Abstract:Extending the length of time series forecasting has a long-term impact on smart grid energy consumption planning, residential electricity monitoring, extreme weather warning, and other real applications. This article studies the reliable long-term load trend forecasting approaches in smart grid environment. In recent years, the improved neural networks models based on self-attention mechanism show good performances in many sequence tasks, but most studies focus on reducing the complexity of networks layer, and can not restrain the increase of calculation error in longer distance forecasting scenarios. Also, most models lack the ability to mine potential high-dimensional features of time series. Based on these problems, we design a reliable hierarchical self-attention model named as long-term stability network (LTSNet), which adopts a tree-shaped decomposition neural network architecture based on hierarchical residual self-attention blocks, to top-down incrementally mine high-dimensional features of temporal components. At the same time, the attention matrix is used for feature interaction at each layer, to reduce the distribution gap between time series fragments in different domains. Compared with existing study models, LTSNet maintains stable forecasting performance and speed in long-term forecasting services, achieved the most reliable multivariate and univariate forecasting results in multiple domain scenarios. Compared with the latest models, the forecasting accuracy of the proposed model is improved by 22.8 and 13.8, respectively, covering three applications: energy consumption, residential electricity consumption, and weather forecasting. At the same time, our experiments verify that the hierarchical decomposition networks can be used as a backbone architecture to effectively extended to longer dimensional load trend forecasting scenarios.

Temporal Attention Based TCN-BIGRU Model for Energy Time Series Forecasting

The Temperature Prediction of Hydro-generating Units based on Temporal Convolutional Network and Recurrent Neural Network

TCN-Attention-BIGRU: Building energy modelling based on attention mechanisms and temporal convolutional networks

Hybrid Model Based on Temporal Convolutional Network and Attention Mechanism for Short-term Wind Power Forecasting

Temporal Convolutional Networks Applied to Energy-related Time Series Forecasting

TADNet: Temporal Attention Decomposition Networks for Probabilistic Energy Forecasting

Temporal pattern attention for multivariate time series forecasting

Improved Load Forecasting Method Based on TCN-BiGRU and Attention Mechanism

AMTCN: An Attention-Based Multivariate Temporal Convolutional Network for Electricity Consumption Prediction

TSARNet :Integrated Spatio-Temporal Attention Mechanism for Multivariate Time Series Prediction Deep Learning Framework

Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networks

Reliable Long-Term Energy Load Trend Prediction Model for Smart Grid Using Hierarchical Decomposition Self-Attention Network

TATCN: Time Series Prediction Model Based on Time Attention Mechanism and TCN

Short-Term Load Forecasting Based on CNNSeq2seq Model with Temporal Pattern Attention

Short-term multi-energy load forecasting for integrated energy systems based on CNN-BiGRU optimized by attention mechanism

Temporal Chain Network with Intuitive Attention Mechanism for Long-Term Series Forecasting.

Short-mid Term Electricity Consumption Prediction Using Non-Intrusive Attention-Augmented Deep Learning Model

A TCN-BiGRU-based Multi-Energy Consumption Evaluation Approach for Integrated Energy System

Temporal Pattern Attention-Based Sequence to Sequence Model for Multistep Individual Load Forecasting.

Dual-stage Attention-Based Long-Short-term Memory Neural Networks for Energy Demand Prediction

A Novel Short-term Load Forecasting Model by TCN-LSTM Structure with Attention Mechanism