Abstract:The remarkable achievements of large models in the fields of natural language processing (NLP) and computer vision (CV) have sparked interest in their application to time series forecasting within industrial contexts. This paper explores the application of a pre-trained large time series model, Timer, which was initially trained on a wide range of time series data from multiple domains, in the prediction of Supervisory Control and Data Acquisition (SCADA) data collected from wind turbines. The model was fine-tuned on SCADA datasets sourced from two wind farms, which exhibited differing characteristics, and its accuracy was subsequently evaluated. Additionally, the impact of data volume was studied to evaluate the few-shot ability of the Timer. Finally, an application study on one-turbine fine-tuning for whole-plant prediction was implemented where both few-shot and cross-turbine generalization capacity is required. The results reveal that the pre-trained large model does not consistently outperform other baseline models in terms of prediction accuracy whenever the data is abundant or not, but demonstrates superior performance in the application study. This result underscores the distinctive advantages of the pre-trained large time series model in facilitating swift deployment.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the time - series prediction problem of wind turbine SCADA (Supervisory Control and Data Acquisition) data in wind power plants. Specifically, the researchers try to apply pre - trained large - scale time - series models to improve the prediction accuracy of wind turbine SCADA data and evaluate their performance under different data volumes and prediction scenarios. #### Main problem background 1. **Importance of wind turbine SCADA data**: - The data collected by the SCADA system plays a crucial role in the operation and maintenance (O&M) of wind turbines, including data pre - processing, fault diagnosis, and wind power generation prediction, etc. - Improving the accuracy of time - series prediction is essential for enhancing the effectiveness and reliability of these applications, thereby helping to improve the efficiency and profitability of renewable energy. 2. **Limitations of existing methods**: - Existing time - series prediction methods are mainly based on deep learning, such as Recurrent Neural Networks (RNN), Graph Neural Networks (GNN), and Transformer, etc. Although these methods have improved in accuracy, they have not fully utilized the advantages brought by large - scale models. - A method that can be quickly deployed and has strong adaptability is needed to meet the diverse needs in wind turbine operation and maintenance. #### Research objectives 1. **Apply large - scale time - series models**: - Use pre - trained large - scale time - series models (such as Timer) to predict wind turbine SCADA data and evaluate their performance under different data volumes and prediction scenarios. 2. **Compare with traditional models**: - By comparing large - scale time - series models with other traditional models (such as LSTM, Transformer, etc.), analyze their advantages and disadvantages, especially in few - shot learning and generalization ability. 3. **Explore specific application scenarios**: - Design an application scenario where a single turbine is fine - tuned for the prediction of the entire wind farm, and verify the ability of large - scale time - series models in few - shot learning and cross - turbine generalization. #### Experimental design and results - **Experimental setup**: - Experiments were carried out on SCADA datasets of two actual wind power plants, located in different regions in China, one is an onshore wind farm (Plant 1), and the other is an offshore wind farm (Plant 2). - The data was cleaned and pre - processed to ensure data quality. - **Experimental results**: - In the case of sufficient data, the large - scale time - series model is not significantly better than other baseline models. - In the case of limited data, the pre - trained and fine - tuned large - scale time - series model shows better few - shot learning ability, especially outstanding in short - term prediction. - The experiment of fine - tuning a single turbine for the prediction of the entire wind farm shows that the large - scale time - series model not only has few - shot learning ability but also can achieve cross - turbine generalization, with comprehensive accuracy advantages. In summary, this paper explores the applicability and advantages of pre - trained large - scale time - series models in the prediction of wind turbine SCADA data through application, especially their abilities in few - shot learning and cross - device generalization.

Fine-Tuning Pre-trained Large Time Series Models for Prediction of Wind Turbine SCADA Data

Winds of Change: Enhancing Wind Power Generation Forecasting with LSTM Models and Advanced Techniques in a Single Turbine Wind Farm

Fatigue Load Prediction of Large Wind Turbine by Big Data and Deep Learning

Wind Power Forecasting with Deep Learning Networks: Time-Series Forecasting

WindTrans: Transformer-Based Wind Speed Forecasting Method for High-Speed Railway

Equipment Health Assessment: Time Series Analysis for Wind Turbine Performance

Exploring Time Series Models for Wind Speed Forecasting: A Comparative Analysis

A Hybrid Wind Speed Prediction Approach Based on Ensemble Empirical Mode Decomposition and BO-LSTM Neural Networks for Digital Twin

Power prediction using high-resolution SCADA data with a farm-wide deep neural network approach

Comparison of deep learning models for multivariate prediction of time series wind power generation and temperature

Time Series Data-Driven Online Prognosis of Wind Turbine Faults in Presence of SCADA Data Loss

A hybrid deep learning model based on parallel architecture TCN-LSTM with Savitzky-Golay filter for wind power prediction

Wind power prediction based on high-frequency SCADA data along with isolation forest and deep learning neural networks

Short-Term Wind Power Prediction Based on Multi-Parameters Similarity Wind Process Matching and Weighed-Voting-Based Deep Learning Model Selection

Time series cross-correlation network for wind power prediction

Predictive maintenance of abnormal wind turbine events by using machine learning based on condition monitoring for anomaly detection

Modeling Wind Turbine Performance and Wake Interactions with Machine Learning

Ultra-short-term multi-step wind speed prediction for wind farms based on adaptive noise reduction technology and temporal convolutional network

An Ultra‐Short‐Term Multi‐Step Prediction Model for Wind Power Based on Sparrow Search Algorithm, Variational Mode Decomposition, Gated Recurrent Unit, and Support Vector Regression

Auxiliary Knowledge-Based Fine-Tuning Mechanism for Industrial Time-Lag Parameter Prediction

An ultra‐short‐term wind speed prediction model using LSTM based on modified tuna swarm optimization and successive variational mode decomposition