Prediction Of Remaining Time Of Wind Power Process Based On Deep Meta-Learning

Wind energy has gained widespread attention as a new type of energy,which promotes the rapid development of wind power generation technology,and the resulting maintenance problems of fan equipment restrict the production efficiency of wind power generation.Wind power maintenance work plays a key role in the efficient output of wind energy.It involves various operations such as filling in work tickets,personnel changes,and process delays,which makes the wind power operation and maintenance process increasingly complex,reduces the efficiency of personnel maintenance and the efficiency of wind power equipment.Wind power generation equipment cannot be maintained in a timely manner.At present,the complexity of wind power business process is high,and there are many business process change information.Due to the long process cycle,the number of operation and maintenance event logs is limited,which makes the existing remaining time prediction model prone to model overfitting and weak generalization ability,etc.question.Therefore,there is room for improvement in the prediction effect of the remaining time of the business process.Aiming at the shortage of existing remaining time prediction models and the limited number of event log samples,a remaining time prediction method is improved.Then,a recommendation method for the remaining time prediction model is proposed,which is suitable for specific event logs in different scenarios.The main contents are as follows:(1)Aiming at the problem that the wind power business process is complex,the process is changed repeatedly,and the resulting long process trajectory results in a small number of existing event logs,a remaining time prediction model based on meta-learning and temporal convolutional network is proposed.First,process trajectory prefix data of different lengths are obtained,and a prediction model is established using a temporal convolutional network.Then,considering the small number of event log samples,meta-learning technology is added to provide better initial hyperparameters for the training task of small sample data,and improve the prediction accuracy of the model proposed.(2)Aiming at the problem that the selection of the remaining time prediction model in a specific business scenario is time-consuming and easily affected by human subjective decision-making,a recommendation method for the remaining time prediction model is proposed.Using meta-learning strategies(meta-feature extraction,meta-target model selection,meta-database construction,recommendation model construction,and remaining-time prediction model recommendation)to implement a recommendation method for remaining time prediction models.The validity of the method is proved by designing comparative experiments.(3)By constructing the wind power operation and maintenance process data set,a wind power business process management and forecasting prototype system is established,and the wind power operation and maintenance process is visualized through the business process modeling language,and the remaining time prediction result of the unfinished work ticket process is used to time out judgment and display the timeout time to realize the timeout alarm of the process.