Intelligent sensing system and method for offshore wind power tower transformer

Abstract

The invention relates to the technical field of online monitoring of transformers, in particular to an intelligent sensing system and method for an offshore wind power tower transformer. According to the invention, the state information of the offshore tower transformer is monitored in real time, the on-site diagnosis of the transformer fault is realized by means of the convolutional neural network, and the cloud edge pressure is reduced. A domestic intelligent processor based on an RISC-V architecture is adopted, the processor adopts an asymmetric multi-process operation mode, and a plurality of operation systems are operated at the same time to process multiple tasks in parallel, so that the hardware cost is reduced, and the processing of real-time sampling data and the processing of a tower transformer fault recognition algorithm can be ensured. Communication with a cloud server is realized by adopting a LoRa ad hoc network technology and a 5G technology, and the problem of insufficient signal coverage of offshore wind power remote communication is solved. According to the method, the fault occurring in the offshore wind power tower transformer can be effectively judged, the early fault can be processed in time, the serious fault can be effectively avoided, and the safe operation of the tower transformer is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of on-line monitoring of transformers, and particularly to an intelligent perception system and method for an offshore wind power tower transformer. Background Art

[0002] The offshore wind power tower transformer is installed inside the fan tower and operates in a high-salt and highly corrosive offshore environment. The operating load of the fan changes frequently, which has high requirements for the miniaturization, safety and reliability of the transformer. Therefore, there is an urgent need for a miniaturized, low-cost and highly reliable on-line monitoring technology to solve the operation and maintenance needs of offshore tower transformers. The on-line monitoring technology obtains diverse transformer signals in a timely manner by monitoring the status of the tower transformer, and then processes and comprehensively analyzes them. According to the magnitude and change trend of the values, it makes a real-time judgment on the reliability of the equipment, so as to detect potential faults early. At the same time, the offshore wind power tower transformer is far from land and base stations, and the reliability of communication transmission has become a restrictive factor. The cost of laying traditional wired networks is high. Therefore, a reliable, flexible and easy-to-maintain wireless communication method is needed, which does not rely on base stations and has the characteristics of self-organizing network, self-management and self-repair, and meets the multi-link communication and network transmission capabilities under multi-node conditions. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide an intelligent perception system and method for an offshore wind power tower transformer, and provide a miniaturized, low-cost and highly reliable solution to solve the technical problem that the existing methods cannot timely detect and handle the faults of offshore tower transformers.

[0004] To solve the above technical problem, the technical solution adopted by the present invention is: an intelligent perception system for an offshore wind power tower transformer, including a signal acquisition module, a signal analysis module, a fault identification module and a network communication module. The signal acquisition module includes means for acquiring high-speed signals and low-speed signals characterizing the status of the tower transformer; the signal analysis module is connected to the signal acquisition module and is used for preliminarily analyzing and calculating the data collected by the signal acquisition module; the fault identification module is connected to the signal analysis module and is used for receiving the data of the signal analysis module to perform partial discharge fault identification and comprehensive judgment of the transformer status. The signal analysis module and the fault identification module adopt domestic intelligent processors based on the RISC-V architecture. The domestic intelligent processors are multi-core processors, including multiple processor cores and NPUs. The NPU runs convolutional neural networks as a neural network processing unit. Through an asymmetric multi-process operation mode, the multi-core processor is divided into relatively independent cores. An independent core is used as a real-time core, and other cores are used as management cores. Among them, the real-time core is responsible for undertaking the function of the signal analysis module and performing real-time analysis on the data collected by the signal acquisition module. The management core is responsible for managing the data processed by the real-time core and communicating externally. The NPU runs convolutional neural networks to identify and analyze the data processed by the real-time core. The network communication module includes a 5G communication component, a LoRa gateway component, and a LoRa node component. Through the network communication module, self-organizing communication between the intelligent sensing systems of offshore tower transformers and communication with the cloud server are realized.

[0005] Furthermore, the real-time core and the management core run different operating systems independently, and the real-time core and the management core communicate through a shared memory queue.

[0006] Furthermore, the intelligent sensing systems of transformers are wirelessly bridged to each other through the LoRa network to form a LoRa self-organizing network; according to the coverage range of the 5G base station, they serve as LoRa gateways, LoRa relays, and LoRa nodes respectively according to their functions. The network communication module of the LoRa gateway enables the 5G communication component and the LoRa gateway component, and the network communication modules of the LoRa relay and the LoRa node only enable the LoRa node component; during communication, the LoRa terminal transmits the transformer status information to the LoRa gateway through the LoRa relay. The intelligent sensing device of the tower transformer serving as the LoRa relay undertakes the routing and relay work. On the one hand, it forwards the transformer status information of the previous node, and on the other hand, it transmits its own transformer status information to the next LoRa relay. The LoRa gateway aggregates the information of each tower transformer intelligent sensing device and transmits it to the cloud server through 5G.

[0007] Furthermore, the LoRa self-organizing network adopts a dynamic topology management method. According to the network link status, the roles of the LoRa relay and the LoRa node can be switched with each other. The LoRa gateway periodically sends broadcast messages to the LoRa relay and the LoRa node, monitors the reply frames of the LoRa node and the LoRa relay, and updates the routing table of the LoRa node. The messages transmitted by the LoRa node and the LoRa relay to the LoRa gateway adopt a variable frame structure. The message includes the transformer status information and all the LoRa relay and node information in the routing. The LoRa gateway monitors the number of connected terminals in real time, and through the message interaction between multiple gateways, the balanced load of each gateway is realized.

[0008] Furthermore, the signal acquisition module includes a high-speed signal acquisition module and a low-speed signal acquisition module. The high-speed signal acquisition module is used to acquire high-speed signals including transformer partial discharge, transformer core current, transformer bushing voltage and current, and transformer vibration acceleration. The low-speed signal acquisition module is used to acquire low-speed signals including transformer oil temperature, ambient temperature and humidity, and transformer tilt acceleration.

[0009] Furthermore, it also includes a data cache module. The high-speed signal acquisition module is connected to the signal analysis module through the data cache module. The data cache module realizes the first-in-first-out data storage and processing mechanism for high-speed signals through a programmable gate array.

[0010] Further, the signal analysis module analyzes signals in the following ways: processing partial discharge data, calculating the partial discharge spectrogram according to the time-frequency analysis method, calculating the effective value of the core grounding current from the core grounding current data through windowed short-time discrete Fourier transform, calculating the amplitude, phase, and frequency of the bushing voltage and current from the bushing data through windowed short-time discrete Fourier transform; calculating the vibration velocity and vibration displacement from the vibration acceleration; obtaining in real time the low-speed signals including oil temperature, ambient temperature and humidity, and tilt acceleration transmitted by the signal acquisition module, and processing these raw data to calculate the oil temperature, ambient temperature and humidity, and tilt angle.

[0011] Further, the fault identification module identifies faults in the following way: The fault identification module obtains in real time the multi-source data including the partial discharge spectrogram, core current, bushing voltage and current, vibration velocity and displacement, tilt angle, oil temperature, and ambient temperature and humidity calculated by the signal analysis module, uses the improved MobileNetV4 lightweight neural network algorithm to identify partial discharge from the partial discharge spectrogram, then preprocesses the multi-source data including the partial discharge identification result, core current, bushing voltage and current, vibration velocity and displacement, performs index normalization processing, determines the fuzzy membership degree, then constructs a fuzzy judgment matrix, performs consistency verification on the judgment matrix, and finally calculates the index weights to obtain the operating state of the tower transformer.

[0012] Further, encryption technology and authentication mechanisms are adopted between LoRa self-organizing network devices.

[0013] The present invention also discloses an intelligent perception method for an offshore wind power tower transformer, including the following steps: S01. System deployment: Install sensors on the tower transformer to collect multi-source data of the transformer. The multi-source data of the transformer includes partial discharge, vibration acceleration, core grounding, bushing voltage and current, tilt acceleration, oil temperature, and temperature and humidity; deploy the above-mentioned transformer intelligent perception system on the tower transformer; S02. Implement monitoring and data collection: The signal acquisition module obtains in real time the multi-source data of the transformer collected by the sensors, and transmits the collected multi-source data of the transformer to the signal analysis module for processing and analysis through the data cache module in real time; S03. Transformer data processing and analysis. The signal analysis module processes the partial discharge data, calculates the partial discharge spectrogram according to the time-frequency analysis method, calculates the effective value of the core grounding current through the windowed short-time discrete Fourier transform for the core grounding current data, and calculates the amplitude, phase, and frequency of the bushing voltage and current through the windowed short-time discrete Fourier transform for the bushing data; calculates the vibration velocity and vibration displacement through the vibration acceleration; obtains in real time the low-speed signals including oil temperature, ambient temperature and humidity, and tilt acceleration transmitted by the signal acquisition module, and processes these raw data to calculate the oil temperature, ambient temperature and humidity, and tilt angle; S04. Transformer fault identification. The fault identification module conducts partial discharge identification on the partial discharge spectrogram based on a lightweight neural network. The fault identification module conducts fault identification based on multivariate data, and the multivariate data includes partial discharge identification results, core current, bushing voltage and current, vibration, tilt angle, oil temperature, and ambient temperature and humidity. First, the multivariate data is preprocessed, and a preliminary judgment is made according to the initial values and warning values of each index specified by the standard specification. Then, the relative deterioration degree is calculated for index normalization, and the fuzzy membership degree is determined; a fuzzy judgment matrix is constructed according to the membership function, the consistency of the judgment matrix is verified, and the probability of the output result is calculated; the weight of the variable is calculated by the entropy method, and the fuzzy evaluation result is obtained according to the weight index to obtain the status score of the tower barrel transformer; S05. Transformer status uploading. The status information of the tower barrel transformer is transmitted to the cloud server in real time through the network communication module. The network communication module uses the LoRa self-organizing network method. Through the LoRa relay, the transformer status information of each offshore wind power tower barrel transformer intelligent sensing device is aggregated to the LoRa gateway, and the LoRa gateway sends the information to the cloud server through 5G.

[0014] Advantages of the present invention: The intelligent sensing method for offshore wind power tower barrel transformers proposed by the present invention solves the technical problem that the existing methods cannot timely detect and handle the faults of the tower barrel transformers. By online monitoring the status indicators including partial discharge of the tower barrel transformer, core current, vibration velocity and displacement, bushing voltage and current, oil temperature, tilt angle, and ambient temperature and humidity, the deterioration and severity of the tower barrel transformer are analyzed, the faults occurring in the tower barrel transformer are effectively judged, and the information is transmitted to the cloud server through the wireless self-organizing network technology, so as to timely handle the faults and effectively avoid the occurrence of serious faults, ensuring the safe operation of the tower barrel transformer.

[0015] The present invention adopts a low-cost hardware solution and uses a domestic intelligent processor based on the RISC-V heterogeneous architecture. The domestic intelligent processor adopts an asymmetric multi-process operation mode, where a single processor runs multiple operating systems simultaneously to process multiple tasks in parallel. This not only reduces the hardware cost but also ensures the processing of real-time sampled data and the processing of the tower barrel transformer fault identification algorithm. At the same time, the flexible, modular, and scalable characteristics of the RISC-V architecture instruction set make it very suitable for resource-constrained embedded computing environments.

[0016] The present invention directly processes and analyzes a large amount of sampled data on-site, including partial discharge, vibration, core current, bushing, oil temperature, temperature, and humidity, which is suitable for offshore platforms where space resources are extremely precious. On-site processing reduces the pressure on the cloud server side and decreases the data interaction latency with the cloud server, thereby improving the efficiency of the entire system.

[0017] The present invention uses LoRa self-organizing network technology to replace the traditional wired network connection method, reducing costs and maintenance difficulties. The wireless self-organizing network has the characteristics of strong mobility, effectively overcomes the environmental constraints of complex sea areas, has the ability of dynamic topology management, can adjust the network structure and routing in real time, and realizes effective communication and data transmission between the offshore wind turbine tower barrel transformer and the remote cloud server. The intelligent perception device of the offshore tower barrel transformer directly uploads the transformer status information to the cloud server, providing auxiliary decision-making for monitoring the operation status of the offshore wind turbine tower barrel transformer.

[0018] These advantages together enhance the practicality and effectiveness of the patented technology and effectively monitor the operation status of the offshore wind turbine tower barrel transformer. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic block diagram of the perception system described in Embodiment 1; Figure 2 It is a schematic block diagram of the intelligent processor; Figure 3 It is a LoRa self-organizing network architecture diagram; Figure 4 It is a LoRa self-organizing network flow chart; Figure 5 It is a flow chart of the perception method described in Embodiment 2. DETAILED DESCRIPTION OF THE INVENTION

[0020] The following further describes the present invention in conjunction with the drawings and specific embodiments.

[0021] Embodiment 1 This embodiment discloses an intelligent perception system for an offshore wind turbine tower transformer. By online monitoring the state indicators of partial discharge, core current, vibration velocity displacement, bushing voltage and current, oil temperature, inclination angle, ambient temperature and humidity of the tower transformer, it analyzes the deterioration and severity of the tower transformer, effectively judges the faults occurring in the tower transformer, and transmits the information to the cloud server through wireless ad-hoc network technology to timely process the faults and effectively avoid the occurrence of serious faults, ensuring the safe operation of the tower transformer.

[0022] As Figure 1 shown, this system includes a signal acquisition module, a data cache module, a signal analysis module, a fault identification module and a network communication module. The signal acquisition module includes a high-speed signal acquisition module and a low-speed signal acquisition module, which are composed of high-speed signal sensors and low-speed signal sensors. The high-speed signal acquisition module includes sensors for partial discharge, core current, bushing voltage and current, vibration acceleration, etc., and acquires high-speed signals such as partial discharge, core current, bushing voltage and current, vibration, etc. The low-speed signal acquisition module includes sensors for transformer oil temperature, transformer inclination angle, ambient temperature and humidity, etc., and is used to acquire low-speed signals such as transformer oil temperature, transformer inclination angle, ambient temperature and humidity, etc. Since the high-speed signal acquisition rate is fast and the data volume is large, there is a rate difference with the real-time nuclear acquisition speed, and a data cache module is required to implement data caching. The data cache module realizes the first-in-first-out data storage and processing mechanism for high-speed signals through a programmable gate array. The main processor real-time core serves as the signal analysis module. On the one hand, it analyzes the partial discharge data, core current, bushing voltage and current, vibration acceleration data, calculates the partial discharge spectrogram, core current effective value, bushing voltage and current, vibration velocity and displacement. On the other hand, it obtains the original data of oil temperature, inclination angle acceleration, ambient temperature and humidity through the low-speed signal acquisition module and calculates the oil temperature, inclination angle, ambient temperature and humidity.

[0023] The management core and NPU serve as the fault identification module. The fault identification module obtains the multi-source data calculated by the real-time core including the partial discharge spectrogram, core current, bushing voltage and current, vibration velocity displacement, inclination angle, oil temperature, ambient temperature and humidity through the shared memory queue. The partial discharge fault identification adopts an improved MobileNetV4 lightweight neural network algorithm. The lightweight neural network has a smaller volume, less computational complexity and higher accuracy, and is suitable for resource-constrained embedded computing environments. The fault identification module realizes the fault identification of the partial discharge spectrogram through the lightweight neural network algorithm. If there is a discharge phenomenon, it identifies the discharge type; if there is no discharge, it judges that there is no discharge. The fault identification module combines the partial discharge fault identification result with the multi-source data including core current, vibration velocity displacement, bushing voltage and current, inclination angle, oil temperature, ambient temperature and humidity to obtain a comprehensive judgment of the transformer state.

[0024] The network communication module is responsible for the self-organizing communication among the intelligent sensing devices of the offshore tower transformer and the communication with the cloud server.

[0025] In this embodiment, based on the performance, cost, power consumption, and self-control factors of the intelligent sensing device for the offshore wind turbine tower transformer, the signal analysis module and the fault identification module adopt domestic intelligent processors based on the RISC-V architecture. The RISC-V architecture processor uses the highly flexible RISC-V instruction set to match algorithms that require high computing power, and is very suitable for low-cost computing power requirements. As Figure 2 shown, the domestic intelligent processor is a multi-core processor, including multiple processor cores and an NPU. The NPU runs convolutional neural networks as a neural network processing unit. The multi-core processor is divided into relatively independent cores through an asymmetric multi-process operation mode, which improves the hardware real-time performance while ensuring the computing power of the processor, and reduces the number of processors, thereby reducing power consumption and cost. The asymmetric multi-process operation mode enables multiple cores of the main processor to run different tasks relatively independently. According to the functions, the multiple cores of the processor are divided into real-time cores and management cores. The real-time cores are responsible for undertaking the functions of the signal analysis module and performing real-time analysis on the data collected by the signal acquisition module. The management cores are responsible for managing the data processed by the real-time cores and external communication. The NPU runs convolutional neural networks to identify and analyze the data processed by the real-time cores. The real-time cores and the management cores run different operating systems and execute different tasks, and the real-time cores and the management cores communicate through a shared memory queue.

[0026] In this embodiment, the network communication module includes a 5G communication component, a LoRa gateway component, and a LoRa node component, and realizes the self-organizing communication among the intelligent sensing systems of the offshore tower transformer and the communication with the cloud server through the network communication module. As Figure 3As shown in the figure, the intelligent perception device of the offshore wind turbine tower transformer serves as a LoRa gateway, a LoRa relay, and a LoRa node according to its function based on the coverage range of the 5G base station. The intelligent perception system component of the tower transformer near the 5G base station serves as the LoRa gateway, the intelligent perception system of the tower transformer in the edge area serves as the LoRa terminal, and the intelligent perception system of the tower transformer in other areas serves as the LoRa relay. The LoRa gateway, LoRa relay, and LoRa node adopt a hybrid topology structure. The network communication module of the LoRa gateway enables the 5G communication component and the LoRa gateway component, and the network communication modules of the LoRa relay and LoRa node only enable the LoRa node component. The LoRa network is managed by means of multi-node linking and dynamic routing. According to the network link status, the functions of the LoRa relay and LoRa node can be switched with each other. The intelligent perception devices of the offshore wind turbine tower transformer are wirelessly bridged to each other through the LoRa network to form a LoRa dynamic self-organizing network. The LoRa terminal transmits the transformer status information to the LoRa gateway through the LoRa relay. The intelligent perception device of the tower transformer serving as the LoRa relay undertakes the routing and relay work. On the one hand, it forwards the transformer status information of the previous node, and on the other hand, it transmits its own transformer status information to the next LoRa relay. The LoRa gateway aggregates the information of each intelligent perception device of the tower transformer and transmits it to the cloud server through 5G.

[0027] In this embodiment, the intelligent perception systems of the offshore wind turbine tower transformers are wirelessly bridged to each other through LoRa to form a LoRa self-organizing network; the LoRa relay covers the area without signal and realizes signal coverage over a longer distance through the multi-node linking method.

[0028] Figure 4It is a self - networking flowchart of the intelligent perception device for the offshore wind turbine tower transformer. In the wireless self - networking of the offshore wind turbine tower transformer, each intelligent perception device has at least one communication component. The LoRa self - networking of the offshore wind turbine tower transformer adopts dynamic topology management capabilities, and its main functions include the transmission of transformer status information and the maintenance of dynamic routing links. In the LoRa self - networking, each intelligent perception device of the tower transformer has a fixed number. The LoRa gateway periodically sends broadcast messages to the LoRa relay and LoRa nodes. After receiving the broadcast message, the LoRa relay and LoRa nodes send reply frames to the LoRa gateway; the reply frame adopts a variable frame structure, and the frame carries the fixed number and signal strength information of the local node. After receiving the reply frame, the LoRa gateway updates the routing table of the LoRa node and sends an acknowledgment frame to the LoRa node according to this route. If the LoRa node cannot receive the LoRa gateway broadcast message or the message signal is too weak within the period, it is transmitted to the LoRa gateway through the relay method. The LoRa node first sends a broadcast message, and the nearby LoRa nodes or relays reply with the local node address, signal strength, and whether it is received in the gateway routing table. The LoRa node receiving the reply frame preferentially sends an acknowledgment message to the LoRa node or relay in the gateway routing table to determine the relay of the LoRa node. If the nodes sending the reply frame are not in the gateway routing table, a request message is sent to the node or relay with stronger signal according to the signal strength, asking whether it can be used as a relay. The node or relay receiving the request message continues to send a broadcast frame to the surrounding to determine whether it can be used as a relay, and the two already - connected nodes will not be re - connected. After determining the relay, the transformer status information is forwarded to the LoRa gateway through this relay, and the data frame carries the address and signal strength information of the relay node. If the LoRa relay cannot receive the LoRa gateway broadcast message or the message signal is too weak within the period, it switches to the role of a LoRa node and searches for the route to the LoRa gateway in the same way as the LoRa node. The LoRa gateway monitors the number of connected terminals in real - time, and through the message interaction between multiple gateways, the balanced load of each gateway is achieved. Encryption technology and authentication mechanisms are adopted between LoRa self - networking devices to ensure the security and reliability of data transmission.

[0029] In this embodiment, the process of the intelligent perception system for realizing the intelligent perception of the transformer device is as follows: Install partial discharge, vibration acceleration, iron core grounding, bushing voltage and current, inclination acceleration, oil temperature, temperature and humidity sensors on the offshore wind power tower transformer to collect diverse transformer data. The intelligent perception system of the offshore wind power tower transformer obtains the diverse transformer data transmitted by the sensors through the signal acquisition module, and calculates the partial discharge spectrogram, vibration velocity and displacement, iron core current, bushing current and voltage, oil temperature, ambient temperature and humidity through the signal analysis module. The signal analysis module inputs the calculated diverse data into the fault identification module for comprehensive analysis. The fault identification module performs partial discharge identification on the partial discharge spectrogram based on a lightweight neural network, and performs fault identification based on diverse data including the partial discharge identification result, iron core current, bushing voltage and current, vibration, inclination, oil temperature, ambient temperature and humidity. First, perform data preprocessing on the diverse data, make a preliminary judgment according to the initial values and warning values of each index specified by the standard specification, then calculate the relative deterioration degree for index normalization processing, and determine the fuzzy membership degree; construct a fuzzy judgment matrix according to the membership function, perform consistency verification on the judgment matrix, and perform probability calculation of the output result; calculate the weights of the variables by the entropy method, and perform fuzzy evaluation results according to the weight index to obtain the status score of the tower transformer.

[0030] Embodiment 2 This embodiment discloses an intelligent perception method for an offshore wind power tower transformer, as Figure 5 shown, including the following steps: S01. System deployment: Install sensors on the tower transformer to collect diverse transformer data, where the diverse transformer data includes partial discharge, vibration acceleration, iron core grounding, bushing voltage and current, inclination acceleration, oil temperature, temperature and humidity; deploy the above-mentioned transformer intelligent perception system on the tower transformer; S02. Implement monitoring and data collection: The signal acquisition module obtains the diverse transformer data collected by the sensors in real time, and transmits the collected diverse transformer data to the signal analysis module for processing and analysis in real time through the data cache module; S03. Transformer data processing and analysis: The signal analysis module processes the partial discharge data, calculates the partial discharge spectrogram according to the time-frequency analysis method, calculates the effective value of the iron core grounding current through the windowed short-time discrete Fourier transform for the iron core grounding current data, and calculates the amplitude, phase and frequency of the bushing voltage and current through the windowed short-time discrete Fourier transform for the bushing data; calculates the vibration velocity and vibration displacement through the vibration acceleration; obtains the low-speed signals including oil temperature, ambient temperature and humidity, inclination acceleration transmitted by the signal acquisition module in real time, and processes these raw data to calculate the oil temperature, ambient temperature and humidity, inclination; S04. Transformer fault identification. The fault identification module performs partial discharge identification on the partial discharge spectrogram based on a lightweight neural network. The fault identification module performs fault identification based on multi-source data including partial discharge identification results, core current, bushing voltage and current, vibration, inclination angle, oil temperature, ambient temperature and humidity. First, the multi-source data is preprocessed, and preliminary judgments are made according to the initial values and warning values of each index specified by the standard specifications. Then, the relative deterioration degree is calculated for index normalization to determine the fuzzy membership degree. A fuzzy judgment matrix is constructed according to the membership function, the consistency of the judgment matrix is verified, and the probability of the output result is calculated. The weight of the variable is calculated by the entropy method, and the fuzzy evaluation result is obtained according to the weight index to obtain the status score of the tower barrel transformer. S05. Transformer status upload. The status information of the tower barrel transformer is transmitted to the cloud server in real time through the network communication module. The network communication module uses the LoRa self-organizing network method. The transformer status information of each offshore wind power tower barrel transformer intelligent sensing device is aggregated to the LoRa gateway through LoRa relay, and the LoRa gateway sends the information to the cloud server through 5G.

[0031] The above description only presents the basic principles and preferred embodiments of the present invention. The improvements and replacements made by those skilled in the art based on the present invention fall within the protection scope of the present invention.

Claims

1. An intelligent sensing system for offshore wind power tower transformer, characterized in that: It includes a signal acquisition module, a signal analysis module, a fault identification module and a network communication module. The signal acquisition module is used to collect high-speed signals and low-speed signals representing the status of the tower transformer; the signal analysis module is connected to the signal acquisition module and is used to perform preliminary analysis and calculation on the data collected by the signal acquisition module; the fault identification module is connected to the signal analysis module and is used to receive data from the signal analysis module to perform partial discharge fault identification and comprehensive judgment of the transformer status; The signal analysis module and fault identification module use a multi-core processor based on the RISC-V architecture, including multiple processor cores and NPU. The NPU, as a neural network processing unit, separates the multi-core processor into relatively independent cores through an asymmetric multi-process operation mode. One independent core serves as a real-time core, and the other cores serve as management cores. The real-time core is responsible for assuming the functions of the signal analysis module and performing real-time analysis on the data collected by the signal acquisition module. The management core is responsible for managing the data processed by the real-time core and communicating with the outside world. The NPU runs a convolutional neural network to identify and analyze the data processed by the real-time core. The network communication module includes 5G communication components, LoRa gateway components and LoRa node components. The network communication module is used to realize self-organizing network communication between the intelligent sensing systems of offshore tower transformers and communication with cloud servers.

2. The offshore wind power tower transformer intelligent sensing system according to claim 1 is characterized in that: The real-time core and the management core independently run different operating systems, and the real-time core and the management core communicate through a shared memory queue.

3. The offshore wind power tower transformer intelligent sensing system according to claim 1 is characterized in that: The transformer intelligent sensing systems are wirelessly bridged to each other through LoRa to form a LoRa self-organizing network; according to the coverage range of the 5G base station, they serve as LoRa gateways, LoRa relays and LoRa nodes respectively according to their functions. The network communication module of the LoRa gateway enables 5G communication components and LoRa gateway components, and the network communication modules of the LoRa relay and LoRa node only enable LoRa node components; during communication, the LoRa terminal transmits the transformer status information to the LoRa gateway through the LoRa relay. The tower transformer intelligent sensing device acting as a LoRa relay undertakes routing and relaying work. On the one hand, it forwards the transformer status information of the previous node, and on the other hand, it transmits its own transformer status information to the next LoRa relay. The LoRa gateway aggregates the information of each tower transformer intelligent sensing device and transmits it to the cloud server via 5G.

4. The offshore wind power tower transformer intelligent sensing system according to claim 3 is characterized in that: The LoRa self-organizing network adopts a dynamic topology management method. The roles of LoRa relays and LoRa nodes can be switched according to the network link status. The LoRa gateway periodically sends broadcast messages to the LoRa relays and LoRa nodes, monitors the reply frames of the LoRa nodes and LoRa relays and updates the routing table of the LoRa node. The messages transmitted by the LoRa nodes and LoRa relays to the LoRa gateway adopt a variable frame structure. The messages include transformer status information and all LoRa relay and node information in the route. The LoRa gateway monitors the number of connected terminals in real time and realizes balanced load of each gateway through message interaction between multiple gateways.

5. The offshore wind power tower transformer intelligent sensing system according to claim 1 is characterized in that: The signal acquisition module includes a high-speed signal acquisition module and a low-speed signal acquisition module. The high-speed signal acquisition module is used to collect high-speed signals including transformer partial discharge, transformer core current, transformer bushing voltage and current, and transformer vibration acceleration. The low-speed signal acquisition module is used to collect low-speed signals including transformer oil temperature, ambient temperature and humidity, and transformer inclination acceleration.

6. The offshore wind power tower transformer intelligent sensing system according to claim 5 is characterized in that: It also includes a data cache module. The high-speed signal acquisition module is connected to the signal analysis module through the data cache module. The data cache module implements a first-in-first-out data storage and processing mechanism for high-speed signals through a programmable gate array.

7. The offshore wind power tower transformer intelligent sensing system according to claim 1 is characterized by: The signal analysis module performs signal analysis in the following way: processes the partial discharge data, calculates the partial discharge spectrum according to the time-frequency analysis method, calculates the effective value of the core grounding current data through windowed short-time discrete Fourier transform, and calculates the bushing voltage and current amplitude, phase, and frequency through windowed short-time discrete Fourier transform; calculates the vibration velocity and vibration displacement through vibration acceleration; obtains the low-speed signals including oil temperature, ambient temperature and humidity, and inclination acceleration transmitted by the signal acquisition module in real time, processes these raw data, and calculates the oil temperature, ambient temperature and humidity, and inclination.

8. The offshore wind power tower transformer intelligent sensing system according to claim 6 is characterized by: The fault identification module performs fault identification in the following way: the fault identification module obtains in real time the multivariate data including partial discharge spectrum, core current, bushing voltage and current, vibration velocity displacement, inclination, oil temperature, and ambient temperature and humidity calculated by the signal analysis module, and performs partial discharge identification on the partial discharge spectrum through the improved MobileNetV4 lightweight neural network algorithm. Then, the multivariate data including partial discharge identification results, core current, bushing voltage and current, vibration velocity displacement, inclination, oil temperature, and ambient temperature and humidity are preprocessed, the indicators are normalized, the fuzzy membership is determined, and then the fuzzy judgment matrix is ​​constructed, the judgment matrix is ​​checked for consistency, and finally the indicator weight is calculated to obtain the operating status of the tower transformer.

9. The offshore wind power tower transformer intelligent sensing system according to claim 3 is characterized by: Encryption technology and authentication mechanism are used between LoRa self-organizing network devices.

10. An intelligent sensing method for offshore wind power tower transformer, characterized in that: The following steps are involved: S01. System deployment: installing sensors on the tower transformer to collect transformer multivariate data, including partial discharge, vibration acceleration, core grounding, bushing voltage and current, inclination acceleration, oil temperature, temperature and humidity; deploying the transformer intelligent sensing system as described in any one of claims 1 to 9 on the tower transformer; S02, implement monitoring and data collection, the signal acquisition module acquires the transformer multivariate data collected by the sensor in real time, and transmits the collected transformer multivariate data to the signal analysis module in real time through the data cache module for processing and analysis; S03, transformer data processing and analysis, the signal analysis module processes the partial discharge data, calculates the partial discharge spectrum according to the time-frequency analysis method, calculates the effective value of the core grounding current data through windowed short-time discrete Fourier transform, and calculates the bushing voltage and current amplitude, phase, and frequency through windowed short-time discrete Fourier transform; calculates the vibration velocity and vibration displacement through vibration acceleration; obtains the low-speed signals including oil temperature, ambient temperature and humidity, and inclination acceleration transmitted by the signal acquisition module in real time, and processes these raw data to calculate the oil temperature, ambient temperature and humidity, and inclination; S04, transformer fault identification, the fault identification module performs partial discharge identification on the partial discharge spectrum based on a lightweight neural network, and the fault identification module performs fault identification based on multivariate data, which includes partial discharge identification results, core current, bushing voltage and current, vibration, inclination, oil temperature, ambient temperature and humidity. First, the multivariate data is preprocessed, and a preliminary judgment is made based on the initial values ​​and warning values ​​of each indicator specified in the standard specification, and then the relative degradation degree is calculated to perform indicator normalization processing and determine the fuzzy membership degree; Construct a fuzzy judgment matrix based on the membership function, perform consistency check on the judgment matrix, and calculate the probability of the output result; The weight of the variable is calculated by entropy method, and the fuzzy evaluation result is carried out according to the weight index to obtain the tower transformer status score; S05. Upload the transformer status. The tower transformer status information is transmitted to the cloud server in real time through the network communication module. The network communication module adopts the LoRa self-organizing network mode. The transformer status information of each offshore wind power tower transformer intelligent sensing device is aggregated to the LoRa gateway through the LoRa relay. The LoRa gateway sends the information to the cloud server through 5G.

Images