Fault self-diagnosis and secondary protection system for wind power converter

Abstract

The invention discloses a fault self-diagnosis and secondary protection system for a wind power converter, which relates to the technical field of wind power generation and comprises a multi-sensor data acquisition module used for acquiring electric, thermal and mechanical parameters of the converter in real time; the data processing and feature extraction module is used for preprocessing the original data and extracting multi-dimensional features; the intelligent fault diagnosis module identifies a fault type and predicts fault time based on a deep learning model; the secondary protection execution module is used for triggering redundancy control or standby switching according to the fault level; and the remote monitoring and self-learning module realizes system state monitoring and model optimization. Through multi-sensor data fusion, deep learning diagnosis and a standby switching mechanism, real-time monitoring, accurate classification and seamless switching of converter faults are realized. The system comprises a multi-sensor data acquisition module, a data processing module, an intelligent diagnosis module, a secondary protection execution module and a remote monitoring module, and the reliability and the generating capacity of the wind power plant can be remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power generation, specifically to a wind power converter fault self-diagnosis and secondary protection system for real-time monitoring of converter operation status, automatic diagnosis of fault types and triggering of backup switching to ensure the continuous and stable operation of wind farms. BACKGROUND

[0002] As the core component of wind turbine generators, the reliability of wind power converters directly affects the power generation efficiency and grid stability of wind farms. Existing converter fault diagnosis techniques rely on single sensors or simple threshold detection, resulting in low diagnostic accuracy and high false alarm rates. For example, traditional systems only trigger shutdown through temperature or current abnormalities, lacking precise identification of fault types. In addition, existing protection measures are usually single protection (such as direct shutdown), leading to long-term power outage of wind farms and resulting in loss of power generation.

[0003] Although multi-sensor data fusion techniques (such as the method proposed by Hang Jun et al.) and machine learning models (such as the deep learning anomaly identification model of Huaneng Group) have been applied to fault diagnosis, they have not been combined with backup switching mechanisms. The redundancy control patent of Hua Wang Electrical only targets control module faults and does not achieve redundancy switching of the entire converter. Although the modular converter switching technology optimizes the number of sub-modules, it does not address the backup problem after a fault.

[0004] Therefore, there is an urgent need for a system that integrates real-time diagnosis, accurate classification, and secondary protection. SUMMARY

[0005] The present application aims to provide a wind power converter fault self-diagnosis and secondary protection system that achieves rapid positioning of converter faults and continuous power supply through multi-source data fusion, intelligent diagnosis algorithms, and backup switching mechanisms, thereby improving the reliability of wind farms.

[0006] A wind power converter fault self-diagnosis and secondary protection system includes: A multi-sensor data acquisition module integrates current sensors, voltage sensors, temperature sensors, and vibration sensors to real-time collect electrical parameters (such as grid-side voltage and machine-side current), thermal parameters (such as IGBT module temperature), and mechanical parameters (such as radiator vibration) of the converter. The sensors are deployed at key locations of the converter (such as the DC bus and power modules) to ensure comprehensive coverage of the operating status.

[0007] A data processing and feature extraction module pre-processes the raw data (including data synchronization and denoising filtering) and extracts time domain, frequency domain, and entropy domain features. For example, the five-point cubic smoothing algorithm is used to remove noise, the fast Fourier transform (FFT) is used to analyze current harmonic components, and the wavelet packet decomposition is used to extract energy features of vibration signals.

[0008] Intelligent fault diagnosis module: a multi-scale deep feature fusion model based on ResNet network is adopted, combined with long short-term memory network (LSTM) for time series analysis. The model is trained by historical data, which can identify 12 types of faults such as overcurrent, overvoltage, overheating and IGBT open circuit, and predict the fault occurrence time. The diagnosis accuracy can reach more than 95%, and the false positive rate is less than 3%.

[0009] Secondary protection execution module: Fault classification response: Primary response: for transient disturbance (such as voltage fluctuation), adjust the PI controller parameters to realize adaptive optimization.

[0010] Secondary response: for recoverable faults (such as short-term overload), trigger the redundant control module to switch to the backup control channel.

[0011] Tertiary response: for non-recoverable faults (such as power module damage), start the backup converter switching program.

[0012] Backup switching mechanism: modular design, main converter and backup converter are connected through intelligent contactors. After the diagnosis module issues the switching command, the faulty converter is first disconnected from the power grid, and the load is seamlessly switched to the backup converter, with a switching time of less than 20ms, ensuring power supply continuity. During the switching process, a pre-charging circuit is used to avoid impact current.

[0013] Remote monitoring and self-learning module: based on edge computing technology, real-time upload of diagnosis results to wind farm SCADA system, and data analysis through cloud platform. The system can automatically update the fault feature library and optimize the diagnosis model parameters to adapt to different operating environments.

[0014] The present application provides a wind power converter fault self-diagnosis and secondary protection system, which has the following beneficial effects: 1. High diagnosis accuracy: multi-sensor data fusion combined with deep learning significantly improves fault recognition accuracy, more than 30% higher than traditional methods.

[0015] 2. Fast response: backup switching time is less than 20ms, reducing downtime by more than 90%, and annual power generation can be increased by 5%-8%.

[0016] 3. High reliability: redundant design and hierarchical protection mechanism enable the system to operate stably under extreme conditions, reducing the failure rate by 40%.

[0017] 4. Intelligent operation and maintenance: remote monitoring and self-learning functions enable predictive maintenance, reducing operation and maintenance costs by more than 20%. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments will be briefly introduced below.

[0019] The drawings described in the following description are merely related to some embodiments of the present application, but not to the limitation of the present application.

[0020] In the drawings: Figure 1 It is a schematic diagram of the overall architecture of the system. Figure 2 It is a deployment flowchart of the multi-sensor data acquisition module. Figure 3 It is a work flowchart of the intelligent fault diagnosis module. Figure 4 It is a logic flowchart of the secondary protection execution module. DETAILED DESCRIPTION

[0021] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, but not all the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0022] The present application proposes a wind power converter fault self-diagnosis and secondary protection system, comprising: A multi-sensor data acquisition module for real-time acquisition of electrical, thermal and mechanical parameters of the converter. A data processing and feature extraction module for preprocessing of the original data and extraction of multi-dimensional features. An intelligent fault diagnosis module for identifying fault types and predicting fault time based on a deep learning model. A secondary protection execution module for triggering redundant control or backup switching according to fault level. A remote monitoring and self-learning module for system state monitoring and model optimization.

[0023] The multi-sensor data acquisition module includes current sensors, voltage sensors, temperature sensors and vibration sensors, and is deployed in the converter.

[0024] The intelligent fault diagnosis module adopts a ResNet-LSTM fusion model, and the diagnosis accuracy is not less than 95%.

[0025] The backup switching time of the secondary protection execution module is less than 20ms, and the switching process avoids impact current through a pre-charge circuit.

[0026] The remote monitoring and self-learning module is based on edge computing and a cloud platform to realize automatic updating of a fault feature library. It also includes hardware deployment: installing current, voltage, temperature and vibration sensors at positions such as converter power modules, DC bus and heat sinks, transmitting data to the data processing module through CAN bus, and combining electrical, thermal and mechanical parameters for comprehensive diagnosis to solve the limitations of single sensors.

[0027] Data processing: After the original data is synchronized and calibrated, multi-dimensional feature vectors are generated through feature extraction algorithms, and input into the ResNet-LSTM model for training and reasoning. The ResNet-LSTM model realizes fault type identification and time prediction, with a diagnosis accuracy rate of 95%.

[0028] Fault diagnosis: The model outputs fault type and prediction time. If the fault level is three, the standby switching instruction is triggered. During the switching process, the intelligent contactor first disconnects the faulty converter, then closes the standby converter, and simultaneously starts the pre-charge circuit. The modular design and intelligent contactor realize seamless switching within 20ms, ensuring power supply continuity.

[0029] System optimization: The cloud platform regularly updates the diagnostic model, and the edge node adjusts the control parameters in real time to ensure continuous optimization of system performance. The combination of cloud self-learning and edge computing continuously improves system performance.

Claims

1. A self-diagnosis and secondary protection system for wind power converter faults, characterized in that, include: The multi-sensor data acquisition module also includes integrated current sensors, voltage sensors, temperature sensors, and vibration sensors, used to acquire the electrical, thermal, and mechanical parameters of the converter in real time; The data processing and feature extraction module preprocesses the raw data and extracts multidimensional features; The intelligent fault diagnosis module identifies fault types and predicts fault times based on a deep learning model. The secondary protection execution module triggers redundant control or backup switching based on the fault level. The remote monitoring and self-learning module enables system status monitoring and model optimization.

2. The system according to claim 1, characterized in that, The multi-sensor data acquisition module includes a current sensor, a voltage sensor, a temperature sensor, and a vibration sensor, and is deployed on the converter.

3. The system according to claim 1, characterized in that, The intelligent fault diagnosis module adopts a ResNet-LSTM fusion model, with a diagnostic accuracy of no less than 95%.

4. The system according to claim 1, characterized in that, The backup switching time of the secondary protection execution module is less than 20ms, and the switching process avoids inrush current through a pre-charging circuit.

5. The system according to claim 1, characterized in that, The remote monitoring and self-learning module is based on edge computing and cloud platform to realize automatic updates of the fault feature database.

Images