1
A Multi-task End-to-End Multivariate
Long-sequence Time Series Prediction Model for
Load Forecasting
Ziyuan Zhang, Yuanzheng Li, Senior Member, IEEE, Yang Li, Senior Member, IEEE,
Yun Liu, Senior Member, IEEE, Xiangpeng Xie, Senior Member, IEEE and Zhigang Zeng, Fellow, IEEE
Abstract—With the increasing complexity of the power system,
and the growing global demand for electricity, accurate and
effective forecasting of electricity load, electricity prices, and
related meteorological features becomes increasingly crucial.
However, existing methods tend to use only one element, for
example, electricity load, as their prediction target instead of
generating multiple indexes (e.g., electricity load, electricity
prices, and related meteorological features) predictions jointly.
This is insufficient for electricity producers to make decisions.
While some time series forecasting methods can predict multiple
indicators simultaneously, they often overlook the correlations
among these features. These approaches miss the potential
performance improvement that modeling correlations among
multiple series could bring to the prediction results. To address
the aforementioned concerns, this study presents a model named
PatchGRU. Once trained, PatchGRU is capable of predicting
long-sequences of electricity loads, electricity prices, and related
meteorological features in a single forward propagation. The
model differentiates between time-variant and time-invariant
components using a Temporal Variance Separator, processing
them separately. For the time-variant part, the network uses
the proposed multi-scale patch input in place of traditional
point inputs, then feeds it through the Gate Recurrent Unit
(GRU), mean supervision, and Local-Global Feature Interaction
modules, ultimately generating the forecast results. Moreover,
this model relies solely on GRU and Multilayer Perceptrons
(MLPs) and is much simpler than the Transformer structure.
Experimental results show that the proposed model outperforms
state-of-the-art models in different regions, time periods, and
sampling frequencies, achieving the best results in forecasts with
a maximum lead time of up to 30 days.
Index Terms—Load forecasting, Time series forecasting, Long-
sequence, Neural networks, Deep learning
I. I NTRODUCTION
A
S global energy systems expand and electricity con-
sumption patterns diversify, the controllability of power
systems decreases. Meteorological factors like temperature and
The work is supported by the National Natural Science Foundation of China
under Grant No. 62422308 and No. 52377081.
Ziyuan Zhang, Yuanzheng Li and Zhigang Zeng are with the School
of Artificial Intelligence and Automation, Huazhong University of Sci-
ence and Technology, Wuhan 430074, China (e-mail:zhang zy@hust.edu.cn;
Yuanzheng Li@hust.edu.cn; zgzeng@hust.edu.cn).
Yang Li is with the School of Electrical Engineering, Northeast Electric
Power University, Jilin 132012, China (e-mail: liyang@neepu.edu.cn).
Yun Liu is with the School of Electric Power Engineering, South
China University of Technology, Guangzhou 510641, China (e-mail: li-
uyun19881026@gmail.com).
Xiangpeng Xie is with the Institute of Advanced Technology, Nanjing
University of Posts and Telecommunications, Nanjing 210023, China (e-mail:
xiexiangpeng1953@163.com).
humidity significantly impact electricity load, while complex
factors like electricity prices interact with load in deeper, un-
derlying ways. With enhanced data accessibility, constructing
an accurate and integrated forecasting system has become
feasible [1], [2].
Considerable research has been conducted on electricity
load forecasting, but two main issues remain. First, most
studies focus on short-term load forecasts, while long-term
predictions could better support system scheduling and man-
agement. Second, many models predict only a single load
variable. In practice, forecasting multiple variables—such as
electricity prices, regional renewable energy data (wind speed,
solar irradiance), and meteorological features—could reveal
hidden relationships and provide managers with more reliable
decision-making information, thereby facilitating subsequent
operations for power system [3]–[5]. Forecasting electricity
load and its related features can be approached as a time
series problem, using either statistical or machine learning-
based methods.
Statistical methods remain widely used in electricity load
forecasting. Early approaches included the naive forecast (NF)
[6], which uses historical data as forecast values. Common
methods include the Autoregressive (AR) model [7], Autore-
gressive Moving Average (ARMA) [8], and Autoregressive
Integrated Moving Average (ARIMA) [9]. These models,
while advantageous for their fast computation and low data
requirements, are limited by their linear nature, making them
less effective for complex, nonlinear load patterns [10].
Machine learning methods are also applied to forecast elec-
tricity load, prices, and meteorological features. Some scholars
use Support Vector Machines (SVM) with optimized param-
eters [11], [12]. As machine learning advances, regression
algorithms such as eXtreme Gradient Boosting (XGBoost)
and Light Gradient Boosting Machine (LightGBM) gained
popularity. These models excel in nonlinear mapping and
generally offer better predictive accuracy without assuming
data characteristics. However, they often lack strong support
for multivariate and multi-step forecasting.
Following the substantial achievements of artificial neural
networks in computer vision and natural language processing,
many researchers have also attempted to use various neural
network architectures for time series forecasting. Some of
these have already been applied to the prediction of electricity
load, while others have not yet been utilized in this field.
From the perspective of neural network architectures, existing
This article has been accepted for publication in IEEE Transactions on Smart Grid. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSG.2025.3605653
© 2025 IEEE. All rights reserved, including rights for text and data mining and training of artificial intelligence and similar technologies. Personal use is permitted,
but republication/redistribution requires IEEE permission. See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: Tamkang Univ.. Downloaded on October 10,2025 at 14:41:36 UTC from IEEE Xplore. Restrictions apply.