Abstract:Industrial time series involves a large amount of production process information, which effectively reflects the production status of the industrial process. To better understand characteristics and patterns of changes in production conditions, it is crucial to analyze and predict industrial time series data. Given the involvement of numerous parameters and complex physical-chemical reactions in industrial processes, attaining precise predictive performance utilizing a single model remains a formidable challenge. In this paper, we propose a novel hybrid deep learning prediction method based on spatio-temporal attention and temporal convolution network. The proposed method aims to handle the multivariate coupling characteristics and dynamic nonlinear features in industrial time series through different model structures for accurate prediction. In this method, historical data are first segmented into multiple consecutive inputs along the temporal dimension, which are then used as inputs to the subsequent attention mechanism module. To realize the mapping from points to series in the temporal dimension, the segmented input is processed using both the adaptive attention mechanism and one-dimensional convolution. Then the spatio-temporal coupling features are further explored through the spatio-temporal attention model. In addition, to extract dynamic nonlinear features from historical data, a parallel temporal convolutional network with temporal pattern attention is utilized. In order to evaluate the prediction performance of the proposed model, we use two different real-world industrial time series datasets for comprehensive evaluation. The experimental results demonstrate the effectiveness and accuracy of the proposed method. Code is available at https://github.com/TensorPulse/MACnet.

Industrial Time Series Prediction Based on Incremental DBSCAN-KNN with Self-learning Scheme

Multi Time Scale Inception-Time Network for Soft Sensor of Blast Furnace Ironmaking Process

Time Series Prediction Method of Industrial Process With Limited Data Based on Transfer Learning

Prediction intervals for industrial data with incomplete input using kernel-based dynamic Bayesian networks

A Process Knowledge-Based Hybrid Method for Univariate Time Series Prediction with Uncertain Inputs in Process Industry

Variational Inference Based Kernel Dynamic Bayesian Networks for Construction of Prediction Intervals for Industrial Time Series with Incomplete Input

Self-Supervised Learning for Time-Series Anomaly Detection in Industrial Internet of Things

Spatio-temporal attention-based hybrid deep network for time series prediction of industrial process

A Spatiotemporal Industrial Soft Sensor Modeling Scheme for Quality Prediction With Missing Data

Chaotic Time Series Online Prediction Based on Improved Kernel Adaptive Filter.

Time Series Online Prediction Based On Adaptive Dynamic Adjustment Kernel Recursive Least Squares Algorithm

Hierarchical Neural Networks for Multivariate Time Series Prediction

Incremental Learning and Conditional Drift Adaptation for Nonstationary Industrial Process Fault Diagnosis.

Improved multi-kernel LS-SVR for time series online prediction with incremental learning

An Active Learning-Based Incremental Deep-Broad Learning Algorithm for Unbalanced Time Series Prediction

Time Series Prediction with Input Noise Based on the ESN and the EM and Its Industrial Applications

Industrial Prediction Intervals With Data Uncertainty

A General Data Renewal Model for Prediction Algorithms in Industrial Data Analytics

Self-Supervised Time Series Clustering With Model-Based Dynamics

Fractional stochastic configuration networks-based nonstationary time series prediction and confidence interval estimation

Multivariate Time Series Online Prediction Based on Adaptive Normalized Sparse Kernel Recursive Least Squares Algorithm