Artificial Neural Networks for Resources Optimization in Energetic Environment

Abstract

Resource Planning Optimization (RPO) is a common task that many companies need to face to obtain several benefits, like budget improvements and run-time analyses. It is often addressed by using several software products and tools, based on sophisticated mathematical artifacts. However, these tools are not able to provide a practical solution because they are often expensive and time-consuming. On the other hand, Artificial Intelligence-based approaches have been increasingly used in many industrial and scientific fields in last decades, and have demonstrated to be a valid alternative to the classical mathematical-based methods. For this purpose, the following paper aims to investigate the use of multiple Artificial Neural Networks (ANNs) for solving a RPO problem related to the scheduling of different Combined Heat & Power (CHP) generators. The experimental results, carried out by using data extracted by considering a real Microgrid system, have confirmed the effectiveness of the proposed approach. Additionally, we show that multiple neural networks achieve up to a 6% improvement in average accuracy over Naive Bayes classifier, up to a 12% over Multi-Layer Perceptron classifier and up to a 13% over state-of-the-art ANNs in the presence of unbalanced training dataset.

Full text

Articial Neural Networks for Resources
Optimization in Energetic Environment
Gianni D'Angelo (  giadangelo@unisa.it )
University of Salerno: Universita degli Studi di Salerno
Francesco Palmieri
University of Salerno: Universita degli Studi di Salerno
Antonio Robustelli
University of Salerno: Universita degli Studi di Salerno
Research Article
Keywords: Articial Neural Network, Resources Planning Optimization, Energetic Environment, Energetic
Generators, Microgrid System, Articial Intelligence
DOI: https://doi.org/10.21203/rs.3.rs-405315/v1
License:   This work is licensed under a Creative Commons Attribution 4.0 International License. 
Read Full License

· ·

· ·
· · ·

J
J

k T Ng
δ
P SU
SD i −th
k
δi(k) = (1
0
Pi(k)≥0,
SUi(k)≥0,
SDi(k)≥0,
i= 1, ..., Ngk= 0, ..., T −1
J
J:=
T−1
X
k=0
Ng
X
i=1
[CDG
i(Pi(k)) + OMiδi(k) + SUi(k) + SDi(k)]
+Cgrid (k) + ρc
Nc
X
h=1
βh(k)Dc
h(k)
J
i
PNg
i=1[CDG
i(Pi(k)) + OMiδi(k) + SUi(k) + SDi(k)]

Cgrid (k)
ρcPNc
h=1 βh(k)Dc
h(k)
k
T
DG
Ng
Nc
P
CDG(P)
OM
δ
SUi(k)ith
SDi(k)ith
Cgrid (k)
ρc
β
Dc
Pg
δg
cP
cS
Tup
Tup
cSU
cDU
Pg(k)
Pg(k)
δg(k) = 1 δg(k) = 0
k
Pg

δg
k
Pg(k) = (>0
<0
δg(k) = (1
0
k= 0, ..., T −1
k
Pg(k)>0⇐⇒ δg(k) = 1
cPcS
Cg
k
Cg(k) = (cP(k)Pg(k)δg(k) = 1
cS(k)Pg(k)
δi(k)≥δi(k−tup −1) −δi(k−tup −2),
1−δi(k)≥δi(k−tdown −2) −δi(k−tdown −1),
i= 1, ..., Ngtup = 0, ..., min(Tup
i, k−Tup
i+2) tdown = 0, ..., min(Tup
i, k−
Tup
i+ 2)
ith

SUi(k)≥cS U
i(k)[δi(k)−δi(k−1)],
SDi(k)≥cS D
i(k)[δi(k−1) −δi(k)],
SUi(k)≥0,
SDi(k)≥0,
i= 1, ..., Ng
J

batch size =
256

Figures
Figure 1
Boxplot representation of input data distribution.
Figure 2

Distribution of electricity output categories of Jenbacher and Caterpillar generators.
Figure 3
Distribution of thermal output categories of Jenbacher and Caterpillar generators.
Figure 4
Distribution of thermal output categories of Chiller generators.

Figure 5
High-level architecture of each network.