A remote monitoring and early warning system for the operating status of energy facilities

Abstract

This application discloses a remote monitoring and early warning system for the operational status of energy facilities, relating to the field of energy facility operational status monitoring technology. In the system, a distributed sensor network collects multi-type status time-series data from various nodes of the energy facility; an edge computing gateway cleans and extracts features from the received data; and a cloud platform is used to: perform quantum dynamic encryption and transmission of feature data packets to obtain plaintext feature data; calculate the coupling coefficient of current feature parameters based on a hidden Markov model and Bayesian network; and then determine the equipment health index and health risk level of each node of the energy facility based on a neural network; and determine corresponding prompt information based on the health risk level, including early warning signals, equipment topology diagrams, and suggestions for collaborative handling of associated equipment. This application can improve monitoring efficiency, ensure the quality of collected data, thereby ensuring the accuracy of early warnings, while also ensuring data security and improving the level of energy facility operation and management.

Description

Technical Field

[0001] This application relates to the field of energy facility operation status monitoring technology, and in particular to a remote monitoring and early warning system for energy facility operation status. Background Technology

[0002] Currently, manual inspection is widely used in power facility monitoring and the operation and management of new energy power plants. However, it has significant drawbacks. For example, manual inspections are time-consuming and cannot promptly detect unexpected problems during facility operation, leading to potential faults remaining undetected for extended periods and affecting the stability of power supply. Furthermore, the accuracy of manual inspection data largely depends on the professional competence and work attitude of the inspectors, making data bias prone to occur. When dealing with large areas of power facilities, manual inspections are inefficient and lack comprehensive coverage, causing many potential faults to go undetected. Once a fault occurs, it can result in severe economic losses and social impact.

[0003] In addition, there is a detection method that combines basic automated control systems to monitor the operating status. However, this method has limited data acquisition frequency and accuracy, and cannot fully capture the detailed operating information of new energy power plants. It has weak real-time perception of the status of new energy power plants, making it difficult to adjust the operating strategy in a timely manner according to the actual situation, thus affecting energy production efficiency. Summary of the Invention

[0004] The purpose of this application is to provide a remote monitoring and early warning system for the operating status of energy facilities, which can improve monitoring efficiency, ensure the quality of collected data, thereby ensuring the accuracy of early warning, while also ensuring data security and improving the level of energy facility operation and management.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] This application provides a remote monitoring and early warning system for the operating status of energy facilities, including a distributed sensor network, an edge computing gateway, and a cloud platform;

[0007] The distributed sensor network is used to: collect multi-type status time-series data of each node of the energy facility and upload it to the edge computing gateway;

[0008] The edge computing gateway is used to: perform sliding window-based data cleaning and wavelet packet decomposition-based feature extraction on the received multi-type state time-series data to obtain feature data packets;

[0009] The cloud platform and the edge computing gateway communicate bidirectionally. The cloud platform is used for: performing quantum dynamic encryption and transmission on the feature data packets to obtain feature plaintext data; calculating the current feature parameter coupling coefficient based on the feature plaintext data using a hidden Markov model and a Bayesian network; determining the equipment health index of each node of the energy facility based on the current feature parameter coupling coefficient using a neural network; determining the health risk level based on the equipment health index of each node; and determining the corresponding prompt information based on the health risk level. The prompt information includes a warning signal, an equipment topology diagram, and suggestions for collaborative handling of related equipment.

[0010] According to the specific embodiments provided in this application, this application achieves the following technical effects: By setting up a distributed sensor network, this application can collect various types of time-series data in real time and accurately, providing a reliable data foundation for subsequent analysis and early warning. Through the setting up of an edge computing gateway, data cleaning is achieved, automatically removing outliers caused by sensor malfunctions or environmental interference, ensuring the accuracy and reliability of the data; feature extraction is implemented to capture feature information within different frequency bands during device operation, providing more refined data input for subsequent correlation analysis and status prediction. By calculating the coupling coefficient of current feature parameters between device operating parameters in real time through a cloud platform, the device health index is predicted and early warning prompts are triggered. The intelligent management method provided in this application can promptly detect potential device malfunctions, prevent the escalation of malfunctions, and reduce economic losses and social impact. Simultaneously, the encrypted communication settings in the cloud platform ensure absolute confidentiality during data transmission, improving system security. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of a remote monitoring and early warning system for the operating status of energy facilities in one embodiment of this application.

[0013] Figure reference numerals: 101-Distributed sensor network; 102-Edge computing gateway; 102a-Data cleaning module; 102b-Feature extraction module; 103-Cloud platform; 103a-Encrypted communication module; 103b-Correlation analysis module; 103c-State prediction module; 103d-Early warning decision module; 104-User terminal. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] This application system can collect data accurately in real time, clean and extract features to ensure data quality, and the cloud platform's intelligent management can promptly detect potential faults. Encrypted communication ensures data security, effectively solving the problems of existing monitoring methods and improving the operation and management level of energy facilities.

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] In one exemplary embodiment, such as Figure 1 As shown, a remote monitoring and early warning system for the operation status of energy facilities is provided, including a distributed sensor network 101, an edge computing gateway 102, and a cloud platform 103.

[0018] (i) The distributed sensor network 101 is used to: collect multi-type status time-series data of each node of the energy facility and upload it to the edge computing gateway.

[0019] The distributed sensor network 101 includes multiple types of sensors installed at various nodes of the energy facility to collect various types of state time-series data. These sensors include accelerometers, temperature sensors, pressure sensors, and energy consumption metering sensors. All of these sensors have periodic self-calibration functions, such as automatically adjusting the acquisition accuracy within a preset 24-hour cycle to ensure data accuracy and reliability. In a specific application, the accelerometer is model ADXL355, sampling vibration signals in real-time from 1Hz to 100Hz; the temperature sensor is model DS18B20; the pressure sensor is model MPX5010, acquiring temperature and pressure data 10 times per second; and the energy consumption metering sensor has an accuracy of 0.1%, aggregating energy consumption data once per minute.

[0020] Correspondingly, the data acquisition steps every 24 hours are as follows: When the sensor's internal timer reaches its 24-hour cycle, it automatically triggers a calibration command and starts the calibration process. The sensor internally uses a preset reference signal as a standard reference value. The acquired reference signal value is compared with the standard reference value to calculate the offset. Taking the temperature sensor mentioned above as an example, the temperature sensor acquires a reference temperature signal and compares it with the known accurate temperature value to obtain the temperature offset. The ADC conversion coefficient is dynamically adjusted: based on the calculated offset, the conversion coefficient of the analog-to-digital converter (ADC) is dynamically adjusted. The ADC is responsible for converting the analog signal acquired by the temperature sensor into a digital signal. By adjusting the conversion coefficient, measurement errors caused by sensor aging, environmental changes, etc., can be compensated. Calibration parameters are stored in the EEPROM: key parameters generated during the calibration process, such as the offset and the adjusted ADC conversion coefficient, are stored in the electrically erasable programmable read-only memory. These parameters are called in subsequent data acquisition processes to correct the acquired data and ensure its accuracy.

[0021] The various types of sensors establish communication with the edge computing gateway 102 via the Modbus protocol, and collect time-series data of equipment vibration, temperature, pressure and energy consumption parameters in real time at a frequency of 1Hz-100Hz (which can be set by relevant technicians).

[0022] (ii) The edge computing gateway 102 is used to: perform sliding window-based data cleaning and wavelet packet decomposition-based feature extraction on the received multi-type state time-series data to obtain feature data packets. Specifically, the edge computing gateway includes a data cleaning module 102a and a feature extraction module 102b.

[0023] The data cleaning module 102a is used to: detect and delete outliers based on a sliding window on the received multi-type state time series data, and remove noise data caused by sensor failure or environmental interference to obtain cleaned multi-type state time series data.

[0024] In a specific application, the data cleaning module 102a includes the following data processing steps:

[0025] Calculate the 25th percentile (Q1) and the 75th percentile (Q3). The 25th percentile represents 25% of the data points below this value, and the 75th percentile represents 75% of the data points below this value. The outlier range is determined using the following formula: IQR = Q3 - Q1. Typically, data points less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR are considered outliers. Data points exceeding the outlier range are marked as outliers and subsequently removed during data processing to reduce the impact of noisy data on the analysis results.

[0026] The feature extraction module 102b is used to: extract multi-scale features from the cleaned multi-type state time series data using a wavelet packet decomposition algorithm, and then convert the data according to a preset interface standard to obtain feature data packets. The wavelet packet decomposition algorithm includes a multi-level decomposition strategy, which can adaptively select the number of decomposition levels based on the device type and operating characteristics to capture feature information within different frequency bands and enhance the comprehensiveness of feature representation.

[0027] In a specific application, the wavelet packet decomposition algorithm uses the db4 wavelet basis function, and the dynamic selection strategy for the number of decomposition layers is as follows: for high-frequency vibration signals: decompose 5 layers (frequency resolution 1 / 32); for low-frequency temperature signals: decompose 3 layers (frequency resolution 1 / 8).

[0028] In a specific application, the feature extraction module includes the following data processing steps:

[0029] (1) For any type of state time series data in the cleaned multi-type state time series data, match the corresponding decomposition level based on the preset multi-level decomposition strategy.

[0030] (2) Based on the number of decomposition layers, wavelet decomposition is performed on the current type state time series data to obtain sub-signals of different frequency bands.

[0031] (3) For any frequency band sub-signal, calculate the energy entropy and root mean square value; the energy entropy and the root mean square value constitute the frequency band characteristics. Among them, the energy entropy can reflect the energy distribution of the signal in different frequency bands, while the root mean square value can reflect the intensity characteristics of the signal.

[0032] (4) Normalize all frequency band features of the current type state time series data to map them to the [0,1] interval, eliminate the dimensional differences between different feature values, and facilitate subsequent data analysis and processing; then encapsulate them according to the preset interface standard to obtain feature data packets, such as encapsulating them according to the IEEE1451.4 standard to form a standardized TEDS format feature data packet, so as to upload it to the cloud platform for further analysis.

[0033] (III) The cloud platform 103 and the edge computing gateway 102 communicate bidirectionally. The cloud platform 103 is used to: perform quantum dynamic encryption and transmission on the feature data packets to obtain feature plaintext data; calculate the current feature parameter coupling coefficient based on the feature plaintext data using a hidden Markov model and a Bayesian network; determine the equipment health index of each node of the energy facility based on the current feature parameter coupling coefficient using a neural network; determine the health risk level based on the equipment health index of each node; and determine the corresponding prompt information based on the health risk level. The prompt information includes a warning signal, an equipment topology association diagram, and suggestions for collaborative handling of associated equipment.

[0034] Specifically, the cloud platform 103 includes an encrypted communication module 103a, a correlation analysis module 103b, a status prediction module 103c, and an early warning decision module 103d. Through the settings of these multiple modules, parameter coupling coefficients are calculated, device status is predicted, and early warning decisions are made.

[0035] The encrypted communication module 103a is used to: distribute a key to the edge computing gateway using dynamic encryption technology based on quantum key distribution, so as to realize the encrypted transmission of the feature data packet, and obtain the feature plaintext data by decryption.

[0036] Specifically, the encrypted communication module 103a includes a key management subsystem for securely storing, updating, and distributing quantum keys, ensuring absolute confidentiality during data transmission, and possessing a key leakage detection mechanism. Upon detecting a potential security threat, it immediately initiates an emergency response process. In a specific application, a dedicated quantum key distribution device and a hardware encryption module based on the SM4 algorithm are deployed on a cloud platform server. The software system implements key generation, management, distribution, security monitoring, and emergency response functions. The hardware encryption module based on the Chinese national cryptographic algorithm SM4 performs encrypted key storage and processing. The SM4 algorithm is a symmetric encryption algorithm independently designed in China, possessing high security and encryption efficiency. The encrypted communication module includes the following data processing steps:

[0037] (1) At preset intervals, a new key is generated based on the quantum key distribution protocol, while the old key is discarded; the preset interval can be ten minutes; generating a new key based on the quantum key distribution protocol ensures the randomness and security of the key. After the new key is generated, the old key is immediately discarded to prevent the key from being cracked and causing data leakage.

[0038] (2) The new key is encrypted based on the national cryptographic SM4 algorithm and transmitted to the edge computing gateway.

[0039] (3) Receive the feature data packet uploaded by the edge computing gateway and encrypted with the new key, and calculate the bit error rate. That is, this application detects the security of the key transmission process by monitoring the bit error rate.

[0040] (4) Determine whether the bit error rate is within a preset threshold range; specifically, the preset threshold range can be set to be greater than 5%.

[0041] (5) If the bit error rate is within the preset threshold range, the emergency response process is initiated. Corresponding to the previous step, when the bit error rate is >5%, it is determined that there may be a potential security threat, and the emergency response process is initiated immediately. The emergency response process includes cutting off the current communication link, redistributing the key and authenticating the link, etc., to ensure the security of data transmission.

[0042] (6) If the bit error rate is not within the preset threshold range, the received feature data packet is decrypted based on the national cryptographic SM4 algorithm to obtain the feature plaintext data.

[0043] The correlation analysis module 103b is used to: integrate a Bayesian network into a Hidden Markov Model (HMM) to determine a preset HMM; input the plaintext feature data into the preset HMM to obtain the corresponding current feature parameter coupling coefficients; in short, the correlation analysis module 103b constructs a correlation matrix of device operating parameters to obtain a preset HMM based on multi-source data fusion, and calculates the dynamic coupling coefficients between parameters in real time. Because the preset HMM integrates a Bayesian network structure and introduces prior knowledge to optimize model parameters, it can improve the accuracy and real-time performance of calculating the dynamic coupling coefficients between parameters, providing more reliable input information for state prediction.

[0044] In a specific application, the correlation analysis module includes the following data processing steps:

[0045] (1) Obtain historical multi-type state time series data and corresponding equipment fault status; the historical multi-type state time series data includes: vibration amplitude, temperature and pressure, and the corresponding fault evaluation index system is: vibration amplitude-temperature / pressure-equipment fault status.

[0046] (2) Based on the aforementioned fault evaluation index system, a Bayesian network structure is constructed. Bayesian networks can utilize prior knowledge to optimize model parameters. In energy facilities, there is a certain causal relationship between vibration amplitude, temperature / pressure, and equipment fault status. This prior knowledge can be introduced into the Hidden Markov Model through a Bayesian network.

[0047] (3) Based on the historical multi-type state time series data and the corresponding equipment fault states, train the parameters of the Bayesian network structure to determine the conditional probability table; the conditional probability table includes the probability distribution of vibration amplitude under different temperature and pressure conditions.

[0048] (4) The conditional probability table is used as the prior distribution of the hidden Markov model to provide initial values ​​for parameter estimation of the hidden Markov model.

[0049] (5) The parameters of the Hidden Markov Model are iteratively optimized using the Baum-Welch algorithm to obtain the preset Hidden Markov Model. The Baum-Welch algorithm is an iterative algorithm based on maximum likelihood estimation. By continuously adjusting the state transition probability and observation probability of the Hidden Markov Model, the likelihood of the model to the observation data is maximized, thereby improving the accuracy of the model.

[0050] The state prediction module 103c is used to: construct an LSTM-GRU hybrid neural network; and input the coupling coefficient of the current feature parameter into the LSTM-GRU hybrid neural network to obtain the equipment health index of each node of the energy facility.

[0051] The LSTM-GRU hybrid neural network employs an ensemble learning strategy, combining the prediction results of multiple independently trained neural network models. Weighted averaging or voting mechanisms are used to improve the stability and accuracy of the predictions. The resulting equipment health index ranges from 0 to 100, where 0 indicates complete equipment failure and 100 indicates the equipment is in optimal operating condition, providing a predicted remaining lifespan.

[0052] The state prediction module includes the following data processing steps:

[0053] (1) Train multiple LSTM-GRU models independently using different initialization parameters and training data subsets to obtain multiple LSTM-GRU hybrid neural networks; for example, train 5 different LSTM-GRU models, each using a different 20% training data subset.

[0054] (2) Assign different weights to each hybrid neural network based on the performance of multiple LSTM-GRU hybrid neural networks on the validation data subset; this step and the previous step constitute the training process.

[0055] (3) Input the current feature parameter coupling coefficient into multiple LSTM-GRU hybrid neural networks to obtain the initial prediction result, and then combine the weights of each hybrid neural network to determine the final prediction result by weighted averaging; the final prediction result is the equipment health index.

[0056] The early warning decision-making module 103d is configured to: match the health risk level of the device health index based on a preset health threshold; generate an early warning signal when the health risk level is at the first level; when the health risk level is at the second level, generate a device topology association graph through graph database technology, activate the linkage protection mechanism, and determine the collaborative disposal suggestions for associated devices.

[0057] Among them, the device topology association graph shows the connection relationship and dependency relationship between the components of the device, helping the operation and maintenance personnel quickly locate the source of the fault. The device topology association graph is implemented through graph database technology, supporting efficient storage and query of the physical connection relationship and logical dependency relationship between devices, providing intuitive visual support for the cross-device collaborative protection mechanism. The linkage protection mechanism automatically controls relevant devices or systems according to preset rules, such as reducing the operating power of the device, cutting off some non-critical circuits, etc., to prevent the fault from expanding further.

[0058] Specifically, the early warning decision-making module 103d triggers multi-level alarm signals according to the preset health threshold. When the health index is lower than the safety threshold (i.e., the health risk level is at the second level), a device topology association graph is generated synchronously and the linkage protection mechanism of the control system is activated. Among them, the preset health threshold is dynamically adjusted based on the historical operation status data of the energy facility and expert experience, and has a self-learning function, which can automatically optimize the early warning strategy according to the change trend of the operation status, reducing false alarms and missed alarms. The calculation formula of the preset health threshold is: threshold = α × historical mean + (1 - α) × expert experience value, and α is dynamically adjusted according to the degree of device aging, with a value range of 0.3 - 0.7.

[0059] (4) The system of the present application further includes a user terminal 104; the user terminal 104 is connected to the cloud platform 103, and the user terminal 104 is provided with a visual interface. The user terminal 104 is configured to: receive the prompt information; use three-dimensional modeling technology to display the three-dimensional operation status map of the energy facility; use natural language processing technology to convert the collaborative disposal suggestions for associated devices into suggestion texts (specifically, convert them into text descriptions that are easy to understand to assist users in making decisions quickly); display the suggestion texts, the early warning signals, and the device topology association graph on the three-dimensional operation status map, thereby realizing the display of the device appearance, the status indication of key components, and the highlighting of abnormal areas.

[0060] In a specific practical application, based on the system of this application, multiple types of sensors can be deployed at eight key locations, including the gearbox and generator bearings. The sensors are deployed at these eight key locations, and the edge computing gateway uploads a feature data packet every 5 seconds. When the health index drops to 70, an alert is triggered: a correlation diagram is generated showing the relationship between the gearbox and the converter, and a handling suggestion is sent to the user terminal: "It is recommended to check the gearbox lubricating oil temperature; the threshold has exceeded 110℃."

[0061] Compared with the prior art, this application has the following advantages:

[0062] (1) The distributed sensor network consists of multiple types of sensors deployed at various nodes of the energy facility. It establishes bidirectional communication with the edge computing gateway through the Modbus protocol, and can capture key parameters such as equipment vibration, temperature, pressure and energy consumption in real time and accurately, providing a reliable data foundation for subsequent analysis and early warning.

[0063] (2) The edge computing gateway has a built-in data cleaning module that uses a sliding window-based outlier detection method to remove noisy data. Through data cleaning, the system can automatically remove outliers caused by sensor failures or environmental interference, ensuring the accuracy and reliability of the data. The feature extraction module uses a wavelet packet decomposition algorithm to extract features from the original signal at multiple scales, which can capture feature information in different frequency bands during device operation, providing more refined data input for subsequent correlation analysis and state prediction.

[0064] (3) The cloud platform can calculate the dynamic coupling coefficient between equipment operating parameters in real time, predict equipment health and remaining lifespan, and trigger early warning signals based on preset thresholds. This intelligent management method can promptly detect potential equipment failures, prevent the escalation of failures, and reduce economic losses and social impact. At the same time, the encrypted communication module ensures absolute confidentiality during data transmission, improving system security.

[0065] (4) The user terminal displays the device status visually, receives warnings and handling suggestions, so that users can quickly understand the device status.

[0066] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0067] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A remote monitoring and early warning system for the operating status of energy facilities, characterized in that, The system includes a distributed sensor network, an edge computing gateway, and a cloud platform; The distributed sensor network is used to: collect multi-type status time-series data of each node of the energy facility and upload it to the edge computing gateway; The edge computing gateway is used to: perform sliding window-based data cleaning and wavelet packet decomposition-based feature extraction on the received multi-type state time-series data to obtain feature data packets; The cloud platform communicates bidirectionally with the edge computing gateway. The cloud platform is used to: perform quantum dynamic encryption and transmission on the feature data packet to obtain feature plaintext data. Based on the Hidden Markov Model and Bayesian Network, the coupling coefficient of the current feature parameters is calculated according to the plaintext feature data. Based on the neural network, the equipment health index of each node of the energy facility is determined according to the coupling coefficient of the current feature parameters; and the health risk level is determined based on the equipment health index of each node. Based on the health risk level, corresponding alert information is determined; the alert information includes warning signals, equipment topology diagrams, and suggestions for collaborative handling of related equipment. The cloud platform includes an encrypted communication module, a correlation analysis module, a status prediction module, and an early warning decision module; The encrypted communication module is used to: distribute a key to the edge computing gateway using dynamic encryption technology based on quantum key distribution, so as to realize the encrypted transmission of the feature data packet, and obtain the feature plaintext data by decryption; The association analysis module is used to: integrate a Bayesian network into a hidden Markov model to determine a preset hidden Markov model; input the feature plaintext data into the preset hidden Markov model to obtain the corresponding current feature parameter coupling coefficient; The state prediction module is used to: construct an LSTM-GRU hybrid neural network; input the coupling coefficient of the current feature parameters into the LSTM-GRU hybrid neural network to obtain the equipment health index of each node of the energy facility; The early warning decision module is used to: match the health risk level of the device health index based on a preset health threshold; When the health risk level is classified as Level 1, an early warning signal is generated; When the health risk level is Level 2, a device topology association diagram is generated using graph database technology, a linkage protection mechanism is activated, and recommendations for collaborative handling of associated devices are determined.

2. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The system also includes a user terminal; The user terminal is connected to the cloud platform, and the user terminal is used to: receive the prompt information; use 3D modeling technology to display a 3D map of the operating status of the energy facility; use natural language processing technology to convert the collaborative handling suggestions of the associated equipment into suggestion text; and display the suggestion text, the warning signal, and the equipment topology association diagram on the 3D map of the operating status.

3. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The distributed sensor network includes multiple types of sensors installed at various nodes of the energy facility to collect multiple types of state time-series data; The various types of sensors include an accelerometer, a temperature sensor, a pressure sensor, and an energy consumption metering sensor. The accelerometer, the temperature sensor, the pressure sensor, and the energy consumption metering sensor all have a periodic self-calibration function.

4. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The edge computing gateway includes a data cleaning module and a feature extraction module; The data cleaning module is used to: perform outlier detection and deletion on the received multi-type state time series data based on a sliding window, so as to obtain cleaned multi-type state time series data; The feature extraction module is used to: extract multi-scale features from the cleaned multi-type state time series data using a wavelet packet decomposition algorithm, and then convert the format according to a preset interface standard to obtain a feature data packet.

5. The remote monitoring and early warning system for the operation status of energy facilities according to claim 4, characterized in that, The feature extraction module includes the following data processing steps: For any type of state time series data in the cleaned multi-type state time series data, the corresponding decomposition level is matched based on the preset multi-level decomposition strategy. Based on the number of decomposition layers, wavelet decomposition is performed on the current type state time series data to obtain sub-signals of different frequency bands; For any frequency band sub-signal, calculate the energy entropy and the root mean square value; the energy entropy and the root mean square value constitute the frequency band characteristics. All frequency band features of the current type state time series data are normalized and then encapsulated according to a preset interface standard to obtain a feature data packet.

6. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The encrypted communication module includes the following data processing steps: At preset intervals, a new key is generated based on the quantum key distribution protocol, while the old key is discarded. The new key is encrypted based on the national cryptographic algorithm SM4 and transmitted to the edge computing gateway; Receive the feature data packet uploaded by the edge computing gateway and encrypted with the new key, and calculate the bit error rate; Determine whether the bit error rate is within a preset threshold range; If the bit error rate is within a preset threshold range, an emergency response procedure is initiated. If the bit error rate is not within the preset threshold range, the received feature data packet is decrypted based on the national cryptographic SM4 algorithm to obtain the feature plaintext data.

7. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The correlation analysis module includes the following data processing steps: Acquire historical multi-type state time-series data and corresponding equipment fault status; the historical multi-type state time-series data includes: vibration amplitude, temperature and pressure, and the corresponding fault evaluation index system is: vibration amplitude-temperature / pressure-equipment fault status. Based on the aforementioned fault evaluation index system, a Bayesian network structure is constructed. Based on the historical multi-type state time series data and the corresponding equipment fault states, the parameters of the Bayesian network structure are trained to determine the conditional probability table; the conditional probability table includes the probability distribution of vibration amplitude under different temperature and pressure conditions. The conditional probability table is used as the prior distribution of the Hidden Markov Model to provide initial values ​​for parameter estimation of the Hidden Markov Model. The parameters of the hidden Markov model are iteratively optimized using the Baum-Welch algorithm to obtain the preset hidden Markov model.

8. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The state prediction module includes the following data processing steps: Multiple LSTM-GRU models were trained independently using different initialization parameters and subsets of training data to obtain multiple LSTM-GRU hybrid neural networks; Different weights are assigned to each hybrid neural network based on the performance of multiple LSTM-GRU hybrid neural networks on a subset of validation data. The current feature parameter coupling coefficients are input into multiple LSTM-GRU hybrid neural networks to obtain initial prediction results. Then, the weights of each hybrid neural network are combined to determine the final prediction result by weighted averaging. The final prediction result is the equipment health index.

9. The remote monitoring and early warning system for the operation status of energy facilities according to claim 1, characterized in that, The preset health threshold in the early warning decision module is dynamically adjusted based on the historical operating status data of the energy facility and expert experience, and has a self-learning function, which can automatically optimize the early warning strategy according to the changing trend of the operating status.

Images