CN-121676236-B - Cloud-edge coordinated wind turbine heat management self-supervision prediction control system and method

CN121676236BCN 121676236 BCN121676236 BCN 121676236BCN-121676236-B

Abstract

The invention discloses a cloud-edge coordinated wind turbine heat management self-supervision predictive control system and a cloud-edge coordinated wind turbine heat management self-supervision predictive control method, wherein the system comprises a data acquisition module, a data preprocessing module, a self-supervision detection module and an execution module which are positioned at an edge layer, and the modules are connected through communication; the cloud side cooperative interaction module is positioned on the transmission layer and is in communication connection with the execution module of the edge layer, the thermal fault diagnosis unit and the global thermal management optimization unit are positioned on the cloud layer and are in communication connection, and the global thermal management optimization unit is in communication connection with the thermal fault diagnosis unit. By adopting the cloud-edge coordinated wind turbine heat management self-supervision predictive control system and the cloud-edge coordinated wind turbine heat management self-supervision predictive control method, the shutdown faults caused by heat management failure are reduced, and the power generation loss and the operation and maintenance cost are effectively reduced.

Inventors

Diao Qianwen
LOU CHAOYAN
ZHANG PENGYUE

Assignees

中国计量大学

Dates

Publication Date: 20260505
Application Date: 20260211

Claims (8)

1. The cloud-edge coordinated wind turbine heat management self-supervision predictive control system is characterized by comprising an edge layer, a transmission layer and a cloud layer; The edge layer comprises a data acquisition module, a data preprocessing module, a self-supervision detection module and an execution module, wherein the data acquisition module, the data preprocessing module, the self-supervision detection module and the execution module are connected through communication; The execution module is connected with the self-supervision detection module and used for outputting a thermal state grade based on the connection of the self-supervision detection module and executing differential regulation; The normal working condition is that the heat is dissipated in a self-adaptive way by adjusting the rotating speed of the fan, and the data of the normal working condition is stored locally; the slight abnormality is that the semiconductor refrigerating power is self-adaptively adjusted by dynamically distributing the flow of the micro-channel cooling liquid, and an abnormal data message is generated and uploaded to a transmission layer; Emergency control, locking the position of a fault component, triggering an alarm, generating an abnormal data message, and uploading the abnormal data message to a transmission layer; the transmission layer comprises a cloud edge cooperative interaction module, wherein the cloud edge cooperative interaction module is in communication connection with an execution module of the edge layer; The cloud layer comprises a thermal fault diagnosis unit and a global thermal management optimization unit, wherein the thermal fault diagnosis unit is in communication connection with the cloud side cooperative interaction module; The thermal fault diagnosis unit in the cloud layer receives the abnormal data message uploaded by the transmission layer, and completes fault tracing and residual service life prediction by combining with the historical fault case library, and outputs maintenance suggestions; and the global thermal management optimizing unit is used for constructing an optimizing model by summarizing the thermal state grade and the working condition data of the wind turbine generator, generating optimizing parameters and transmitting the optimizing parameters to the execution module and the self-supervision detection module.
2. The cloud-edge coordinated wind turbine heat management self-supervision predictive control system of claim 1, wherein the data acquisition module is used for acquiring wind turbine multi-dimensional data in real time, wherein the wind turbine multi-dimensional data comprises wind turbine operation data and environmental parameters; (1) The wind turbine generator operation data comprise thermal state data and working condition data; The thermal state data comprises real-time temperature data of a gear box, a generator stator, a generator rotor and a converter IGBT module; the working condition data comprise wind speed, rotating speed, torque and power of a power grid; (2) Environmental parameters include ambient temperature, humidity and precipitation.
3. The cloud-edge collaborative wind turbine heat management self-supervision predictive control system according to claim 2 is characterized in that the data preprocessing module is used for preprocessing the wind turbine multi-dimensional data acquired by the data acquisition module to obtain standardized multi-dimensional data; the pretreatment comprises the following specific contents: firstly, carrying out high-frequency denoising treatment on multidimensional data of a wind turbine generator by a wavelet transformation method; Secondly, adopting a mean value filling method to fill up missing data of the multi-dimensional data of the wind turbine after denoising; Finally, mapping the multi-dimensional data of the wind turbine generator after the completion of the deletion to the maximum and minimum normalization method Interval.
4. The cloud-edge collaborative wind turbine heat management self-supervision predictive control system according to claim 3, wherein the self-supervision detection module comprises a self-supervision pre-training unit, a feature extraction unit and an anomaly scoring unit; The self-supervision pre-training unit is used for constructing and training a CNN-transducer model based on the pre-training data set to generate a normal working condition feature library; the feature extraction unit is used for extracting the real-time feature vector of the preprocessed standardized multidimensional data based on the trained CNN-transducer model; and the anomaly scoring unit is used for comparing the real-time feature vector with the feature vector of the normal working condition feature library to obtain feature similarity, wherein the feature similarity is in negative correlation with the anomaly degree of the real-time data and is used for representing the anomaly degree of the real-time data.
5. The cloud-edge collaborative wind turbine heat management self-supervision predictive control system according to claim 4, wherein the CNN-converter model comprises a CNN subunit for extracting local dimension coupling characteristics of the standardized multidimensional data; The transducer subunit performs global time sequence feature extraction on the local dimension coupling features output by the CNN subunit through a self-attention mechanism; And the feature fusion subunit is used for carrying out feature fusion on the local dimension coupling features of the CNN subunit and the global time sequence features extracted by the transducer subunit to generate a comprehensive feature vector.
6. The cloud-edge collaborative wind turbine heat management self-supervision predictive control system according to claim 5, wherein the self-supervision pre-training is performed on a CNN-converter model based on a pre-training data set, and a comprehensive feature vector generated by the self-supervision pre-training is constructed as a normal working condition feature library.
7. The cloud-edge collaborative wind turbine heat management self-supervision predictive control system according to claim 6, wherein the anomaly scoring unit classifies heat state levels based on feature similarity according to the following rules: Setting a first threshold A and a second threshold B by combining an operation scene of the wind turbine and engineering practice experience, wherein the first threshold A is the lowest matching degree threshold of a real-time characteristic vector and a normal working condition characteristic vector, the second threshold B is a critical matching degree threshold of the real-time characteristic vector and the normal working condition characteristic vector based on characteristic vector setting corresponding to thermal safety critical conditions of a gear box, a generator stator, a generator rotor and a converter, and the first threshold A is a critical matching degree threshold of the real-time characteristic vector and the normal working condition characteristic vector A second threshold B; (1) Feature similarity The first threshold A is judged to be in a normal working condition; (2) Second threshold B Feature similarity A first threshold A, which is determined to be slightly abnormal; (3) Feature similarity And the second threshold B is used for judging that the alarm is in serious abnormality and triggering early warning.
8. The cloud-edge coordinated wind turbine heat management self-supervision predictive control method is applied to the cloud-edge coordinated wind turbine heat management self-supervision predictive control system according to any one of claims 1 to 7, and is characterized by comprising the following steps: s1, acquiring multidimensional data of a wind turbine generator in real time through a data acquisition module and inputting the multidimensional data into a data preprocessing module; s2, preprocessing the collected multidimensional data of the wind turbine generator by using a data preprocessing module to obtain standardized multidimensional data; screening out standardized multidimensional data under normal working conditions based on the history operation record without abnormality to form a pre-training data set; S3, constructing and training a CNN-transducer model through a self-supervision pre-training unit of a self-supervision detection module based on the pre-training data set obtained in the S2, and generating a normal working condition feature library; S4, calling a trained CNN-converter model through a feature extraction unit of the self-supervision detection module, extracting real-time feature vectors of the standardized multidimensional data preprocessed in the S2, calculating feature similarity between the real-time feature vectors and feature vectors of a normal working condition feature library obtained in the S3 through an anomaly scoring unit of the self-supervision detection module, and classifying the thermal state grade; s5, based on the thermal state level divided in the S4, performing differential regulation and control through an execution module, and simultaneously generating an abnormal data message which is uploaded to a cloud layer through a transmission layer; and S6, after the thermal fault diagnosis unit of the cloud layer receives the abnormal data message, fault tracing and residual service life prediction are completed, the global thermal management optimization unit of the cloud layer generates optimization parameters, and the optimization parameters are transmitted to the execution module and the self-supervision detection module through the transmission layer.

Description

Cloud-edge coordinated wind turbine heat management self-supervision prediction control system and method Technical Field The invention relates to the technical field of wind power generation, in particular to a cloud-edge coordinated wind turbine heat management self-supervision predictive control system and method. Background Wind energy is a clean and pollution-free renewable energy source, the amount of the wind energy is huge, and wind power generation becomes one of the core development directions of the global energy field and is highly valued in all countries of the world. The wind generating set (wind generating set for short) is a system for converting kinetic energy of wind into electric energy, and the safe and stable operation of the wind generating set directly determines the supply quality of wind power energy. Along with the increasingly severe operation conditions of the wind turbine, how to ensure the safe and stable operation of the wind turbine and how to improve the operation reliability of the wind turbine have become hot problems in the wind power research field. The core components of the wind turbine generator are mainly a gear box, a generator and a converter, but the wind turbine generator is in a complex working condition for a long time, and is very easy to cause problems of scale deposition, fan performance attenuation and the like of a heat dissipation system due to the influence of environmental factors (such as wind sand or humidity), so that the heat management capability is directly reduced. The shutdown failure caused by thermal management failure has increasingly high proportion, not only causes power generation loss, but also increases operation and maintenance cost and equipment loss. Therefore, a cloud-edge coordinated wind turbine heat management self-supervision predictive control system and a cloud-edge coordinated wind turbine heat management self-supervision predictive control method are needed. Disclosure of Invention The invention aims to provide a cloud-edge collaborative wind turbine heat management self-supervision predictive control system and a cloud-edge collaborative wind turbine heat management self-supervision predictive control method, which not only solve the delay problem of the traditional system and reduce the shutdown fault caused by heat management failure, but also avoid the local optimal trap of a single edge system, can be suitable for wind turbines in different scenes such as land or sea, and effectively reduce the power generation loss and the operation and maintenance cost. In order to achieve the purpose, the invention provides a cloud-edge coordinated wind turbine heat management self-supervision predictive control system, which comprises an edge layer, a transmission layer and a cloud layer; The edge layer comprises a data acquisition module, a data preprocessing module, a self-supervision detection module and an execution module, wherein the data acquisition module, the data preprocessing module, the self-supervision detection module and the execution module are connected through communication; the transmission layer comprises a cloud edge cooperative interaction module, wherein the cloud edge cooperative interaction module is in communication connection with an execution module of the edge layer; the cloud layer comprises a thermal fault diagnosis unit and a global thermal management optimization unit, wherein the thermal fault diagnosis unit is in communication connection with the cloud side cooperative interaction module, and the global thermal management optimization unit is in communication connection with the thermal fault diagnosis unit. Preferably, the data acquisition module is used for acquiring multidimensional data of the wind turbine generator in real time, wherein the multidimensional data of the wind turbine generator comprises operation data and environmental parameters of the wind turbine generator; (1) The wind turbine generator operation data comprise thermal state data and working condition data; The thermal state data comprises a gear box, a generator stator/rotor temperature and a converter IGBT module temperature; the working condition data comprise wind speed, rotating speed, torque and power of a power grid; (2) Environmental parameters include ambient temperature, humidity and precipitation. Preferably, the data preprocessing module is used for preprocessing the multidimensional data of the wind turbine generator set acquired by the data acquisition module to obtain standardized multidimensional data, screening the standardized multidimensional data under normal working conditions based on the abnormal history-free operation record, and forming a pre-training data set; the pretreatment comprises the following specific contents: firstly, carrying out high-frequency denoising treatment on multidimensional data of a wind turbine generator by a wavelet transformation method; Secondly, adopting a mean value filling