly or wired, depending on the communication capabilities of the smart gateway and the miniature circuit breaker device.

[0155] The smart gateway uploads the data to the cloud through the Ethernet protocol. Ethernet is a widely used local area network technology that provides high-speed and reliable data transmission. During the data transmission process, the security and privacy protection of the data should be ensured, and encryption and security authentication mechanisms may be required.

[0156] Specifically, the second load recognition model can be deployed in the cloud. This model can be based on more complex or advanced algorithms, such as deep learning models. The performance of the second load recognition model is better than that of the first load recognition model because it can use more data and more complex algorithms for training and optimization.

[0157] The second load recognition model in the cloud can conduct in-depth analysis on the uploaded data, use more advanced feature processing and model optimization techniques to improve the recognition accuracy. The cloud model can also conduct global data analysis to identify electricity consumption patterns across devices or regions.

[0158] After the cloud analysis is completed, a second recognition result can be generated. The second recognition result can be fed back to the miniature circuit breaker device for updating the local model or providing immediate control actions.

[0159] Through the above method, this application utilizes the cloud-edge-end collaborative architecture, combines the real-time processing capabilities of edge devices and the powerful computing resources of the cloud, and realizes more efficient and accurate load recognition and management. This multi-level recognition process not only improves the recognition accuracy but also enhances the adaptability and scalability of the system.

[0160] In some embodiments, the method of this application may further include:

[0161] First, conduct a preliminary recognition on the first recognition result. If the first recognition result is that the load to be recognized violates electricity usage regulations, trigger the local alarm mechanism;

[0162] Next, compare the first recognition result with the second recognition result. If the two are consistent, confirm to adopt the first recognition result as the final recognition result of the load;

[0163] If the two are inconsistent, update the first load recognition model in the cloud.

[0164] In specific implementation, according to this solution, the process of processing the first recognition result and the second recognition result is as follows:

[0165] The first load recognition model deployed on the miniature circuit breaker device can be used to analyze the collected load characteristic data to obtain the first recognition result. The first recognition result may include the load type, access status, and a preliminary judgment on whether there is illegal electricity use. If the first recognition result indicates that there is illegal electricity use for the load to be recognized (for example, it is detected that an electric bicycle is charging indoors or a high-power prohibited electrical appliance is used), the local alarm mechanism is immediately triggered. The local alarm can be achieved through methods such as sound alarms, light warnings, or directly disconnecting the power supply to quickly remind users or management personnel of potential electricity use safety issues.

[0166] Specifically, the first recognition result and the relevant load characteristic data can be uploaded to the cloud through the intelligent gateway. The second load recognition model deployed in the cloud conducts a more in-depth analysis of the uploaded data to obtain the second recognition result.

[0167] Specifically, the second recognition result obtained at the cloud end can be compared with the first recognition result on the miniature circuit breaker device. If the first recognition result is consistent with the second recognition result, it indicates that the two models have the same recognition of the load. At this time, the first recognition result can be confirmed as the final recognition result of the load.

[0168] If the first recognition result is inconsistent with the second recognition result, it indicates that there may be recognition errors or areas where the model needs to be improved. The first load recognition model is updated in the cloud, and the model is retrained using a large amount of data and more complex algorithms in the cloud. The updated model will have more accurate feature recognition capabilities to improve the accuracy of future recognition.

[0169] In some embodiments, the updated model can be sent down and deployed on the miniature circuit breaker device to replace the original first load recognition model. Ensure that the model on the miniature circuit breaker device can continuously learn and improve to adapt to the changing electricity use environment and load characteristics. Through the above methods, the present application realizes a system with dynamic learning and self-improvement, which can continuously improve the accuracy and reliability of load recognition while ensuring electricity use safety.

[0170] In some embodiments, the load characteristic data in the cloud can be used as samples first, and the second recognition result can be used as labels to train the first load recognition model to obtain an updated first load recognition model; then the updated first load recognition model is sent down and deployed on the miniature circuit breaker device.

[0171] Specifically, according to this solution, a large amount of load characteristic data and the corresponding second recognition results can be collected in the cloud, which can include the time domain and frequency domain characteristics of various types of loads, as well as the more accurate load types and states obtained through cloud model analysis.

[0172] Next, the load feature data collected by the cloud can be used as samples, and the second recognition result as labels to retrain the first load recognition model, which can help the model learn the mapping relationship between more accurate features and labels, thereby improving the recognition accuracy.

[0173] After the training is completed, an updated first load recognition model can be obtained. This model contains new knowledge and recognition capabilities obtained from cloud data analysis and has better performance than the previous model. The updated first load recognition model can be sent from the cloud to the miniature circuit breaker device. This usually involves the transmission of model parameters, and data compression and encryption may need to be considered to improve transmission efficiency and security.

[0174] Finally, the updated first load recognition model can be deployed on the miniature circuit breaker device. This may involve loading new model parameters on the device and performing necessary configuration and testing to ensure that the model can run normally on the device. The updated first load recognition model starts to run on the miniature circuit breaker device, processes and recognizes the load feature data collected in real time, and provides more accurate recognition results.

[0175] Through the above method, this application realizes the collaborative work between the cloud and the edge side, and uses the powerful computing power and data analysis ability of the cloud to improve the performance of the edge side model. The model update and deployment mechanism not only improves the recognition accuracy, but also enhances the adaptability and intelligence level of the system, enabling the miniature circuit breaker device to better serve the power consumption management and safety monitoring of smart grids and intelligent buildings.

[0176] As Figure 4 shown below, the process of cloud-edge-end integrated load recognition is introduced, showing a flowchart of a non-intrusive load recognition system, which is divided into two main parts: edge side (local) processing and cloud processing. The following is a simple introduction to this flowchart:

[0177] First, during edge side processing, electrical equipment is connected to the system and starts using electricity. The system collects the current and voltage signals of the connected electrical appliances. The collected original current and voltage signals are preprocessed, such as denoising and filtering, to improve the signal quality. Time-domain and frequency-domain features are extracted from the preprocessed signals, and these features can describe the electricity consumption characteristics of the electrical appliances. The first load recognition model deployed locally is used to analyze the extracted features to obtain a preliminary load recognition result. Finally, according to the first recognition result, it is judged whether there is illegal electricity consumption behavior. If a violation is detected, the local alarm mechanism is triggered, and control actions may be executed or alarm information may be uploaded.

[0178] During cloud processing, the communication module is responsible for communicating with the end-side device, using the 485 or Modbus protocol. The sending end of the intelligent gateway can send the feature data and the first recognition result obtained from the end-side processing to the cloud through the intelligent gateway. The receiving end of the intelligent gateway receives data from the end-side. The cloud server uses a more complex second load recognition model to deeply analyze the uploaded data and obtain a second recognition result. Finally, the cloud analysis is completed to generate the second recognition result, which may be more accurate than the first recognition result.

[0179] In summary, through this architecture of collaborative work between the end-side and the cloud, this application can achieve more efficient and accurate load recognition and management. The end-side device is responsible for real-time monitoring and preliminary recognition, while the cloud provides more in-depth analysis and optimization. The combination of the two improves the recognition accuracy and reliability of the entire system.

[0180] In a possible example, in order to better implement the non-intrusive single-load recognition method in the embodiments of this application, on top of the non-intrusive single-load recognition method, this application also provides a non-intrusive miniature circuit breaker device, such as Figure 6 shown, the non-intrusive miniature circuit breaker device includes:

[0181] A metering chip 201, configured to acquire an analog signal of the load to be recognized and convert the analog signal into a digital signal;

[0182] A feature extraction module 202, configured to perform time-domain and frequency-domain feature extraction on the digital signal to obtain load feature data;

[0183] A load recognition module 203, configured to input the load feature data into the first load recognition model deployed in the miniature circuit breaker device to recognize the information of the load to be recognized and obtain a first recognition result;

[0184] A cloud-edge-end collaborative architecture 204, configured to upload at least part of the load feature data and the first recognition result to the cloud for analysis to obtain a second recognition result.

[0185] In the embodiments of this application, the metering chip 201, the feature extraction module 202, the load recognition module 203, and the cloud-edge-end collaborative architecture 204 can be respectively used to execute the steps S101-S104 of the foregoing method embodiments. For more detailed content or specific implementation manners of each functional module, reference can be made to the description of the corresponding method steps, which will not be elaborated here.

[0186] The embodiments of this application also provide a computer device. For details, please refer to Figure 7 , Figure 7 which is the basic structural schematic diagram of the computer device in this embodiment.

[0187] The computer device includes a memory 310 and a processor 320 that are communicatively connected to each other via a system bus. It should be noted that only the computer device with components 310 - 320 is shown in the figure, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Among them, those skilled in the art of the present technology can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0188] The computer device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touchpad or a voice control device and other means.

[0189] The memory 310 includes at least one type of readable storage medium, which includes non-volatile memory or volatile memory, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. The RAM can include static RAM or dynamic RAM. In some embodiments, the memory 310 can be an internal storage unit of a computer device, such as the hard disk or memory of the computer device. In other embodiments, the memory 310 can also be an external storage device of the computer device, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device. Of course, the memory 310 can also include both the internal storage unit and the external storage device of the computer device. In this embodiment, the memory 310 is generally used to store the operating system installed on the computer device and various application software, such as the program code of the above method. In addition, the memory 310 can also be used to temporarily store various data that have been output or will be output. In some embodiments, the memory 310 can also be a register.

[0190] The processor 320 is generally used to execute the overall operations of the computer device. In this embodiment, the memory 310 is used to store program code or instructions, and the program code includes computer operation instructions. The processor 320 is used to execute the program code or instructions stored in the memory 310 or process data, such as running the program code of the above method.

[0191] In this text, the bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. This bus system can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, only a thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0192] Another embodiment of the present application further provides a computer-readable medium, which can be a computer-readable signal medium or a computer-readable medium. A processor in the computer reads the computer-readable program code stored in the computer-readable medium, so that the processor can execute the functional actions specified in each step or the combination of steps in the above method; and generate a device for implementing the functional actions specified in each block or the combination of blocks in the schematic diagram.

[0193] The computer-readable medium includes but is not limited to electronic, magnetic, optical, electromagnetic, infrared memories or semiconductor systems, devices or apparatuses, or any suitable combination of the foregoing. The memory is used to store program codes or instructions, and the program codes include computer operation instructions. The processor is used to execute the program codes or instructions of the above method stored in the memory.

[0194] For the definitions of the memory and the processor, reference can be made to the description of the foregoing computer device embodiments, which will not be elaborated here.

[0195] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the devices or units can be in electrical, mechanical or other forms.

[0196] In each embodiment of the present application, each functional unit or module can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0197] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods in various embodiments of this application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0198] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The "including" described in this application does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In the claims listing several units of a device, several of these units of the device can be embodied by the same item of hardware. The use of the first, second, and third, etc. does not denote any order and these words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

[0199] The above embodiments are only used to illustrate the technical solutions of this application, rather than to limit them; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of this application.

Claims

1. A non-invasive single load identification method, characterized in that Applied to a miniature circuit breaker device, the method includes: Obtain an analog signal of a load to be recognized, and convert the analog signal into a digital signal; Extract time-domain and frequency-domain features from the digital signal to obtain load feature data; Input the load feature data into a first load recognition model deployed in the miniature circuit breaker device to recognize information of the load to be recognized, and obtain a first recognition result; Upload at least part of the load feature data and the first recognition result to the cloud for analysis to obtain a second recognition result.

2. The method according to claim 1, wherein The miniature circuit breaker device includes a current detection device, a voltage detection device, and an analog-to-digital converter; obtaining an analog signal of a load to be recognized and converting the analog signal into a digital signal includes: Collect the current signal of the load to be recognized through the current detection device; Collect the voltage signal of the load to be recognized through the voltage detection device; Convert the current signal and the voltage signal into the digital signal through the analog-to-digital converter.

3. The method according to claim 2, wherein The extracting time-domain and frequency-domain features from the digital signal to obtain load feature data includes: Perform time-domain analysis on the characteristics of the voltage signal and the current signal changing with time to obtain a time-domain signal, and extract time-domain features from the time-domain signal; Convert the time-domain signal into a frequency-domain signal, and extract frequency-domain features from the frequency-domain signal.

4. The method according to claim 1, characterized in that, The first recognition result and the second recognition result at least include the type and access status of the load to be recognized.

5. The method according to claim 4, wherein The first load recognition model is a tree model based on machine learning, and the method further includes a training step for the tree model, specifically including: Obtain a load with a known load type and access status, and extract the load feature data; Use the load feature data as a sample, and use the load type and access status as labels to train an initial tree model to obtain the trained tree model and deploy it on the miniature circuit breaker device.

6. The method according to claim 1, characterized in that, The uploading at least part of the load feature data and the first recognition result to the cloud for analysis to obtain a second recognition result includes: Transmit the at least part of the load feature data and the first recognition result to an intelligent gateway, and upload them to the cloud through the Ethernet protocol; Through a second load recognition model deployed in the cloud, analyze and optimize the uploaded data to obtain the second recognition result; Wherein, the performance of the second load recognition model is better than that of the first load recognition model.

7. The method according to claim 1, characterized in that The method further includes: Perform a preliminary recognition on the first recognition result. If the first recognition result indicates that the load to be recognized uses electricity illegally, trigger a local alarm mechanism; Compare the first recognition result with the second recognition result. If the two are consistent, confirm to use the first recognition result as the final recognition result of the load; If the two are inconsistent, update the first load recognition model in the cloud.

8. The method according to claim 7, wherein The updating the first load recognition model in the cloud includes: Use the load feature data in the cloud as a sample, and use the second recognition result as a label to train the first load recognition model to obtain an updated first load recognition model; Deploy and install the updated first load recognition model to the miniature circuit breaker device.

9. A miniature circuit breaker device, characterized in that, It includes: A metering chip, configured to obtain the analog signal of the load to be recognized and convert the analog signal into a digital signal; A feature extraction module, configured to perform time-domain and frequency-domain feature extraction on the digital signal to obtain load feature data; A load recognition module, configured to input the load feature data into the first load recognition model deployed in the miniature circuit breaker device, recognize the information of the load to be recognized, and obtain a first recognition result; A cloud-edge-end collaborative architecture, configured to upload at least part of the load feature data and the first recognition result to the cloud for analysis to obtain a second recognition result.

10. A computer device, characterized in that, It includes: A processor and a memory; The memory is configured to store a computer program; The processor is configured to execute the computer program stored in the memory to implement the steps of the non-intrusive single load recognition method according to any one of claims 1-8.

Images