1 Introduction: problem identification
2 Objective of the paper
3 Methodology
3.1 Main machine learning algorithms and techniques
3.1.1 Artificial neural networks (ANN)
-
Input layer. It is the one that gets the information from the outside world.
-
Hidden layers. They process the information internally.
-
Output layer. It is the one that gets the response from the network and transfers it to the outside world.
-
E: Error.
-
y: Correct value.
-
S: Neural network output.
3.1.2 Genetic algorithms
3.2 Selecting specific tools to solve the problem
3.2.1 A systematic literature review
-
+ Article [9] approached the process of optimising thermal energy consumption in La Robla cement factory in Tudela Veguín, Spain, by means of using statistical tools (AI was not applied). The initial planning, along with several initial methodological aspects of this study, served as a reference for our study.
-
+ Article [37] approached a simulation model conducted in four cement plants in South Africa, which enabled as much as 7.1% energy savings (electricity savings and thermal energy savings). To reach those savings, a managing methodology, which considered the crude mill, the kiln, the coal mill, crushers, cement mills and other auxiliary machinery, was carried out. Hence, a model was developed to distribute the load among the different days of the week and seasons of the year and always searching for the off-peak energy times. Although it also addressed thermal energy cost, the analysis of this methodology was extremely useful for the present research when shifting electricity consumption to the most economical off-peak periods.
-
+ Article [36] approached the problem of the cement market in South Africa, where a remarkable increase in electricity costs took place, which brought about an important pressure on international producers. An energy management system (EMS) was implemented in order to cut the electricity bill. This system, apart from controlling the periods of highest consumption, shifted electricity demand to the most economical daily off-peak times.
-
+ Article [35] focussed on managing and controlling thermal energy. Some of the factors detailed in this article were quite useful for our research, although they did not refer to electricity consumption.
3.2.2 Surveys and expert panel
3.3 Development of the working methodology: application of specific tools
3.3.1 A methodology for operational electricity optimisation
-
CE(ti): electricity consumption/crusher electricity power (kW).
-
CE(M): electricity consumption/the raw mill electricity power (kW).
-
CE(H): electricity consumption/kiln electricity power (kW).
-
CE(MC): electricity consumption/coal mill electricity power (kW).
-
CE(MCEM): electricity consumption/cement mill electricity power (kW).
-
CE(Fx): stage X net electricity consumption (kWh).
-
V(n)FX: “n” variable with impact on electricity consumption in stage X, being n = 1, 2, …, X = 1, 3, 4, 5 y 6.
-
\(\omega\)nX: variable “n” weight in stage X.
-
\(u\): the value obtained in each of the neurons in the network will be a linear regression model to which a bias (u) should be added.
-
Size of crusher 1 output product (primary crusher): V(1)F1(t1).
-
Size of crusher 2 output product (secondary crusher): V(2)F1(t2).
V(1)F1(t1) | V(2)F1(t1) | CE(F1)(t1) |
---|---|---|
a | b | X |
c | d | Z |
e | f | Y |
-
CE(t1): Primary crusher electricity consumption.
-
H(t1): Primary crusher operating hours.
-
u1: Bias.
-
CE(t2): Secondary crusher electricity consumption.
-
H(t2): Secondary crusher operating hours.
-
u2: Bias.
-
Input raw material size: V(1)F3.
-
Mill rotational speed: V(2)F3.
-
Mill filling level: V(3)F3.
-
Balls diameter: V(4)F3.
-
CE(M): Raw mill power consumption.
-
H(M): Raw mill operating hours.
-
u3: Bias.
-
False air entering cyclones: V(1)F4
-
Clinker granulometry: V(2)F4
-
CE(H): Kiln electricity consumption.
-
H(H): Kiln operating hours.
-
u4: Bias.
-
Carbon size/granulometry: V(1)F5.
-
Mill rotational speed: V(2)F5.
-
CE(MC): Coal mill electricity consumption.
-
H(H): coal mill operating hours.
-
u5: Bias.
-
Clinker size: V(1)F6(M5,M6,M7).
-
Plaster size: V(2)F6(M5,M6, M7).
-
Other additions size: V(3)F6(M5,M6,M7).
-
Mill rotational speed: V(4)F6(M5,M6,M7).
-
CE(MCEM): Cement mill electricity consumption.
-
H(H): Cement mill operating hours.
-
u6: bias.
V(1)F6(M5, M6, M7) | V(2)F6 (M5, M6, M7) | V(3)F6(M5, M6, M7) | V(4)F6(M5, M6, M7) | CE(F6)(M5, M6, M7) |
---|---|---|---|---|
a | B | c | d | X |
e | F | g | h | Z |
i | J | k | l | Y |
-
i: Number of the variable.
-
j: Stage it affects.
3.3.2 A methodology for the optimisation of regulated electricity prices
-
CE(FX). Total electricity consumption in stage X. It will be measured in kWh.
-
CEFX P(Y). Hourly electricity consumption in stage X during Y period. It will be measured in kw.
-
HFX P(Y). Working hours in stage X during period Y. It will be measured in h.
-
CEF(1)(t1,t2)=CE F1 P(1)(t1)*HF1P(1)(t1) + CEF1P(1)(t2)*HF1P(1)(t2) + CEF1P(2)(t1)*HF1P(2)(t1) + CEF1P(2)(t2)*HF1P(2)(t2) + CEF1P(3)(t1)*HF1P(3)(t1) + CEF1P(3)(t2)*HF1P(3)(t2) + CEF1P(4)(t1)*HF1P(4)(t1) + CEF1P(4)(t2)*HF1P(4)(t2) + CEF1P(5)(t1)*HF1P(5)(t1) + CEF1P(5)(t2)*HF1P(5)(t2) + CEF1P(6)(t1)*HF1P(6)(t1) + CEF1P(6)(t2)*HF1P(6)(t2).
-
CE(F2) = CEF2P(1)*HF2P(1) + CEF2P(2)*HF2P(2) + CEF2P(3)*HF2P(3) + CEF2P(4)*HF2P(4) + CEF2P(5)*HF2P(5) + CEF2P(6)*HF2P(6).
-
CE(F3) = CEF3P(1)*HF3P(1) + CEF3P(2)*HF3P(2) + CEF3P(3)*HF3P(3) + CEF3P(4)*HF3P(4) + CEF3P(5)*HF3P(5) + CEF3P(6)*HF3P(6)
-
CE(F4) = CEF4P(1)*HF4P(1) + CEF4P(2)*HF4P(2) + CEF4P(3)*HF4P(3) + CEF4P(4)*HF4P(4) + CEF4P(5)*HF4P(5) + CEF4P(6)*HF4P(6)
-
CE(F4) = CEF5P(1)*HF5P(1) + CEF5P(2)*HF5P(2) + CEF5P(3)*HF5P(3) + CEF5P(4)*HF5P(4) + CEF5P(5)*HF5P(5) + CEF5P(6)*HF5P(6).
-
CE(F6)(M5,M6,M7) = CEF6P(1)(M5)*HF6P(1)(M5) + CEF6P(1)(M6)*HF6P(1)(M6) + CEF6P(1)(M7)*HF6P(1)(M7) + CEF6P(2)(M5)*HF6P(2)(M5) + CEF6P(2)(M6)*HF6P(2)(M6) + CEF6P(2)(M7)*HF6P(2)(M7) + CEF6P(3)(M5)*HF5P(3)(M5) + CEF6P(3)(M6)*HF5P(3)(M6) + CEF6P(3)(M7)*HF5P(3)(M7) + CEF6P(4)(M5)*HF6P(4)(M5) + CEF6P(4)(M6)*HF6P(4)(M6) + CEF6P(4)(M7)*HF6P(4)(M7) + CEF6P(5)(M5)*HF6P(5)(M5) + CEF6P(5)(M6)*HF6P(5)(M6) + CEF6P(5)(M7)*HF6P(5)(M7)+CEF6P(6)(M5)*HF6P(6)(M5) + CEF6P(6)(M6)*HF6P(6)(M6)+CEF6P(6)(M7)*HF6P(6)(M7)
-
CE(F7) = CEF7P(1)*HF7P(1) + CEF7P(2)*HF7P(2) + CEF7P(3)*HF7P(3) + CEF7P(4)*HF7P(4) + CEF7P(5)*HF7P(5) + CEF7P(6)*HF7P(6)
-
CE(F8) = CEF8P(1)*HF8P(1) + CEF8P(2)*HF8P(2) + CEF8P(3)*HF8P(3) + CEF8P(4)*HF8P(4) + CEF8P(5)*HF8P(5) + CEF8P(6)*HF8P(6).
-
CE(F9) = CEF9P(1)*HF9P(1) + CEF9P(2)*HF9P(2) + CEF9P(3)*HF9P(3) + CEF9P(4)*HF7P(4) + CEF9P(5)*HF9P(5) + CEF9P(6)*HF9P(6)
Mw | |
---|---|
MAXIMUM POWER—TWO TRANFORMERS | 20 + 20 |
CONTRACTED POWER IN P1 | 14 |
CONTRACTED POWER IN P2 | 14 |
CONTRACTED POWER IN P3 | 17.2 |
CONTRACTED POWER IN P4 | 17.2 |
CONTRACTED POWER IN P5 | 17.2 |
CONTRACTED POWER IN P6 | 17.4 |
-
CE(P1) = CEF1P(1) + CEF2P(1) + CEF3P(1) + CEF4P(1) + CEF5P(1) + CEF6P(1) + CEF7P(1) + CEF8P(1) + CEF9P(1) ≤ 14 Mw
-
If 0 < = CE(P1) ≤ C′ Mw (1 month under normal conditions) → Access rate may be reduced.
-
CE(P2) = CEF1P(2) + CEF2P(2) + CEF3P(2) + CEF4P(2) + CEF5P(2) + CEF6P(2) + CEF7P(2) + CEF8P(2) + CEF9P(2) ≤ 14 Mw.
-
If 0 < = CE(P2) ≤ C′ Mw (1 month under normal conditions) → Access rate may be reduced.
-
CE(P3) = CEF1P(3) + CEF2P(3) + CEF3P(3) + CEF4P(3) + CEF5P(3) + CEF6P(3) + CEF7P(3) + CEF8P(3) + CEF9P(3) ≤ 17.2 Mw.
-
If 0 ≤ CE(P3) ≤ C′ (1 month under normal conditions) → Access rate may be reduced.
-
CE(P4) = CEF1P(4) + CEF2P(4) + CEF3P(4) + CEF4P(4) + CEF5P(4) + CEF6P(4) + CEF7P(4) + CEF8P(4) + CEF9P(4) ≤ 17.2 Mw
-
If 0 ≤ CE(P4) ≤ C′ (1 month under normal conditions) → Access rate may be reduced.
-
CE(P5) = CEF1P(5) + CEF2P(5) + CEF3P(5) + CEF4P(5) + CEF5P(5) + CEF6P(5) + CEF7P(5) CEF8P(5) + CEF9P(5) ≤ 17.2 Mw.
-
If 0 ≤ CE(P5) ≤ C′ (1 month under normal conditions) → Access rate may be reduced.
-
CE(P6) = CEF1P(6) + CEF2P(6) + CEF3P(6) + CEF4P(6) + CEF5P(6) + CEF6P(6) + CEF7P(6) + CEF8P(6) + CEF9P(6) ≤ 17.4 Mw
-
If 0 ≤ CE(P5) ≤ C′ (1 month under normal conditions) → Access rate may be reduced.
-
The maximum production capacity of each stage of the cement manufacturing process minus the maximum capacity of next stage should be > 0, and at the same time ≤ than the storage capacity of raw material, fuel, by-product or final product at each stage of the process (7 functions).
-
Time constraints (2 functions).
-
CPF1(t1): maximum production capacity per hour, in the extraction of raw materials stage. 300 t/h.
-
CPF1(t2): maximum production capacity per hour, in the extraction of raw materials stage. 520 t/h.
-
CAMp: maximum storage capacity for crushed raw materials.
-
CPF2: maximum production capacity per hour in the homogenisation stage. 320 t/h.
-
CAMpH: maximum homogenised raw material capacity.
-
CPF3: maximum production capacity per hour in the crude grinding stage. 550 t/h.
-
CACr: maximum crude oil storage capacity.
-
CPF4: maximum production capacity per hour in clinker manufacturing process/kiln. For raw materials: 240 t/h. For clinker: 154 t/h.
-
CACk: maximum clinker storage capacity.
-
CPF5: maximum production capacity per hour in the coal grinding stage.17.1 t/h.
-
CACb: maximum coal storage capacity.
-
CPF6: maximum production capacity per hour in the cement grinding stage:
-
CPF6(M5)—Cement mill 5: 29.2 t/h.
-
CPF6(M6)—Cement mill 6: 78 t/h.
-
CPF6(M7)—cement mill 7: 34.3 t/h.
-
CPF7: maximum production capacity per hour in the transportation stage.
-
CASL_CEM(Z): maximum cement storage capacity in silo Z.
-
CPF8: maximum production capacity per hour in the packing stage.
-
Line 1: 67 t/h (2700 sacks).
-
Line 2: 55 t/h (2200 sacks).
-
CAEns: maximum storage capacity in sacks storehouse.
-
CPF9: maximum production capacity per hour in the auxiliary services stage.
-
VECEM(Z): SALES of packaged cement type Z, being Z = 1, …, n.
-
VGCEM(Z): sales of bulk cement type Z, being Z = 1, …, n.
-
CC: consumption of coal, measured in t/h.
-
CAMp ≥ CPF1(t1)*[HF1P(1)(t1) + HF1P(2)(t1) + HF1P(3)(t1) + HF1P(4)(t1) + HF1P(5)(t1) + HF1P(6)(t1)] + CPF1(t2)*[HF1P(1)(t2) + HF1P(2)(t2) + HF1P(3)(t2) + HF1P(4)(t2) + HF1P(5)(t2) + HF1P(6)(t2)] − CPF2*[HF1P(1) + HF1P(2) + HF1P(3) + HF1P(4) + HF1P(5) + HF1P(6)] ≥ 0.
-
CAMpH ≥ CPF2*[HF2P(1) + HF2P(2) + HF2P(3) + HF2P(4) + HF2P(5) + HF2P(6)] − CPF3*[HF3P(1) + HF3P(2) + HF3P(3) + HF3P(4) + HF3P(5) + HF3P(6)]) ≥ 0
-
CACr ≥ CPF3*[HF3P(1) + HF3P(2) + HF3P(3) + HF3P(4) + HF3P(5) + HF3P(6)] − CPF4*[HF4P(1) + HF4P(2) + HF4P(3) + HF4P(4) + HF4P(5) + HF4P(6)] ≥ 0
-
CACk ≥ CPF4*[HF4P(1) + HF4P(2) + HF4P(3) + HF4P(4) + HF4P(5) + HF4P(6)] − CPF6*[HF6P(1) + HF6P(2) + HF6P(3) + HF6P(4) + HF6P(5) + HF6P(6) ≥ 0
-
CACb ≥ CPF5*[HF5P(1) + HF5P(2) + HF5P(3) + HF5P(4) + HF5P(5) + HF5P(6)] − CC*[HF6P(1) + HF6P(2) + HF6P(3) + HF6P(4) + HF6P(5) + HF6P(6)] ≥ 0
-
CASL_CEM(Z) ≥ CPF6(M5)*[HF6P(1)(M5) + HF6P(2)(M5) + HF6P(3)(M5) + HF6P(4)(M5) + HF6P(5)(M5) + HF6P(6)(M5)] + CPF6(M6)*[HF6P(1)(M6) + HF6P(2)(M6) + HF6P(3)(M6) + HF6P(4)(M6) + HF6P(5)(M6) + HF6P(6)(M6)] + CPF6(M7)*[HF6P(1)(M7) + HF6P(2)(M7) + HF6P(3)(M7) + HF6P(4)(M7) + HF6P(5)(M7) + HF6P(6)(M7)] − CPF7[HF7P(1) + HF7P(2) + HF7P(3) + HF7P(4) + HF7P(5) + HF7P(6)] − CPF8*[HF8P(1) + HF8P(2) + HF2P(3) + HF8P(4) + HF8P(5) + HF8P(6)] − VGCEM(Z) ≥ 0
-
CAEns ≥ CPF8*[HF8P(1) + HF8P(2) + HF2P(3) + HF8P(4) + HF8P(5) + HF8P(6)] − VECEM(Z) ≥ 0.