Intelligent voltage protection system and method based on Internet of Things

Abstract

The invention provides an intelligent voltage protection system and method based on the Internet of Things, and the method comprises the steps: carrying out the equipment action recognition according to behavior association data, and obtaining an equipment external action recognition result of target equipment; performing equipment action recognition according to the internal voltage state data to obtain an internal action recognition result of the target equipment; performing behavior consistency check according to the external action recognition result and the internal action recognition result of the equipment to obtain an external behavior consistency degree, performing dynamic behavior prediction on the target equipment according to the image sensing data and the external behavior consistency degree, and determining an equipment behavior prediction result; the equipment voltage protection threshold value is dynamically adjusted according to the behavior prediction result, intelligent voltage protection is performed on the target equipment through the equipment voltage protection threshold value, the behavior prediction of the target equipment can be performed through the image sensing data of the Internet of Things, and the voltage protection threshold value is dynamically adjusted according to the behavior prediction result. Therefore, voltage false alarm caused by frequent start and stop is avoided.

Description

Technical Field

[0001] The present application relates to the field of voltage protection technology, and more specifically, to an intelligent voltage protection system and method based on the Internet of Things. Background Art

[0002] Voltage protection is a relay protection method that uses the measured voltage at the installation point of the protection device as the action quantity. The relay protection formed by excessively low or high voltage is called low voltage protection and overvoltage protection respectively. The operating voltage of the power system is an important indicator of power quality. When it deviates from the rated voltage beyond a certain limit, it may disrupt the stable operation of the system, affect the quantity and quality of factory products, and may also cause damage to electrical equipment. With the advancement of voltage protection technology, relay protection computing intelligent analysis and early warning systems based on the Internet of Things and big data have emerged, providing new technical support for the stable operation of the power system.

[0003] In the existing technology, the electrical parameters of the target equipment are mainly detected through the Internet of Things, and then voltage protection is performed through fixed threshold judgment to prevent the operating voltage of the target equipment from deviating from the rated voltage by more than a certain limit. However, during periods of concentrated industrial electricity consumption, due to the frequent start and stop of large equipment such as motors, air compressors, welding equipment, etc., short-term fluctuations or even impact changes in current and voltage will be caused. Such fluctuations are transient and predictable. The fixed voltage protection threshold can easily misjudge these short-term changes as voltage anomalies and trigger voltage false alarms. Therefore, how to solve the voltage false alarms caused by frequent start and stop of target equipment has become an urgent problem to be solved. Summary of the Invention

[0004] The present application provides an intelligent voltage protection system and method based on the Internet of Things, which can predict the behavior of target equipment through Internet of Things image sensor data, and dynamically adjust the voltage protection threshold according to the behavior prediction results, thereby avoiding false voltage alarms caused by frequent starts and stops.

[0005] In a first aspect, the present application provides an intelligent voltage protection method based on the Internet of Things. The method can be executed by a network device, or can also be executed by a chip configured in the network device. The present application does not limit this.

[0006] Specifically, the method includes: Collecting external behavior association data of the target device through the Internet of Things, performing device action recognition based on the external behavior association data, and obtaining a device external action recognition result of the target device; Collecting internal voltage state data of a target device, performing device action recognition based on the internal voltage state data, and obtaining an internal action recognition result of the target device; Performing a behavior consistency check based on the device's external action recognition result and the internal action recognition result to obtain an external behavior consistency; When the external behavior consistency is higher than a preset trustworthy threshold, acquiring image sensor data of the target device, performing dynamic behavior prediction on the target device based on the image sensor data and the external behavior consistency, and determining a device behavior prediction result; The device voltage protection threshold is dynamically adjusted according to the behavior prediction result, and the target device is intelligently voltage-protected by the device voltage protection threshold.

[0007] In combination with the first aspect, in certain implementations of the first aspect, the internal voltage status data is collected by a voltage sensor installed in a power supply line loop of the target device.

[0008] In conjunction with the first aspect, in certain implementations of the first aspect, performing device action recognition based on the internal voltage state data to obtain an internal action recognition result of the target device specifically includes: Constructing a time series sample set of a voltage signal using a sliding time window according to the internal voltage state data; Extract voltage features from the voltage data in each time window, and construct voltage feature vectors based on multiple voltage features corresponding to each time window to obtain a feature vector set; Each voltage feature vector in the feature vector set is subjected to feature classification through a trained neural network to obtain an internal action state label corresponding to each time window, thereby obtaining an internal action recognition result.

[0009] In conjunction with the first aspect, in certain implementations of the first aspect, performing a behavior consistency check based on the device external action recognition result and the internal action recognition result to obtain the external behavior consistency specifically includes: Perform code value mapping based on the device external action recognition result label corresponding to each time window to obtain the corresponding external action recognition code sequence; Map the code values ​​according to the device internal action recognition result labels corresponding to each time window to obtain the corresponding internal action recognition code sequence; A consistency check is performed based on the external action recognition coding sequence and the internal action recognition coding sequence to determine the external behavior consistency.

[0010] In conjunction with the first aspect, in certain implementations of the first aspect, performing dynamic behavior prediction on the target device based on the image sensing data and the external behavior consistency, and determining the device behavior prediction result specifically includes: After preprocessing the image sensing data, time series image features are extracted using a convolutional neural network; The behavior prediction depth of the target device is determined according to the external behavior consistency, the time series image features are modeled using a moving average autoregressive model, and the device behavior prediction result of the target device is determined using the moving average autoregressive model.

[0011] In combination with the first aspect, in some implementations of the first aspect, the external behavior-related data includes: external image data, external sound data, and external vibration data of the target device.

[0012] In conjunction with the first aspect, in certain implementations of the first aspect, performing device action recognition based on the external behavior association data to obtain a device external action recognition result of the target device specifically includes: acquiring external image data, external sound data, and external vibration data from the behavior-related data; performing image action recognition on the external image data to determine a first action recognition result; performing device sound recognition on the external sound data to determine a second action recognition result; performing vibration abnormality recognition on the external vibration data to determine a third action recognition result; A fusion decision is performed based on the first, second, and third action recognition results to determine an external device action recognition result of the target device.

[0013] In a second aspect, the present application provides an Internet of Things-based intelligent voltage protection system, which includes a voltage protection unit, and the voltage protection unit includes: A device action recognition module is used to collect external behavior association data of a target device through the Internet of Things, perform device action recognition based on the external behavior association data, and obtain a device external action recognition result of the target device; The device action recognition module is further configured to collect internal voltage status data of a target device, perform device action recognition based on the internal voltage status data, and obtain an internal action recognition result of the target device; a decision module, configured to perform a behavior consistency check based on the device's external action recognition result and the internal action recognition result to obtain an external behavior consistency; a device behavior prediction module, configured to, when the external behavior consistency is higher than a preset trustworthy threshold, obtain image sensor data of the target device, dynamically predict the behavior of the target device based on the image sensor data and the external behavior consistency, and determine a device behavior prediction result; The voltage protection adjustment module is used to dynamically adjust the device voltage protection threshold according to the behavior prediction result, and perform intelligent voltage protection on the target device by the device voltage protection threshold.

[0014] In a third aspect, the present application provides a computer terminal device, which includes a memory and a processor, wherein the memory stores a code, and the processor is configured to obtain the code and execute the above-mentioned intelligent voltage protection method based on the Internet of Things.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium, which stores at least one computer program, and the computer program is loaded and executed by a processor to implement the operations performed by the above-mentioned IoT-based intelligent voltage protection method.

[0016] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The present application provides an Internet of Things-based intelligent voltage protection system and method, which collects external behavior association data of a target device through the Internet of Things, performs device action recognition based on the behavior association data, and obtains an external action recognition result of the target device; collects internal voltage status data of the target device, performs device action recognition based on the internal voltage status data, and obtains an internal action recognition result of the target device; performs a behavior consistency check based on the external action recognition result and the internal action recognition result of the device to obtain an external behavior consistency; when the external behavior consistency is higher than a preset trustworthy threshold, obtains image sensor data of the target device, performs dynamic behavior prediction of the target device based on the image sensor data and the external behavior consistency, and determines the device behavior prediction result; dynamically adjusts the device voltage protection threshold based on the behavior prediction result, and performs intelligent voltage protection on the target device based on the device voltage protection threshold.

[0017] Therefore, it can be seen that the present application, by collecting external behavior-related data through IoT sensors, can accurately identify whether the device is in a running, starting, stopping, or abnormal state from a multimodal perspective, avoiding the risk of misjudging the device state based on a single voltage signal, and improving the system's ability to understand actual device behavior. By performing consistency checks on the external behavior recognition results with the internal voltage state recognition results and calculating the external behavior consistency, it ensures that in-depth predictions are only made when the external observation results are consistent with the actual voltage state, fundamentally reducing mispredictions and protection failures caused by perception bias. The device behavior prediction is adaptively adjusted based on the external behavior consistency, improving prediction flexibility and fault tolerance, and reducing the risk of misjudgment. Compared with traditional methods that rely solely on voltage fluctuations for judgment, the present application can capture the preceding characteristics of device operation changes in advance through image sensor data, such as the initial start of motor rotation, indicator light lighting, and changes on the operation panel. This allows for early judgment of the device state, gaining reaction time for dynamic threshold adjustment, and avoiding false alarms caused by threshold hysteresis. In scenarios where the device frequently starts and stops, the voltage alarm threshold can be reasonably raised or lowered based on the prediction results, improving the accuracy and robustness of the voltage protection system.

[0018] In summary, the present application can predict the behavior of target devices through IoT image sensing data, and dynamically adjust the voltage protection threshold according to the behavior prediction results, thereby avoiding false voltage alarms caused by frequent starts and stops. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is an exemplary flow chart of an IoT-based intelligent voltage protection method according to some embodiments of the present application; Figure 2 is a schematic structural diagram of a voltage protection unit according to some embodiments of the present application; Figure 3 This is a structural diagram of a computer terminal device that implements an Internet of Things-based intelligent voltage protection method according to some embodiments of the present application. DETAILED DESCRIPTION

[0020] This application collects external behavior association data of the target device through the Internet of Things, identifies the device action based on the behavior association data, and obtains the device external action recognition result of the target device; collects internal voltage status data of the target device, identifies the device action based on the internal voltage status data, and obtains the internal action recognition result of the target device; performs behavior consistency test based on the device external action recognition result and the internal action recognition result to obtain the external behavior consistency; when the external behavior consistency is higher than a preset trustworthy threshold, obtains the image sensor data of the target device, dynamically predicts the behavior of the target device based on the image sensor data and the external behavior consistency, and determines the device behavior prediction result; dynamically adjusts the device voltage protection threshold based on the behavior prediction result, and performs intelligent voltage protection on the target device by the device voltage protection threshold. It can predict the behavior of the target device through the image sensor data of the Internet of Things, and dynamically adjust the voltage protection threshold based on the behavior prediction result, thereby avoiding false voltage alarms caused by frequent starts and stops.

[0021] In order to better understand the above technical solution, the following will be combined with the accompanying drawings and specific implementation methods to describe the above technical solution in detail. Figure 1 , which is an exemplary flow chart of an IoT-based intelligent voltage protection method according to some embodiments of the present application. The IoT-based intelligent voltage protection method 100 mainly includes the following steps: In step S101, external behavior association data of a target device is collected through the Internet of Things, and device action recognition is performed based on the external behavior association data to obtain a device external action recognition result of the target device.

[0022] It should be noted that the external behavior association data is collected in real time through the Internet of Things sensor network installed around the target device, and is used for non-invasive identification and status judgment of the operating behavior of the target device. The external behavior association data at least includes: external image data, external sound data and external vibration data of the target device, wherein the external image data includes an image sequence or video stream of the target device in operation or standby state, which is used to analyze whether the device has undergone visible movement changes (such as rotation, movement, opening, closing, etc.); the external sound data is the operating noise generated by the target device when it is working, such as high-frequency noise and continuous running sound when the motor starts, etc., which is used to reflect the dynamic state of the device; the external vibration data is the mechanical vibration generated during the operation of the device obtained through a three-axis acceleration sensor, MEMS sensor, etc., which is used to detect whether there is device startup, load fluctuation or abnormal vibration.

[0023] Preferably, in some embodiments, performing device action recognition based on the external behavior association data to obtain a device external action recognition result of the target device specifically includes: acquiring external image data, external sound data, and external vibration data from the behavior-related data; performing image action recognition on the external image data to determine a first action recognition result; performing device sound recognition on the external sound data to determine a second action recognition result; performing vibration abnormality recognition on the external vibration data to determine a third action recognition result; A fusion decision is performed based on the first, second, and third action recognition results to determine an external device action recognition result of the target device.

[0024] In specific implementation, the action recognition of external image data can be achieved by introducing the YOLOV8 target detection model, performing continuous frame difference analysis and feature extraction on the external image data, identifying the motion state of the device in the image, and combining optical flow method, edge detection, regional motion vector and other algorithms to determine whether the target device is in a motion state; the recognition result outputs the corresponding action state label, such as "run", "stop" or "unknown", as the first action recognition result.

[0025] During the action recognition process of external sound data, the collected external sound signal is subjected to Fourier transform or Mel-frequency cepstral coefficient extraction to generate spectrum graph data during the operation of the device. The spectrum data is input into the trained long-short-term memory network model for dynamic time series modeling and cluster analysis, and the current action state label of the motor or device, such as "standby", "starting", "stable operation", etc., is output as the second action recognition result.

[0026] Vibration recognition: By performing wavelet analysis or bandpass filtering on the external vibration data, and extracting time domain and frequency domain features of key frequency bands, including characteristic values ​​such as acceleration mean, frequency peak, and short-time energy, the data are input into a one-dimensional convolutional neural network or other clustering model to complete vibration pattern recognition and output the corresponding vibration action label as the third action recognition result.

[0027] To improve recognition accuracy, the first, second and third action recognition results will be aggregated into a fusion decision module for unified decision making.

[0028] Preferably, the fusion module adopts a confidence-based weighted decision strategy, in which: each action recognition model will calibrate its confidence output during the training process; based on historical verification experiments, dataset quality and model stability, the system assigns fixed confidence weights to image recognition, sound recognition and vibration recognition, for example: Image recognition: weight 0.4; sound recognition: weight 0.35; vibration recognition: weight 0.25; The final recognition result is fused and calculated based on the confidence weighted score to determine the current external behavior recognition status of the target device.

[0029] For example, if the image model identifies the device as "running" with a confidence of 0.92, the sound model is "running" with a confidence of 0.87, and the vibration model is "stopped" with a confidence of 0.78, then the fusion judgment is "running". The confidence weight is calculated as follows: fusion score = (0.4×0.92)+(0.35×0.87)+(0.25×0.22)=0.748. The system can set a fusion threshold (such as 0.65). If it exceeds this threshold, it will be considered as the current valid recognition state.

[0030] In step S102, internal voltage status data of a target device is collected, and device action recognition is performed based on the internal voltage status data to obtain an internal action recognition result of the target device.

[0031] In a preferred embodiment of the present invention, the internal voltage status data refers to the voltage change data reflecting the electrical working status of the equipment, which is monitored and recorded in real time by a voltage acquisition module installed in the power input terminal, control panel or electrical control cabinet of the target equipment, and is used to identify the internal operating behavior of the equipment, including whether it is starting, running, stopping or in an abnormal state.

[0032] Optionally, in some embodiments, the internal voltage status data is collected in real time using a voltage sensor (e.g., a voltage transformer, Hall voltage sensor, Rogowski coil, voltage divider circuit, etc.) installed in the target device's power supply input circuit, with a sampling accuracy of no less than ±1V. In other embodiments, the voltage signal can also be digitally sampled using an analog-to-digital converter module and processed by a microprocessor, edge computing node, or embedded terminal. The sampling frequency can be set based on the response characteristics of the target device, preferably between 1 and 10kHz. The present invention does not limit the specific voltage acquisition hardware model or communication method (e.g., Modbus, CAN, RS485, Wi-Fi). The specific implementation can be adapted to different industrial scenarios and IoT platform architectures.

[0033] Preferably, in some embodiments, performing device action recognition based on the internal voltage state data to obtain an internal action recognition result of the target device specifically includes: Constructing a time series sample set of a voltage signal using a sliding time window according to the internal voltage state data; Extract voltage features from the voltage data in each time window, and construct voltage feature vectors based on multiple voltage features corresponding to each time window to obtain a feature vector set; Each voltage feature vector in the feature vector set is subjected to feature classification through a trained neural network to obtain an internal action state label corresponding to each time window, thereby obtaining an internal action recognition result.

[0034] In specific implementations, based on continuously collected internal voltage state data, a fixed-length sliding time window (e.g., 1s, 2s, or adaptive) is set to construct a time series sample set for the voltage signal. Each time window forms a local voltage data segment containing multiple voltage sampling points. Multidimensional features are then extracted from the voltage data within each time window, including but not limited to: time domain features: maximum, minimum, mean, standard deviation, slope change rate; frequency domain features: FFT spectrum peak, frequency center, bandwidth; mutation features: number of zero crossings, fluctuation amplitude, number of spikes; energy features: short-term energy, waveform entropy, and power estimation. The features extracted from each window form a voltage feature vector. Multiple time windows can form a feature vector set. These feature vectors are input into a pre-trained neural network model (e.g., a convolutional neural network (CNN)). The model identifies the state of the current time window based on historically learned classification boundaries and distribution patterns, and outputs a corresponding internal action state label.

[0035] Among them, the internal action status label may include "start", "run", "stop" and "abnormal", etc. In a specific application example of the present application, the rated voltage of a certain industrial motor equipment is 380V: when the motor starts, the voltage is collected and drops sharply from 380V to 330V, and recovers within 0.8 seconds. This mode is recognized by the system as a typical startup behavior; after starting, the voltage remains within the fluctuation range of 376~384V and lasts for 5 minutes, then the system judges it as "running"; if the voltage is collected to be 0 for 10 seconds, it is determined that the equipment has "stopped"; if the voltage waveform frequently changes suddenly and there are more than 10 spike disturbances within 20ms, the "abnormal" state is output.

[0036] In step S103, a behavior consistency check is performed based on the device external action recognition result and the internal action recognition result to obtain external behavior consistency.

[0037] It should be noted that the external behavior consistency described in this application refers to measuring the degree of match between the external performance of the device and the actual internal action by comparing the consistency between the recognition results of the external perception data (such as images, sounds, vibrations) of the target device and the recognition results of the internal operating status (such as voltage changes). When the external behavior consistency is higher than the preset threshold, it indicates that the external perception data of the device is highly credible, and the dynamic behavior of the target device can be predicted based on the external perception data, and the voltage protection threshold can be dynamically adjusted according to the behavior prediction results.

[0038] Preferably, in some embodiments, performing a behavior consistency check based on the device external action recognition result and the internal action recognition result to obtain the external behavior consistency specifically includes: Perform code value mapping based on the device external action recognition result label corresponding to each time window to obtain the corresponding external action recognition code sequence; Map the code values ​​according to the device internal action recognition result labels corresponding to each time window to obtain the corresponding internal action recognition code sequence; A consistency check is performed based on the external action recognition coding sequence and the internal action recognition coding sequence to determine the external behavior consistency.

[0039] In specific implementation, in order to make the recognition results suitable for correlation coefficient analysis, the "external action state label" and "internal action state label" must first be uniformly encoded as numerical variables. For example, the numerical code of the "start" label is 1, the numerical code of the "run" label is 2, the numerical code of the "stop" label is 0, and the numerical code of the "abnormal" label is -1.

[0040] In this application, taking the sliding window length of 1S as an example, the external action recognition coding sequence and the internal action recognition coding sequence are simultaneously obtained for the target device, and the Pearson correlation coefficient is determined based on the external action recognition coding sequence and the internal action recognition coding sequence as the external behavior consistency.

[0041] In a specific application scenario of the present application, in the time series within the past 10 seconds, the external action recognition coding sequence is: X=[1,1,2,2,2,2,2,0,0,0] (start→run→stop), and the internal action recognition coding sequence is: Y=[1,2,2,2,2,2,1,0,0,0] (slightly faster start, brief fluctuations during operation). The Pearson coefficient is calculated as: ρ=0.97. Therefore, it is judged that the external behavior is highly consistent with the internal state, and the recognition result is considered credible, which can be used as the basis for subsequent behavior prediction and voltage protection adjustment.

[0042] In step S104, when the external behavior consistency is higher than a preset trust threshold, the image sensing data of the target device is obtained, and the behavior of the target device is dynamically predicted based on the image sensing data and the external behavior consistency to determine the device behavior prediction result.

[0043] Optionally, in some embodiments, when the external behavior consistency is lower than a preset trustworthy threshold, due to the large difference between the external behavior recognition result and the internal voltage state, indicating that the current data may be unreliable or abnormal, the system may suspend the dynamic prediction of behavior based on the external behavior data to avoid misleading subsequent threshold adjustments. When the behavior prediction is unavailable or unreliable, the system returns to the preset static voltage protection threshold range to ensure that the basic functions of the protection mechanism are not affected and to prevent false operations or omissions.

[0044] Preferably, in some embodiments, the image sensing data of the target device can be collected by an industrial camera. In some other embodiments, it can also be collected by other devices or equipment that can realize target device image collection. This application does not limit this.

[0045] Preferably, in some embodiments, performing dynamic behavior prediction on the target device based on the image sensing data and the external behavior consistency, and determining the device behavior prediction result specifically includes: After preprocessing the image sensing data, time series image features are extracted using a convolutional neural network; The behavior prediction depth of the target device is determined according to the external behavior consistency, the time series image features are modeled using a moving average autoregressive model, and the device behavior prediction result of the target device is determined using the moving average autoregressive model.

[0046] In specific implementations, the collected image sensor data undergoes standardized preprocessing operations to improve the accuracy and stability of subsequent feature extraction and modeling. Preprocessing steps include, but are not limited to: image grayscale conversion and denoising; image scaling and enhancement, which can utilize histogram equalization and edge enhancement; and temporal alignment of image frames.

[0047] It should be noted that this application can effectively improve the accuracy and credibility of the prediction by introducing external behavior consistency as a regulating factor in the device behavior prediction process. When the external behavior consistency is high, it means that the external perception data and the internal operating status are highly matched. At this time, the prediction depth can be increased, and more historical feature data can be used for trend modeling to improve the accuracy of behavior prediction; when the consistency is low, the system can automatically shorten the prediction window or suspend the prediction to avoid misjudgment due to unreliable data. In addition, the mechanism can also dynamically adjust the voltage protection strategy according to the consistency, realize the refinement and adaptive control of intelligent protection, reduce computing resource consumption, enhance the system's perception and response capabilities to potential abnormal behaviors, and significantly improve the stability and intelligence level of the entire voltage protection system.

[0048] The preprocessed image sequence is then input into a convolutional neural network model to extract the spatial-temporal feature representation corresponding to each frame or each number of frames. The convolutional neural network model can be a trained lightweight network (such as MobileNet, ResNet-18, EfficientNet). The extracted feature vectors are organized into a feature sequence in chronological order, and the behavior prediction depth of the target device is then adjusted by the external behavior consistency. Specifically, based on the range of the external behavior consistency, dynamic mapping is performed according to a preset mapping table, and the number of image frames is determined as the corresponding behavior prediction depth. The behavior prediction depth is the moving average modeling window length. The image feature sequence within the prediction depth range is then selected as the input feature, and moving average autoregressive modeling is performed. The predicted image feature vector for the next cycle is then subjected to feature clustering matching (preferably using the K-Means clustering algorithm) with the historical label library to identify the most likely action state corresponding to the current feature in the label space. For example: cluster center A → "running", cluster center B → "about to start", cluster center C → "abnormal vibration", and finally output the device behavior prediction result of the target device, which can also be accompanied by its corresponding behavior label and prediction confidence score to guide the voltage protection system to dynamically adjust the protection threshold.

[0049] The following is a specific embodiment of the present application for determining the device behavior prediction result of the target device by using a moving average autoregressive model: The system collects image data of the target device over a recent period of time in real time from the image sensor. After image preprocessing, the data is input into a trained lightweight convolutional neural network (such as ResNet-18) to extract the spatial-temporal feature representation of each frame image or image sequence, and generate a continuous image feature vector sequence.

[0050] In order to further reduce the dimension and enhance the feasibility of modeling, principal component analysis (PCA) can be performed on the extracted image feature sequence, and the first 1 to 2 principal components can be retained as the image behavior indicator sequence to form a time series.

[0051] Then, the behavior prediction depth is dynamically adjusted in combination with the external behavior consistency. In this embodiment, the following external behavior consistency and prediction depth mapping rules are set: when the external behavior consistency ρ is greater than or equal to 0.9, the system believes that the current external and internal recognition are highly consistent, and the data credibility is high. At this time, the behavior prediction depth is set to 60, that is, the image features in the past 60 minutes are used for behavior modeling and prediction; when the external behavior consistency ρ is between 0.7 and 0.9, the data consistency is good, and the behavior prediction depth is set to 30; when the external behavior consistency ρ is between 0.5 and 0.7, the data consistency is general, and the prediction depth is reduced to 15; when the external behavior consistency ρ is lower than 0.5, it means that there is a large difference between the external and internal recognition results, and the data credibility is low. At this time, long-term prediction is not recommended. You can choose to only perform short-term prediction (for example, the window length is 5) or skip the prediction stage.

[0052] According to the external behavior consistency calculated at the current moment, the corresponding prediction window length is selected, that is, the corresponding number of data points are intercepted from the image behavior indicator sequence to form a prediction sequence.

[0053] Subsequently, the moving average autoregressive series is tested for stationarity (e.g., ADF test). If it is not stationary, first-order difference or exponential smoothing is performed. The autocorrelation coefficient (ACF) and partial autocorrelation coefficient (PACF) plots are then plotted, and the order (p, q) of the ARMA model is determined based on the significant attenuation points in the plots.

[0054] For example, if the ACF is truncated after the third order and the PACF gradually decays after the second order, the order of the ARMA model can be determined to be (3, 2). The parameters can be estimated using the least squares method or maximum likelihood estimation with a confidence level of 0.05.

[0055] The completed ARMA(p,q) model is used to predict the image behavior indicator value at the next moment. If the predicted value is significantly higher than the historical average, it may indicate that the device is about to start up or enter a high-load operation state; if the predicted value decreases, it may indicate that the device is about to shut down or that its operation is becoming stable.

[0056] Finally, the corresponding behavior label (such as "about to start", "continuous operation", "ready to stop" or "abnormal") is matched according to the prediction result and used as the basis for subsequent dynamic voltage protection threshold adjustment.

[0057] It should be noted that compared with the traditional method of judging only by voltage fluctuations, image sensor data can not only better reflect the actual operation and start-up and shutdown conditions of the equipment, but also capture the preceding characteristics of equipment operation changes in advance, such as the initial rotation of the motor, the lighting of the indicator light, changes in the operation panel, etc., so as to judge the equipment status in advance, gain reaction time for dynamic threshold adjustment, and avoid false alarms caused by threshold lag. Therefore, in scenarios where the equipment frequently starts and stops, if the voltage alarm threshold is reasonably increased or lowered according to the prediction results (for example, temporarily relaxing the threshold range when it is predicted that it is about to start), false alarms caused by short-term voltage drops can be avoided; conversely, when it is predicted that the equipment is about to stop or abnormal fluctuations occur, the threshold range is tightened in time to improve the protection sensitivity, effectively improving the accuracy and robustness of the voltage protection system.

[0058] In step S105, the device voltage protection threshold is dynamically adjusted according to the behavior prediction result, and intelligent voltage protection is performed on the target device by the device voltage protection threshold.

[0059] In a preferred embodiment of the present invention, the device behavior prediction results can be used to proactively detect impending operating states of target devices, such as impending startup, increased load, or a high likelihood of shutdown. Based on these predictions, the device's voltage protection threshold is dynamically adjusted to accommodate impending voltage fluctuations, thereby improving the protection system's sensitivity and fault tolerance.

[0060] Specifically, the following steps may be included: Behavior prediction result acquisition: Based on image sensor data, external behavior consistency, and behavior modeling algorithms (such as a moving average autoregressive model), the target device's behavior prediction results for the next prediction period are obtained. The prediction results may include labels such as "about to start," "continuous operation," "planned shutdown," or "possible abnormality." Threshold adjustment strategy matching: Based on different prediction results, the system will invoke the corresponding voltage protection strategy and dynamically adjust the voltage protection threshold. For example, if the prediction result is "about to start," because motor startup is often accompanied by a transient voltage drop, the system will temporarily lower the lower voltage threshold (for example, from 340V to 310V) to avoid false triggering of protection. If the prediction result is "continuous operation," the system will maintain the normal operating voltage protection threshold (e.g., lower limit 340V, upper limit 420V). If the prediction result is "planned shutdown," the system will adjust the upper and lower thresholds to more stringent values ​​to detect voltage fluctuations during abnormal shutdowns in advance. If the prediction result is "possibly abnormal": the system can enable a redundant protection mechanism, increase the sampling frequency and narrow the protection threshold range (such as setting the upper and lower limits to 370V~390V) to achieve more sensitive abnormal voltage detection.

[0061] During the intelligent voltage protection of the target device using the device voltage protection threshold, dynamic voltage protection is specifically implemented by sending the adjusted voltage protection threshold to the voltage monitoring module and control module in real time. When the actual collected voltage value exceeds the dynamically adjusted protection threshold range, the system triggers the corresponding intelligent protection action. For example, the device emergency actions that can be taken include: issuing an alarm signal; controlling the device to power off; pushing voltage anomaly information to the platform monitoring center; activating the backup power supply or switching the power supply path (if the system has UPS or dual power switching capabilities).

[0062] Among them, in the process of intelligent voltage protection of the target device by the device voltage protection threshold, the recovery and adaptive mechanism specifically includes: after the protection action is completed, the system will automatically restore the normal protection threshold setting according to the device recovery status, external behavior recognition and a new round of prediction results to achieve closed-loop control.

[0063] This application can significantly improve the system's adaptability to voltage fluctuations by linking the future operating trends of the equipment with the voltage protection strategy, avoiding false alarms, missed alarms and unnecessary downtime, while enhancing the equipment's operating stability and intelligent protection level under complex working conditions.

[0064] In addition, in another aspect of the present application, in some embodiments, the present application provides an intelligent voltage protection system based on the Internet of Things, the system includes a voltage protection unit, reference Figure 2 , which is a schematic diagram of exemplary hardware and / or software structure of a voltage protection unit according to some embodiments of the present application. The voltage protection unit 200 includes: a device action recognition module 201, a decision module 202, a device behavior prediction module 203, and a voltage protection adjustment module 204, which are described as follows: The device action recognition module 201 is used to collect external behavior association data of the target device through the Internet of Things, perform device action recognition based on the external behavior association data, and obtain a device external action recognition result of the target device; The device action recognition module 201 is further configured to collect internal voltage status data of a target device, perform device action recognition based on the internal voltage status data, and obtain an internal action recognition result of the target device; A decision module 202 is configured to perform a behavior consistency check based on the device external action recognition result and the internal action recognition result to obtain an external behavior consistency degree; The device behavior prediction module 203 is configured to obtain image sensor data of the target device when the external behavior consistency is higher than a preset trust threshold, dynamically predict the behavior of the target device based on the image sensor data and the external behavior consistency, and determine a device behavior prediction result; The voltage protection adjustment module 204 is configured to dynamically adjust a device voltage protection threshold according to the behavior prediction result, and perform intelligent voltage protection on the target device using the device voltage protection threshold.

[0065] The above describes in detail an example of an IoT-based intelligent voltage protection system and method provided in an embodiment of the present application. It can be understood that in order to achieve the above functions, the corresponding device includes a hardware structure and / or software module corresponding to executing each function.

[0066] Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function in the application is executed in hardware or in a computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Therefore, professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0067] In addition, the present application also provides a computer terminal device, which includes a memory and a processor, the memory stores a code, and the processor is configured to obtain the code and execute the above-mentioned intelligent voltage protection method based on the Internet of Things.

[0068] In some embodiments, reference Figure 3 , which is a schematic diagram of the structure of a computer terminal device that implements an intelligent voltage protection method based on the Internet of Things according to some embodiments of the present application. In the above embodiment, an intelligent voltage protection method based on the Internet of Things can be Figure 3 The computer terminal device 300 shown in FIG. 1 is implemented as shown in FIG. 1 , and the computer terminal device 300 includes at least one communication bus 301 , a communication interface 302 , a processor 303 and a memory 304 .

[0069] The processor 303 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more processors for controlling the execution of an IoT-based intelligent voltage protection method in the present application.

[0070] The communication bus 301 may include a path for transmitting information between the aforementioned components.

[0071] Memory 304 may be, but is not limited to, a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer. Memory 304 may be independent and connected to processor 303 via communication bus 301. Memory 304 may also be integrated with processor 303.

[0072] Among them, the memory 304 is used to store the program code for executing the solution of the present application, and the execution is controlled by the processor 303. The processor 303 is used to execute the program code stored in the memory 304. The program code may include one or more software modules. The determination of the external behavior consistency in the above embodiment can be implemented by the processor 303 and one or more software modules in the program code in the memory 304.

[0073] The communication interface 302 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.

[0074] Optionally, the computer terminal device 300 may further include a power supply 305 for providing power to various devices or circuits in the real-time computer terminal device.

[0075] In a specific implementation, as an example, a computer terminal device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. A processor herein may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0076] The aforementioned computer terminal device can be a general-purpose computer terminal device or a dedicated computer terminal device. In a specific implementation, the computer terminal device can be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of this application do not limit the type of computer terminal device.

[0077] In addition, in other aspects of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one computer program, and the computer program is loaded and executed by a processor to implement the operations performed by the above-mentioned intelligent voltage protection method based on the Internet of Things.

[0078] In summary, in an intelligent voltage protection system and method based on the Internet of Things disclosed in an embodiment of the present application, first, external behavior association data of the target device is collected through the Internet of Things, and device action recognition is performed according to the behavior association data to obtain the device external action recognition result of the target device; internal voltage status data of the target device is collected, and device action recognition is performed according to the internal voltage status data to obtain the internal action recognition result of the target device; behavior consistency is checked based on the device external action recognition result and the internal action recognition result to obtain the external behavior consistency; when the external behavior consistency is higher than a preset trustworthy threshold, image sensor data of the target device is obtained, and the behavior of the target device is dynamically predicted according to the image sensor data and the external behavior consistency to determine the device behavior prediction result; the device voltage protection threshold is dynamically adjusted according to the behavior prediction result, and the target device is intelligently voltage protected by the device voltage protection threshold. The target device behavior can be predicted through the image sensor data of the Internet of Things, and the voltage protection threshold can be dynamically adjusted according to the behavior prediction result, thereby avoiding voltage false alarms caused by frequent starts and stops.

[0079] The above description is merely an embodiment of the present application. Common knowledge such as the specific technical solutions or features of the solutions is not described in detail herein. It should be noted that those skilled in the art may make various modifications and improvements without departing from the technical solution of the present application, and these modifications and improvements should also be considered within the scope of protection of the present application. These modifications and improvements will not affect the effectiveness of the implementation of the present application or the practical application of the patent.

[0080] The scope of protection claimed by this application shall be determined by the content of the claims. The specific embodiments and other descriptions in the specification may be used to interpret the content of the claims. Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of the invention. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application is intended to include such modifications and variations.

Claims

1. An intelligent voltage protection method based on the Internet of Things, characterized in that: include: Collecting external behavior association data of the target device through the Internet of Things, performing device action recognition based on the external behavior association data, and obtaining a device external action recognition result of the target device; Collecting internal voltage state data of a target device, performing device action recognition based on the internal voltage state data, and obtaining an internal action recognition result of the target device; Performing a behavior consistency check based on the device's external action recognition result and the internal action recognition result to obtain an external behavior consistency; When the external behavior consistency is higher than a preset trustworthy threshold, acquiring image sensor data of the target device, performing dynamic behavior prediction on the target device based on the image sensor data and the external behavior consistency, and determining a device behavior prediction result; The device voltage protection threshold is dynamically adjusted according to the behavior prediction result, and the target device is intelligently voltage-protected by the device voltage protection threshold.

2. The method according to claim 1, wherein The internal voltage status data is collected by a voltage sensor installed in the power supply line loop of the target device.

3. The method according to claim 1, wherein Performing device action recognition based on the internal voltage state data to obtain an internal action recognition result of the target device specifically includes: Constructing a time series sample set of a voltage signal using a sliding time window according to the internal voltage state data; Extract voltage features from the voltage data in each time window, and construct voltage feature vectors based on multiple voltage features corresponding to each time window to obtain a feature vector set; Each voltage feature vector in the feature vector set is subjected to feature classification through a trained neural network to obtain an internal action state label corresponding to each time window, thereby obtaining an internal action recognition result.

4. The method according to claim 1, wherein Performing a behavior consistency check based on the device's external action recognition result and the internal action recognition result to obtain the external behavior consistency specifically includes: Perform code value mapping based on the device external action recognition result labels corresponding to each time window to obtain the corresponding external action recognition code sequence; Perform code value mapping based on the device internal action recognition result labels corresponding to each time window to obtain the corresponding internal action recognition code sequence; A consistency check is performed based on the external action recognition coding sequence and the internal action recognition coding sequence to determine the external behavior consistency.

5. The method according to claim 1, wherein Performing a dynamic behavior prediction on the target device based on the image sensing data and the external behavior consistency, and determining the device behavior prediction result specifically includes: After preprocessing the image sensing data, time series image features are extracted using a convolutional neural network; The behavior prediction depth of the target device is determined according to the external behavior consistency, the time series image features are modeled using a moving average autoregressive model, and the device behavior prediction result of the target device is determined using the moving average autoregressive model.

6. The method according to claim 1, wherein The external behavior associated data includes: external image data, external sound data, and external vibration data of the target device.

7. The method according to claim 1, wherein Performing device action recognition based on the external behavior association data to obtain a device external action recognition result of the target device specifically includes: acquiring external image data, external sound data, and external vibration data from the behavior-related data; performing image action recognition on the external image data to determine a first action recognition result; performing device sound recognition on the external sound data to determine a second action recognition result; performing vibration abnormality recognition on the external vibration data to determine a third action recognition result; A fusion decision is performed based on the first, second, and third action recognition results to determine an external device action recognition result of the target device.

8. An intelligent voltage protection system based on the Internet of Things, comprising a voltage protection unit, wherein the voltage protection unit is configured to execute the intelligent switching control method for a household energy storage system according to any one of claims 1 to 7, characterized in that: The voltage protection unit includes: A device action recognition module is used to collect external behavior association data of a target device through the Internet of Things, perform device action recognition based on the external behavior association data, and obtain a device external action recognition result of the target device; The device action recognition module is further configured to collect internal voltage status data of a target device, perform device action recognition based on the internal voltage status data, and obtain an internal action recognition result of the target device; a decision module, configured to perform a behavior consistency check based on the device's external action recognition result and the internal action recognition result to obtain an external behavior consistency; a device behavior prediction module, configured to, when the external behavior consistency is higher than a preset trustworthy threshold, obtain image sensor data of the target device, dynamically predict the behavior of the target device based on the image sensor data and the external behavior consistency, and determine a device behavior prediction result; The voltage protection adjustment module is used to dynamically adjust the device voltage protection threshold according to the behavior prediction result, and perform intelligent voltage protection on the target device by the device voltage protection threshold.

9. A computer terminal device, characterized in that: The computer terminal device includes a memory and a processor, the memory stores a code, and the processor is configured to obtain the code and execute the intelligent voltage protection method based on the Internet of Things as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing at least one computer program, characterized in that: The computer program is loaded and executed by a processor to implement the operations performed by the intelligent voltage protection method based on the Internet of Things as described in any one of claims 1 to 7.

Images