Electrical Engineering
https://doi.org/10.1007/s00202-020-01135-y
ORIGINAL PAPER
Long short-term memory-singular spectrum analysis-based model for
electric load forecasting
Neeraj Neeraj
1
· Jimson Mathew
1
· Mayank Agarwal
1
· Ranjan Kumar Behera
1
Received: 6 May 2019 / Accepted: 20 October 2020
© Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract
Electrical load forecasting is a key player in building sustainable power systems and helps in efficient system planning.
However, the irregular and noisy behavior in the observed data makes it difficult to achieve better forecasting accuracy.
To handle this, we propose a new model, named singular spectrum analysis-long short- term memory (SSA-LSTM). SSA
is a signal processing technique used to eliminate the noisy components of a skewed load series. LSTM model uses the
outcome of SSA to forecast the final load. We have used five publicly available datasets from the Australian Energy Market
Operator (AEMO) repository to assess the performance of the proposed model. The proposed model has superior forecasting
accuracy compared to other existing state-of-the-art methods [persistence, autoregressive (AR), AR-exogenous, ARMA-
exogenous (ARMAX), support vector regression (SVR), random forest (RF), artificial neural network (ANN), deep belief
network (DBN), empirical mode decomposition (EMD-SVR), EMD-ANN, ensemble DBN, and dynamic mode decomposition
(DMD)] for half-hourly and one day ahead load forecasting using RMSE and MAPE error metrics.
Keywords Short-term load forecasting · Singular spectrum analysis · Long short-term memory · Australian energy market
operator
Nomenclature
AEMO Australian Energy Market Operator repos-
itory
AR Autoregressive model
ARMAX Autoregressive Moving Average exogenous
SVR Support Vector Regression
RF Random Forest
ANN Artificial Neural Network
DBN Deep Belief Network
EMD-SVR Empirical Mode Decomposition-Support
Vector Regression
EMD-ANN Empirical Mode Decomposition-Artificial
Neural Network
EDBN ensemble Deep Belief Network
DMD Dynamic Mode Decomposition
STLF Short-Term Load Forecasting
RBM Restricted Boltzmann Machines
IMFs Intrinsic Mode Functions
B Neeraj Neeraj
neeraj.pcs17@iitp.ac.in
1
Department of Computer Science and Engineering, Indian
Institute of Technology, Patna, Bihar, India
VMD Variational Mode Decomposition
SVD Singular Value Decomposition
INFS Integrated Nonlinear Feature Selection
ANFIS Adaptive Neuro-Fuzzy Inference System
SOFM self-organizing feature map
f
t
Forget gate layer of LSTM
i
t
Input gate layer of LSTM
C
t
Cell state of LSTM
o
t
Output gate of LSTM
h
t
Final output of LSTM
Y Hankel matrix having equal elements on the
diagonals
L Window size in constructing the Hankel
matrix
EVG Eigenvalue Grouping
EV Eigenvalues
1 Introduction
Electric load forecasting is increasingly attaining focus day
by day. It is most useful in making power systems more
intelligent, efficient, sustainable, and reliable. The most crit-
123