Electrical equipment remote monitoring and fault diagnosis system based on Internet of Things

Abstract

The invention discloses an electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things. The electrical equipment remote monitoring and fault diagnosis system comprises the following parts: A1, a sensing layer, A2, an edge layer and A3, a cloud platform. The invention relates to the technical field of the Internet of Things, intelligent sensing, edge computing and artificial intelligence, in particular to an electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things, which has the following advantages: 1, multi-dimensional data acquisition and fusion are realized; 2, multi-dimensional data acquisition and fusion are realized, and compared with traditional cloud centralized processing, the response speed is remarkably improved; 3, a semi-supervised transfer learning model is adopted, dependence on a large amount of labeled data is reduced, and the problem of scarcity of actual fault data is effectively solved; a self-calibration module is integrated, sensor drifting is dynamically compensated through Kalman filtering, and measurement accuracy is kept for a long time; and 5, a lightweight encryption transmission technology is adopted, so that the data transmission security and the equipment operation efficiency are improved while the data security is guaranteed.

Description

Technical Field

[0001] The present invention relates to the fields of Internet of Things, intelligent sensing, edge computing and artificial intelligence technology, and in particular to an electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things. Background Art

[0002] In fields such as industrial production and power supply, the stable operation of electrical equipment is crucial. Traditional electrical equipment monitoring methods have many shortcomings. First, most monitoring systems can only collect single electrical parameters (such as current and voltage) and lack comprehensive analysis of multi-dimensional data such as vibration and soundprint, making it difficult to comprehensively and accurately determine equipment failures. Second, data processing often relies on centralized cloud processing, resulting in high transmission latency and failing to meet the needs of real-time monitoring and early warning. Furthermore, existing systems lack effective self-calibration mechanisms to address sensor drift caused by long-term operation, leading to accumulated measurement errors and affecting diagnostic accuracy. Furthermore, with the widespread adoption of IoT technology, the secure transmission of electrical equipment data has become a pressing issue. Traditional encryption methods consume a lot of computing power and are unsuitable for resource-constrained edge devices. Therefore, a more advanced system is urgently needed to address these issues and improve the efficiency and accuracy of electrical equipment monitoring and fault diagnosis. Summary of the Invention

[0003] In view of this, the present invention aims to provide an electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things, which realizes remote real-time monitoring, accurate fault diagnosis and efficient operation and maintenance management of electrical equipment through multi-source data fusion, edge computing, intelligent algorithms and other technologies.

[0004] The technical solution of the embodiment of the present invention is achieved as follows:

[0005] The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things includes the following parts:

[0006] A1. Perception layer: Deploys a multi-source sensor array and integrates a self-calibration module to collect multi-dimensional data such as electrical parameters, voiceprint signals, and vibration signals of electrical equipment, and dynamically compensates for sensor drift through the Kalman filter algorithm.

[0007] A2. Edge layer: Set up edge computing nodes to perform real-time filtering and processing of perception layer data, semi-supervised transfer learning fault diagnosis, and edge resource optimization to achieve local rapid decision-making;

[0008] A3. Cloud platform: used to store historical data, integrate multi-source features through DS evidence theory to comprehensively assess device health status, and regularly update edge layer diagnostic models;

[0009] The multi-source sensor array includes at least a current sensor, a voltage sensor, a temperature sensor, a voiceprint sensor and a vibration sensor.

[0010] Preferably, the self-calibration module in A1 is based on a Kalman filter algorithm and satisfies the following state equation and observation equation:

[0011] Equation of state: x k =Ax k-1 +w k-1 ;

[0012] Observation equation: z k =Hx k +v k ;

[0013] Among them, x k is the true value of the sensor at the kth moment, z k is the measurement value, A and H are the state transfer matrix and observation matrix, w k-1 、v k are process noise and observation noise.

[0014] Preferably, the edge computing node in A2 uses the least mean square (LMS) algorithm to perform adaptive filtering on the original data, and the filter coefficient update formula is:

[0015] w(n+1)=w(n)+2μe(n)x(n);

[0016] Where w(n) is the nth iteration filter coefficient, μ is the step size factor, e(n) is the error signal, and x(n) is the input signal.

[0017] Preferably, the semi-supervised transfer learning model in A2 combines domain adversarial training (DANN) and uncertainty-aware pseudo-label selection (UPS), and the loss function is: L = L cls +λL adv Among them, L cls is the classification loss, L adv is the domain adversarial loss, and λ is the balancing weight coefficient.

[0018] Preferably, the edge resource optimization module in A2 establishes a joint optimization objective function based on digital twins and whale optimization algorithm: min(αP+βP+γU);

[0019] Among them, P is the node power consumption, D is the data processing delay, U is the resource allocation imbalance rate, and α, β, and γ are weight coefficients.

[0020] Preferably, the cloud platform in A3 adopts DS evidence theory to fuse multi-source features, and the combination rule is:

[0021] Among them, m1 and m2 are basic probability distribution functions of different data sources. is the conflict coefficient.

[0022] Preferably, the voiceprint sensor in A1 is an adaptive filtering voiceprint sensor, which can extract discharge characteristics in the 10-20kHz frequency band in a 100dB noise environment; the vibration sensor is used to collect mechanical vibration signals of the equipment, and the frequency range covers 10Hz-10kHz.

[0023] Preferably, the edge computing node and the perception layer sensor are connected via the LoRa wireless communication protocol, with a transmission distance ≥ 2km and a single-node power consumption ≤ 50mW; the cloud platform adopts containerized deployment (Docker+Kubernetes) and supports elastic expansion.

[0024] Preferably, the edge computing nodes undertake ≥90% of real-time diagnostic tasks, the data transmission volume is reduced by ≥70% compared with traditional cloud-based centralized processing, and the diagnostic delay is ≤50ms; the cloud platform performs fusion analysis on multi-source data every day, updates the equipment health score and predicts the remaining life, and the analysis cycle is ≤10min.

[0025] Preferably, the system also includes a data encryption transmission module, which filters valid data through hash value comparison, fills in empty values ​​when the predicted value matches the measured value hash, otherwise uses the AES-128 algorithm to encrypt and transmit, reducing the edge device CPU load by more than 30%.

[0026] The embodiment of the present invention adopts the above technical solution, which has the following advantages:

[0027] 1. Realize multi-dimensional data collection and fusion, which can capture equipment status more comprehensively and accurately than traditional single parameter monitoring, and improve fault diagnosis accuracy by more than 20%.

[0028] 2. Edge computing nodes process data in real time, reducing transmission delays to less than 50ms, meeting real-time warning needs. Compared with traditional cloud-based centralized processing, the response speed is significantly improved.

[0029] 3. Use a semi-supervised transfer learning model to reduce dependence on large amounts of labeled data, effectively solve the problem of scarcity of actual fault data, and improve the model's generalization ability by more than 15%.

[0030] 4. An integrated self-calibration module uses Kalman filtering to dynamically compensate for sensor drift, maintaining long-term measurement accuracy, extending the sensor calibration cycle from three months to one year, and reducing manual calibration costs.

[0031] 5. Use lightweight encryption transmission technology to reduce edge device computing power consumption by 40% while ensuring data security, thereby improving data transmission security and device operation efficiency.

[0032] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 Schematic diagram of the system architecture of the present invention;

[0035] Figure 2 This is a flow chart of edge computing node processing in the present invention;

[0036] Figure 3 This is a flow chart of the overall architecture of the present invention;

[0037] Figure 4 This is a flow chart of the overall architecture of the present invention;

[0038] Figure 5 This is the architecture diagram of the semi-supervised transfer learning model of the present invention. DETAILED DESCRIPTION

[0039] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0040] It should be noted that the terms "first," "second," "symmetrical," "array," etc. are used only to distinguish descriptions from positional descriptions and should not be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, limitations on features such as "first" and "symmetrical" may explicitly or implicitly include one or more of these features; similarly, when the number of certain features is not limited in the form of words such as "two" or "three," it should be noted that these features also explicitly or implicitly include one or more of the number of features.

[0041] In the present invention, unless otherwise expressly specified or limited, terms such as "installation," "connection," and "fixation" should be understood broadly; for example, they may refer to fixed connection, detachable connection, or integral molding; they may refer to mechanical connection, direct connection, welding, or indirect connection through an intermediate medium; they may refer to internal communication between two components or interaction between two components. Those skilled in the art will be able to understand the specific meanings of the above terms in the present invention based on the specification and drawings in conjunction with specific circumstances.

[0042] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0043] like Figure 1-5 The present invention provides an electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things, comprising the following parts:

[0044] A1. Perception layer: Deploys a multi-source sensor array and integrates a self-calibration module to collect multi-dimensional data such as electrical parameters, voiceprint signals, and vibration signals of electrical equipment, and dynamically compensates for sensor drift through the Kalman filter algorithm.

[0045] A2. Edge layer: Set up edge computing nodes to perform real-time filtering and processing of perception layer data, semi-supervised transfer learning fault diagnosis, and edge resource optimization to achieve local rapid decision-making;

[0046] A3. Cloud platform: used to store historical data, integrate multi-source features through DS evidence theory to comprehensively assess device health status, and regularly update edge layer diagnostic models;

[0047] The multi-source sensor array includes at least a current sensor, a voltage sensor, a temperature sensor, a voiceprint sensor and a vibration sensor.

[0048] like Figure 1-5 As shown, the self-calibration module in A1 is based on the Kalman filter algorithm and satisfies the following state equation and observation equation:

[0049] Equation of state: x k =Ax k-1 +w k-1 ;

[0050] Observation equation: z k =Hx k +v k ;

[0051] Among them, x k is the true value of the sensor at the kth moment, z k is the measurement value, A and H are the state transfer matrix and observation matrix, w k-1 、v kare process noise and observation noise.

[0052] like Figure 1-5 As shown, the edge computing node in A2 uses the least mean square (LMS) algorithm to adaptively filter the original data, and the filter coefficient update formula is:

[0053] w(n+1)=w(n)+2μe(n)x(n); where w(n) is the nth iteration filter coefficient, μ is the step size factor, e(n) is the error signal, and x(n) is the input signal.

[0054] like Figure 1-5 As shown, the semi-supervised transfer learning model in A2 combines domain adversarial training (DANN) and uncertainty-aware pseudo-label selection (UPS), and the loss function is: L = L cls +λL adv Among them, L cls is the classification loss, L adv is the domain adversarial loss, and λ is the balancing weight coefficient.

[0055] like Figure 1-5 As shown in the figure, the edge resource optimization module in A2 establishes a joint optimization objective function based on digital twins and the whale optimization algorithm: min(αP+βP+γU); where P is the node power consumption, D is the data processing delay, U is the resource allocation imbalance rate, and α, β, and γ are weight coefficients.

[0056] like Figure 1-5 As shown, the cloud platform in A3 uses DS evidence theory to fuse multi-source features, and the combination rule is: Among them, m1 and m2 are basic probability distribution functions of different data sources. is the conflict coefficient.

[0057] like Figure 1-5As shown, the voiceprint sensor in A1 is an adaptive filtering voiceprint sensor that can extract discharge characteristics in the 10-20kHz frequency band in a 100dB noise environment; the vibration sensor is used to collect mechanical vibration signals of the equipment, with a frequency range of 10Hz-10kHz. The edge computing node and the perception layer sensor are connected through the LoRa wireless communication protocol, with a transmission distance of ≥2km and a single-node power consumption of ≤50mW; the cloud platform adopts containerized deployment (Docker+Kubernetes) to support elastic expansion, and the edge computing node undertakes ≥90% of real-time diagnosis tasks. The data transmission volume is reduced by ≥70% compared with traditional cloud centralized processing, and the diagnosis delay is ≤50ms; the cloud platform performs daily fusion analysis of multi-source data, updates the device health score and predicts the remaining life, with an analysis cycle of ≤10min. The system also includes a data encryption transmission module, which filters valid data through hash value comparison and fills null values ​​when the predicted value matches the measured value hash. Otherwise, the AES-128 algorithm is used to encrypt the transmission, reducing the CPU load of the edge device by more than 30%.

[0058] In this embodiment, the present invention is specifically designed to work as follows:

[0059] 1. Hardware Deployment and Installation

[0060] The hardware deployment of this system is divided into three parts: perception layer, edge layer and cloud platform. The specific implementation steps are as follows:

[0061] 1. Perception layer sensor installation:

[0062] Install a multi-source sensor array on the surface or key parts of target electrical equipment (such as switchgear, transformers, motors, etc.):

[0063] Electrical parameter sensor: Use a high-precision Hall current sensor (such as CHB-25NP) to monitor the equipment current, a voltage sensor (such as LV25-P) to collect voltage signals, and an infrared temperature sensor (such as MLX90614) to measure the surface temperature of the equipment non-contactly.

[0064] Voiceprint sensor: Use an adaptive filtering voiceprint sensor (such as the SW-100 model) installed in the gap of the equipment casing or near the heat dissipation vent to capture high-frequency voiceprint signals (10-20kHz frequency band) generated by abnormalities such as partial discharge.

[0065] Vibration sensor: Use a piezoelectric accelerometer (such as CA-YD-103) and fix it to mechanical vibration-sensitive parts such as equipment bearings or machine bases, covering the vibration frequency range of 10Hz-10kHz.

[0066] All sensors are connected to edge computing nodes via LoRa wireless communication modules (such as the SX1278 chip), with a transmission distance of up to 2km. The power consumption of a single sensor is less than 5mW, ensuring long-term stable operation.

[0067] 2. Edge Computing Node Deployment: The edge computing node uses the NVIDIA Jetson Nano development board (a quad-core ARM CPU and a 128-core Maxwell GPU) and is installed in a protective box near the equipment (waterproof and dustproof with an IP65 rating). The node receives sensor data via an RS485 interface or a LoRa module. It has a built-in 16GB eMMC storage for real-time data caching and supports OTA remote upgrades.

[0068] 3. Cloud platform construction:

[0069] The cloud platform is deployed on an Alibaba Cloud ECS server cluster and uses Docker containerization technology (with Kubernetes orchestration) for elastic scalability. A database (MySQL + HBase) is deployed to store historical data, and a data analysis server (Intel Xeon 8255C CPU + NVIDIA T4 GPU) is configured for multi-source data fusion and model training. A web management interface is also provided for operations personnel to monitor device status.

[0070] 2. Software Operation Process

[0071] The system operation is divided into three core links: data collection, edge processing, and cloud analysis. The specific implementation steps are as follows:

[0072] 1. Data collection and preprocessing:

[0073] The sensor collects data synchronously with a period of 10ms (current and voltage are sampled every 10ms, and voiceprint and vibration signals are sampled at a frequency of 10kHz). The edge computing node performs preliminary cleaning of the raw data:

[0074] Remove obvious outliers (such as sudden changes in current values ​​exceeding 2 times the rated value of the device);

[0075] The voiceprint and vibration signals are framed (1024 sampling points per frame) to extract time domain (RMS) and frequency domain (spectral energy) features;

[0076] Perform sliding average filtering on the temperature data to eliminate interference from ambient temperature fluctuations.

[0077] 2. Real-time diagnosis at the edge layer:

[0078] Edge computing nodes use lightweight neural network models (such as MobileNet-V3) to analyze pre-processed data in real time:

[0079] Electrical parameter analysis: Compare the deviation of current and voltage with the equipment rated value to identify anomalies such as overcurrent and undervoltage;

[0080] Voiceprint analysis: Detects discharge characteristics (such as periodic pulse signals) in the 10-20kHz frequency band to determine partial discharge faults;

[0081] Vibration Analysis: This node uses vibration spectrograms to identify mechanical faults such as bearing wear (triple frequency of the characteristic frequency) and rotor imbalance (sudden energy increase at single frequency). If any analysis result indicates an abnormality (confidence level > 90%), the node immediately triggers a local audible and visual alarm (buzzer + LED light) and pushes an alert (including device ID, fault type, and timestamp) to the cloud and the operator's mobile app via the MQTT protocol.

[0082] 3. Cloud-based in-depth analysis and model updates:

[0083] The cloud aggregates all data uploaded by edge nodes daily (including raw data, pre-processed features, and diagnostic results) and optimizes the system through the following steps:

[0084] Multi-source data fusion: Combines historical data on electrical parameters, voiceprints, and vibration, and uses evidence-based reasoning algorithms to comprehensively assess the health status of equipment (e.g., three-level classification: "healthy," "caution," and "fault").

[0085] Remaining life prediction: Analyze the evolution trend of equipment status based on LSTM neural network and predict the remaining useful life of key components (such as bearings and insulation materials);

[0086] Model iteration: Every two weeks, newly collected fault data (such as actual partial discharge cases) is used to train and optimize the edge diagnosis model, which is then pushed to edge nodes via a secure channel (TLS1.3 encryption) to improve model generalization capabilities.

[0087] 3. Implementation Examples of Typical Application Scenarios

[0088] 1. Power substation GIS switch monitoring:

[0089] This system was deployed on the GIS switchgear of a 220kV substation, installing two sets of current, voltage, soundprint, and vibration sensors. Within three months of operation, the system successfully identified two early-stage insulation aging faults (weak 15kHz discharge pulses in the soundprint signal), providing a 30-day advance warning and avoiding two station-wide power outages caused by insulation breakdown. Compared to traditional regular inspections, this system reduced outage duration by 80%.

[0090] 2. Industrial production line motor health management:

[0091] This system was deployed on a motor production line at an automobile manufacturer, monitoring 50 asynchronous motors. Through vibration analysis, the system issued warnings when three motor bearings showed early signs of wear (a 30% increase in vibration energy at 1st harmonic frequency). Operations and maintenance personnel promptly replaced the bearings, avoiding production line downtime due to bearing breakage and reducing equipment downtime by 50%.

[0092] 3. Intelligent building power distribution system optimization:

[0093] This system was deployed in the power distribution room of a commercial complex to monitor the operating status of transformers and distribution boxes. By integrating and analyzing current and temperature data, the system discovered that two transformers were unbalanced in load (one at 85% and the other at 50%). After recommending adjustments to the distribution lines, the average transformer load dropped to 70%, reducing overall energy consumption by 15% and also avoiding potential safety hazards associated with overheating of heavily loaded transformers.

[0094] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various modifications and substitutions within the technical scope disclosed in the present invention, and such modifications and substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things is characterized by: Includes the following sections: A1. Perception layer: Deploys a multi-source sensor array and integrates a self-calibration module to collect multi-dimensional data such as electrical parameters, voiceprint signals, and vibration signals of electrical equipment, and dynamically compensates for sensor drift through the Kalman filter algorithm. A2. Edge layer: Set up edge computing nodes to perform real-time filtering and processing of perception layer data, semi-supervised transfer learning fault diagnosis, and edge resource optimization to achieve local rapid decision-making; A3. Cloud platform: used to store historical data, integrate multi-source features through DS evidence theory to comprehensively assess device health status, and regularly update edge layer diagnostic models; The multi-source sensor array includes at least a current sensor, a voltage sensor, a temperature sensor, a voiceprint sensor and a vibration sensor.

2. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The self-calibration module in A1 is based on the Kalman filter algorithm and satisfies the following state equation and observation equation: Equation of state: x k =Ax k-1 +w k-1 ; Observation equation: z k =Hx k +v k ; Among them, x k is the true value of the sensor at the kth moment, z k is the measurement value, A and H are the state transfer matrix and observation matrix, w k-1 、v k are process noise and observation noise.

3. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The edge computing node in A2 uses the least mean square algorithm to adaptively filter the original data. The filter coefficient update formula is: w(n+1)=w(n)+2μe(n)x(n); Where w(n) is the nth iteration filter coefficient, μ is the step size factor, e(n) is the error signal, and x(n) is the input signal.

4. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The semi-supervised transfer learning model in A2 combines domain adversarial training and uncertainty-aware pseudo-label selection (UPS), and the loss function is: L = L cls +λL adv Among them, L cls is the classification loss, L adv is the domain adversarial loss, and λ is the balancing weight coefficient.

5. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The edge resource optimization module in A2 establishes a joint optimization objective function based on digital twins and the whale optimization algorithm: min(αP+βP+γU); Among them, P is the node power consumption, D is the data processing delay, U is the resource allocation imbalance rate, and α, β, and γ are weight coefficients.

6. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The cloud platform in A3 adopts DS evidence theory to fuse multi-source features, and the combination rule is: Among them, m1 and m2 are basic probability distribution functions of different data sources. is the conflict coefficient.

7. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The voiceprint sensor in A1 is an adaptive filtering voiceprint sensor that can extract discharge characteristics in the 10-20kHz frequency band in a 100dB noise environment; the vibration sensor is used to collect mechanical vibration signals of the equipment, and the frequency range covers 10Hz-10kHz.

8. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The edge computing node and the perception layer sensor are connected via the LoRa wireless communication protocol, the transmission distance is greater than or equal to the preset value K1 and the single node power consumption is less than or equal to the preset value K2; the cloud platform adopts containerized deployment and supports elastic expansion.

9. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1, characterized in that: The edge computing nodes undertake ≥90% of real-time diagnostic tasks, the data transmission volume is reduced by ≥70% compared with traditional cloud-based centralized processing, and the diagnostic delay is ≤50ms; the cloud platform performs fusion analysis on multi-source data every day, updates equipment health scores and predicts remaining life, and the analysis cycle is ≤10min.

10. The electrical equipment remote monitoring and fault diagnosis system based on the Internet of Things according to claim 1, characterized in that: The system also includes a data encryption transmission module, which filters valid data through hash value comparison and fills in null values ​​when the predicted value matches the measured value hash. Otherwise, it uses the AES-128 algorithm to encrypt the data for transmission, reducing the CPU load of the edge device by more than 30%.

Images