Research On Early Fault Warning Technology Of Wind Energy Information Network

The development of new energy and protecting earth's environment is a global consensus. Wind energy is generally accepted around the world as a clean new energy, and developing the wind power technology contributes to reducing greenhouse gas emissions and the sustainable development of human beings. However, working in the bad condition leads to a big fault probability of wind turbine. If faults can be discovered as early as possible and repaired timely, economic loss will be reduced greatly. Wind energy information network based on the energy internet provides a direction for the accurate early fault warning of wind turbine.The gearbox of wind turbine is the key component in the power transmission system, but fails frequently. It restricts the working state of wind turbine. Based on data mining technology, this thesis takes the gearbox as the research object, uses the historical operation statement data which from the supervisory control and data acquisition (SCADA)system of wind farm, and the optimization mining algorithm to design and build wind farm management system that provides reliable and intuitive wind turbine monitoring platform. The main work and innovations of this thesis include:1. As to the modeling method of early fault warning, this thesis which takes gearbox bearing temperature as the prediction parameter, builds model with the historical data from wind turbine in normal operating condition. Then, real-time state data is put into this model and predicted bearing temperature can be got. Early fault warning can be achieved by monitoring the variation of difference between the actual gearbox bearing temperature and predicted temperature.2. For the selection of data mining algorithm in the early warning model of this thesis, three algorithms which are linear regression, support vector machine and back propagation (BP) neural network are compared by simulation and the conclusion is prediction accuracy and generalization of BP neural network are both better than other algorithms. Finally, the early fault warning model based on this algorithm can advance the gearbox failure warning by 12 days.3. This thesis designs and builds the wind farm management system.This system is divided into offline research platform for data mining research, and online monitoring platform for monitoring all the wind turbines, so that the fault wind turbine and alarm information can be visually displayed.This thesis studies early fault warning technology of wind turbine in wind energy information network, and uses the optimization mining algorithm to develop the wind farm management system which provides technical support for early fault warning.