Wind driven generator fault diagnosis method based on multi-scale space-time convolution deep belief network

Abstract

The invention provides a wind driven generator fault diagnosis method based on a multi-scale space-time convolution deep belief network. According to the method, inherent spatial-temporal correlationand interaction characteristics of an SCADA multivariable time sequence are utilized; a convolution depth belief network with different convolution kernel structures is designed to capture space-timecorrelation information among sensor variables in a cascaded manner; meanwhile, interactive complementary characteristics among variables are mined under a plurality of filter scales in a parallel manner; according to the technical means, a space-time dependence extraction method and a multi-scale feature learning method are fused, so that richer fault diagnosis information can be extracted, and compared with a traditional convolutional deep belief network model and variants thereof, the classification performance of the invention can be enhanced, and a new way is provided for the field of wind driven generator fault diagnosis.

Description

technical field [0001] The invention belongs to the technical field of fault diagnosis of wind power generators, and in particular relates to a fault diagnosis method of wind power generators based on a multi-scale spatio-temporal convolution deep belief network. Background technique [0002] In recent years, wind energy, as an inexhaustible and rapidly developing clean and renewable energy, has attracted widespread attention from all over the world. Due to its important role in wind power generation, wind turbines are widely used on land and at sea. However, in practical applications, wind turbines usually operate in a harsh working environment around the clock, and are affected by various complex effects for a long time, which can easily cause failures, and even cause the unit to shut down in severe cases. These failures and unplanned downtime have seriously affected the economic benefits of wind farms and the healthy development of the wind power industry. Therefore, it ...

AI Technical Summary

Problems solved by technology

[0004] At present, there are methods for wind turbine fault diagnosis by processing SCADA data of wind turbines, but SCADA data is essentially a multivariate time series with typical spatio-temporal correlation and interaction characteristics, while the existing wind turbine fault diagnosis methods are ubiquitous. Lack of ability to capture these features, which in turn affects fault classification performance

Images