Fan blade icing visual state monitoring based on FDA and SOM

Abstract

The invention discloses a fan blade icing visual state monitoring method based on FDA and SOM. The method adopts a mode based on data driving to visually show the state of a fan blade and the state transition process of the fan blade through a visual means during the operation of a fan. The method specifically comprises the following steps: acquiring data of a single fan unit in a certain wind field; cleaning, preprocessing and standardizing the obtained data; carrying out feature extraction by utilizing linear discriminant analysis (FDA), and building an FDA model; and inputting the data processed by the FDA model into a self-organizing mapping neural network (SOM), marking and re-standardizing sample categories, then learning training data to obtain an SOM model, and inputting test datato complete visualization and state monitoring.

Description

technical field [0001] The invention relates to a visualization and state monitoring method for the icing of fan blades, and mainly relates to the visual monitoring of the normal and icing states of the fan blades. Background technique [0002] Today, when the current environmental pollution situation is unprecedentedly severe and the stock of fossil fuels is facing depletion, wind energy, as a clean and renewable energy source, is one of the most ideal new energy sources in the world. The use of wind power generation can effectively reduce greenhouse gas emissions and have a positive impact on the current global climate change. Therefore, it has attracted more and more attention from many countries in the world. With the new energy support policies promulgated by governments of various countries, wind power generation The industry has ushered in the heyday of great development. By comparing the data of new wind turbine assembly machines in various countries in recent years,...

AI Technical Summary

Problems solved by technology

[0003] Although wind power generation has many advantages, in this field, the problem of wind turbine blade icing has not been well resolved, and this has become a long-term problem that plagues the wind power industry

A large amount of operating experience shows that when the blades of the wind turbine freeze, the aerodynamic shape of the blades will be significantly changed, and the load on each blade of the fan will be increased, which will intensify the wear of the fan during long-term operation and affect the stable operation of the fan; When the ice falls off, the falling ice blocks will damage the power plant equipment and even threaten the personal safety of the power plant maintenance personnel.

[0004] Domestic research and achievements on wind turbine blade icing faults are in the initial development stage. At present, the monitoring of wind turbine blade icing mainly uses video visual monitoring or the method of installing sensors to monitor the status of the blades, but these The method will lead to high economic costs and is greatly affected by severe weather conditions. With the continuous use of the equipment, the measurement error of the related equipment will gradually increase, thereby affecting the monitoring effect; in addition, there are some data-driven methods, mainly It uses multivariate statistical analysis and artificial neural network to identify the state of the blade through data mining, and the effect is better, but there are still problems that the model has too high requirements for feature selection and the accuracy of the algorithm is unstable.

Images