UNIT-V FMM.HYDRAULIC TURBINE - Construction and working
Β
KCC2017 28APR2017
1. Evaluation of demand power prediction
performance based on deep learning
algorithm and data preprocessing
Park Jee Hyun
28 APR 2017
-
SungKyunGwan University
Nanyang Technological University
2. Contents
β’ Purpose
β’ Experiments
β’ Training & Testing Algorithm
β’ Raw vs. Preprocessed
β’ Results
β’ Single Layer Perceptron Network
β’ Multiple Layer Perceptron Network
β’ Convolutional Neural Network
β’ Recurrent Neural Network
β’ Conclusions
β’ Further Studies
3. Purpose
β’ Accurate load forecasting is pivotal for the economic and secure
operation of the power system
β’ Goal: Achieving high load forecasting accuracy
β’ Two main way to do this:
β’ improvement of the performance of the learning algorithm
β’ how well the data features are extracted through the pre-processing
process.
ο¨ In this paper, we will going to evaluate of demand power
prediction performance based on deep learning algorithm
and data preprocessing
5. Training & Testing Algorithm
Input Vector Target Vector
Training
Algorithm
Input Vector
Training
Algorithm
Predict Vector
a) Training
b) Testing
transplant
trained
model
feature data predicting
feature data label data
6. Raw vs. Preprocessed
β’ Given data(Australia NEM data):
β’ Daily Max and Mean Temperature data ( 2 points/day)
β’ Daily power load data with 30 mins interval ( 48 points/day)
β’ Raw data:
β’ [ππππ€
π ] = π π, π πππ₯
π , π ππππ
π ( 50 points/day)
β’ Preprocessed data:
β’ [ππππ
π ] = π π, π π.πππ₯.π πππππ
π
, π π.ππππ.π πππππ
π
(144 points/day)
β’ π π.πππ₯.π πππππ
π
= π π
πππ€ππππ ππ¦ π πππππ πππ₯ ππππππππ‘π’ππ
β’ [π π.ππππ.π πππππ
π
] = π π
πππ€ππππ ππ¦ π πππππ ππππ ππππππππ‘π’ππ
12. Conclusions
β’ All algorithms show better and stable RMSE score with
preprocessed data.
β’ RNN model shows the best performance among four
learning algorithm designed.
ο We checked that it is able to get better performance from
the same predictive model with only limited data if the
data preprocessing process is used.
ο The results show that the RNN model is most suitable for
the power demand forecasting.
13. Further Studies
β’ All neural networks in this paper are based on gradient descent algorithm.
β’ Extreme learning machine(ELM) is a learning algorithm based on single
hidden layer feed-forward neural network (SLFN).
β’ Hierarchical ELM(H-ELM) is a framework of multilayer implementation of
the ELM, which improves the learning performance of the original ELM,
while maintaining its advantages of training efficiency.
ο¨ Need further study about power load predicting algorithm based on ELM
and H-ELM.
ο¨ Check ELM and H-ELM can outperform gradient based machine and find
out the reasons if it outperform.