ISSN: 2641-3086
Trends in Computer Science and Information Technology
Research Article       Open Access      Peer-Reviewed

Classification by Artificial Neural Network according to the values affecting Electricity Generation

Furkan Esmeray1 and Sevcan Aytaç Korkmaz2*

1Electric and Energy Department, Munzur University, Tunceli, Turkey
2Electronic and Automation Department, Fırat University, Elazığ, Turkey
*Corresponding author: Sevcan Aytaç Korkmaz, Electronic and Automation Department, Fırat University, Elazığ, Turkey, E-mail: sevcanaytackorkmaz@gmail.com
Received: 03 April, 2018 | Accepted: 12 April, 2018 | Published: 18 April, 2018

Cite this as

Esmeray F, Korkmaz SA (2018) Classification by Artificial Neural Network according to the values affecting Electricity Generation. Trends Comput Sci Inf Technol 3(1): 001-004. DOI: 10.17352/tcsit.000006

Predicting the amount of electricity produced in a power plant is very important for today’s economy. Oven Power (MW), Boiler Input Gas Temperature, Superheated Steam Amount, ID-Fan Speed, Feeding Water Tank data affect the electricity production. In this article, Etikrom A.Ş. The electricity production amount to be produced in Elazığ Etikrom A.Ş. was estimated by using the data of Oven Power (MW), Water Inlet Gas Temperature, Steam Vapor Volume, ID-Fan Speed, Feeding Water Tank data. Electricity generation amount is used as verification data. That is, by the k-means clustering method, the electricity generation amount is divided into 3 classes (low, medium, and high). 3621 data including Oven Power (MW), Boiler Input Gas Temperature, Superheated Steam Amount, ID-Fan Speed, and Feeding Water Tank data were used after class 3 separation. With the K-means clustering method, 2742 of these data were clustered as low electricity, 296 as medium electricity and 583 as high electricity. This clustered data was given to the Artifical Neural Network classifier. The success rate obtained as a result of this classification is 85.81%. Classified data were analyzed by ROC curve.

Introduction

Nowadays, various sources are used to meet the need for electricity. One of the ways to meet energy needs is electricity generation with steam . Steam is the passage of water through various stages and from liquid to gas. In this article, we describe the transformation of electricity to electricity by generating steam and we have examined and classified the variables that affect the electricity production values and production values. The data belong to the year 2017 and are registered for every hour. We have ETİKROM A.Ş. by observing the data we have observed in the cogeneration plant, we have observed by classifying variables affecting electricity generation and electricity generation. In a steam power unit; high pressure steam boiler, high pressure steam turbine, condenser, feed water tank, superheater. In the demineralization unit, water with a maximum of 0.2 μS conductivity and a minimum pH of 9.4 is sent to the feed water tank.

The water temperature in the feed water tank is increased a little. Burning of the arc furnaces with 23 MW installed power, which is used to melt the chrome, is taken into the boiler by vacuuming with the help of the resulting flue gas ID-Fan. Water is circulated through the boiler with the help of pipes. Boiler; high pressure economizer, high pressure evaporator, high pressure superheater, high pressure steam drum. The water in the pipes is interacted with the reverse flow by the gas taken in. The steam obtained as hot steam in the boiler is taken to the high pressure steam drum. The pressurized steam is sent from the high pressure steam drum to the steam turbine. The steam sent to the steam türbine is hot. Steam with 300 ° C and 30 bar pressure is hot and steam turbine is sent to these specifications. Convection steam is condensed in the condenser and the condensate is pumped back to the feed water tank by the condenser pumps and the closed cycle is completed.

In addition, the feed water tank is kept warm continuously. This is possible by constant steam heating from the high pressure steam boiler. The harmful and suspended substances in the system are taken out by the bluff tank. Synchronized generator with steam turbine is operated and electricity generation is provided. The amount of electricity produced depends on some variables, from steam production, steam to the steam turbine, and the synchronous generator connection of the steam turbine with the help of reducers. These variables are the oven power, the boiler gas inlet temperature, the total amount of steam produced, the feed water tank temperature elevated water temperature, the ID fan speed, and the ambient temperature. If these variables are to be explained, the working power of the furnace is directly proportional to the outlet temperature of the furnace. That is, when the oven power is high, the flue gas temperature is also high. Therefore, the flue gas temperature directly refers to the boiler inlet temperature, that is, the temperature of the gas heating the water. The higher this temperature, the better the evaporation will be. It is easier to evaporate the water with high temperature because the feed water is heated by preheating the water in the tank. This means that the steam that needs to be created is easier to build. Because the heat required to evaporate cold water is higher than the heat that must be supplied to heat the hot water. The higher the amount of steam produced, the higher the amount of electricity produced. The number of active stages of the turbine will increase and the force generated will increase in direct proportion to the amount of steam. The ID-Fan devride affects steam production. If the ID-Fan cycle is not adjusted according to the flow of the flue gas, the steam production amount will decrease. If the gas flow is low, and the ID-Fan cycle is not at the proper value and the gas passes quickly through the boiler, the flue gas heat will leave the boiler without passing the water heat through the pipes, which will reduce the amount of steam. It also affects the amount of steam production and the amount of electricity generated at the outside temperature. When the outdoor temperature is low, the temperature of the flue gas in the flue gas channels will decrease. This will cause the boiler inlet gas temperature to decrease. In addition, when the ambient temperature is low on the steam transmission lines, the transmission will be condensed at the surface of the line, and the steam temperature and steam amount will decrease accordingly.

There are many studies about electricity production [1-8]. However, Elazığ Etikrom A.Ş. does not have any studies that classified by artificial neural network according to the values affecting electricity generation. The purpose of this article is to classification by artificial neural network according to the values affecting electricity generation of the Etikrom A.Ş. The novelty of this study, when studies in the literature are examined, it is seen that k-means and Artificial neural network are not used in together to classification Oven Power (MW), Boiler Inlet Gas Temperature, Amount of Steaming Steam, ID-Fan Speed, and Feeding Water Tank Datas taken from Etikrom A.Ş. Also, Analysis Oven Power (MW), Boiler Inlet Gas Temperature, Amount of Steaming Steam, ID-Fan Speed, and Feeding Water Tank Datas taken from Etikrom A.Ş have been done with ROC curve.

Theory and Method of the study are described in Section 2. Experiment results and conclusions are explained in Section 3 and Section 4, respectively.

Theory and Method

Obtaining the data

Electricity generation datas in this article have been taken from Etikrom A.Ş. The electricity generation datas are valid for the first four months of 2017. It consists of total and instant values for each hour. It consists of total and instant values for each hour. The k-means clustering method is defined as low electricity level between 0-2.33 MW, medium electricity level between 2.333-2.39 MW, and high electricity level between 2.391-3.39 MW.

In this article, the proposed system for classification by artificial neural network according to the values affecting electricity generation taken from Elazığ Etikrom A.Ş. is shown in figure 1.

In this article, according to the figure 2, Etikrom A.Ş. den Furnace Power, Boiler Inlet Gas Temperature, Superheated Steam Amount, ID-Fan Speed Speed, Feeding Water Tank, Electricity production values were taken.

Later, only electricity generation values were clustered by the k-means method. The values of kiln power, boiler inlet gas temperature, superheated steam, ID-Fan speed, feeding water tank values which are used to generate electricity by using cluster data as verification data are used as a classification feature. Artificial Neural Network was used as a classifier. Then, the classification of ROC curve was analyzed.

K-means clustering algorithm

The k-means algorithm uses it intuitively to find the center seeds for the k-median clusters. According to Arthur and Vassilvitskii [9,10], k-means improves the working time of the Lloyd algorithm and the quality of the final solution [9-13]. The k-means ++ algorithm chooses the seeds as follows, assuming that the number of clusters is k.

Step 1: Select a random observation of Z from the data a1 set. The selected observation is the first center, designated as.

Step 2: Calculate a1 distances from each observation. Let d( z b , a f ) MathType@MTEF@5@5@+=feaaguart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbnvMCYL2DLfgDOvMCaeXatLxBI9gBaerbd9wDYLwzYbItLDharuavP1wzZbItLDhis9wBH5garqqtubsr4rNCHbGeaGakY=3j0xXdbba91rFfpec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqaq=JfrVkFHe9pgea0dXdar=Jb9hs0dXdbPYxe9vr0=vr0=vqpWqaaeaabiGaciaacaqabeaadaabauaaaOqaaabaaaaaaaaapeGaamizaiaacIcacaWG6bWdamaaBaaaleaapeGaamOyaaWdaeqaaOWdbiaacYcacaWGHbWdamaaBaaaleaapeGaamOzaaWdaeqaaOWdbiaacMcaaaa@43D4@ be the distance between and observation b.

Step 3: Select the next centroid, a2 at random from Z with probability.

d 2 ( z be , a 1 ) f=1 n d 2 ( z f , a 1 ) MathType@MTEF@5@5@+=feaaguart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLnhiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrFfpeea0xe9Lq=Jc9vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=xfr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8qadaWcaaWdaeaapeGaamiza8aadaahaaWcbeqaa8qacaaIYaaaaOGaaiikaiaadQhapaWaaSbaaSqaa8qacaWGIbGaamyzaaWdaeqaaOWdbiaacYcacaWGHbWdamaaBaaaleaapeGaaGymaaWdaeqaaOWdbiaacMcaa8aabaWdbmaavadabeWcpaqaa8qacaWGMbGaeyypa0JaaGymaaWdaeaapeGaamOBaaqdpaqaa8qacqGHris5aaGccaWGKbWdamaaCaaaleqabaWdbiaaikdaaaGccaGGOaGaamOEa8aadaWgaaWcbaWdbiaadAgaa8aabeaak8qacaGGSaGaamyya8aadaWgaaWcbaWdbiaaigdaa8aabeaak8qacaGGPaaaaaaa@5056@

Step 4: To choose center f: Calculate the distance of each observation to each center and assign each observation to the nearest center. For b = 1,...,n and p = 1,...,f – 1, select centroid p at random from Z with probability

d 2 ( z b , a p ) {l; z l a p } d 2 ( z l , a p ) MathType@MTEF@5@5@+=feaaguart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLnhiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr4rNCHbGeaGqk0Jf9crFfpeea0xh9v8qiW7rqqrFfpeea0xe9Lq=Jc9vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=xfr=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@55F4@

Where Ap is the set of all observations closest to centroid ap and ab belongs to Ap.

That is, a probabilistic distance from each center must be selected that is proportional to the distance to the nearest selected center.

Step 5: Repeat step 4 until K centroids are selected. Using a simulation study of several cluster orientations, Arthur and Vassilvitskii [9,10] show that k-tools provide a faster convergence of cluster-centric distances from square points to the sum of a cluster set lower than Lloyd’s. algorithm [9-13].”

Experimental Results and Discussion

Electricity production amount to be produced in Elazığ Etikrom A.Ş. was estimated by using Furnace Power (MW) obtained from Etikrom A.Ş., Boiler Input Gas Temperature, Superheated Steam Quantity, ID-Fan Speed Rate, Feeding.

Water Tank data. Electricity generation amount is used as verification data. That is, by the k-means clustering method, the electricity generation amount is divided into 3 classes (low, medium, and high). 3621 data including Furnace Power (MW), Boiler Input Gas Temperature, Superheated Steam Amount, ID-Fan Speed, and Feeding Water Tank data were used after class 3 separation. With the k-means clustering method, 2742 of these data were clustered as low electricity, 296 as medium electricity and 583 as high electricity. This clustered data was given to the Artifical Neural Network classifier. The success rate obtained as a result of this classification is 85.81%. Classified data were analyzed by ROC curve.

In table 1 is seen examples of data affecting electricity generation. This values is determined as low-level classification with k-means method.

Conclusion

Predicting the amount of electricity produced in a power plant is very important for today’s economy. To date, there are many field work for classification or clustering [14-28]. Electricity generation datas in this article have been taken from Etikrom A.Ş. The electricity generation datas are valid for the first four months of 2017. The k-means clustering method is defined as low electricity level between 0-2.33 MW, medium electricity level between 2.333-2.39 MW, and high electricity level between 2.391-3.39 MW. Furnace Power (MW), Boiler Input Gas Temperature, Superheated Steam Amount, ID-Fan Speed, Feeding Water Tank data affect the electricity production. In this article, Etikrom A.Ş. The electricity production amount to be produced in Elazığ Etikrom A.Ş. was estimated by using the data of Furnace Power (MW), Water Inlet Gas Temperature, Steam Vapor Volume, ID-Fan Speed, Feeding Water Tank data. Electricity generation amount is used as verification data. That is, by the k-means clustering method, the electricity generation amount is divided into 3 classes (low, medium, and high). 3621 data including Furnace Power (MW), Boiler Input Gas Temperature, Superheated Steam Amount, ID-Fan Speed, and Feeding Water Tank data were used after class 3 separation. With the k-means clustering method, 2742 of these data were clustered as low electricity, 296 as medium electricity and 583 as high electricity. This clustered data was given to the Artificial Neural Network classifier. The success rate obtained as a result of this classification is 85.81%. Classified data were analyzed by ROC curve.

  1. Üstüntaş T, Şahin AD (2008) Wind turbine power curve estimation based on cluster center fuzzy logic modeling. Journal of Wind Engineering and Industrial Aerodynamics 96: 611-620. Link: https://goo.gl/8xnNus
  2. Agrawal S, Panigrahi BK, Tiwari MK (2008) Multiobjective particle swarm algorithm with fuzzy clustering for electrical power dispatch. IEEE Transactions on Evolutionary Computation 12: 529-541. Link: https://goo.gl/XXzvLA
  3. Grubb M, Butler L, Twomey P (2006) Diversity and security in UK electricity generation: The influence of low-carbon objectives. Energy Policy 34: 4050-4062. Link: https://goo.gl/8UC56a
  4. Aelterman P, Versichele M, Marzorati M, Boon N, Verstraete W (2008) Loading rate and external resistance control the electricity generation of microbial fuel cells with different three-dimensional anodes. Bioresource Technology 99: 8895-8902. Link: https://goo.gl/ZUpQQM
  5. Seabra JEA, Tao L, Chum HL, Macedo IC (2010) A techno-economic evaluation of the effects of centralized cellulosic ethanol and co-products refinery options with sugarcane mill clustering. Biomass and Bioenergy 34: 1065-1078. Link: https://goo.gl/aKor6f
  6. Dastrup SR, Zivin JG, Costa DL, Kahn ME (2012) Understanding the Solar Home price premium: Electricity generation and “Green” social status. European Economic Review 56: 961-973. Link: https://goo.gl/TwcHEU
  7. Tekiner H, Coit DW, Felder FA (2010) Multi-period multi-objective electricity generation expansion planning problem with Monte-Carlo simulation. Electric Power Systems Research 80: 1394-1405. Link: https://goo.gl/CDQrSW
  8. Marnay C, Venkataramanan G (2006) Microgrids in the evolving electricity generation and delivery infrastructure. In Power Engineering Society General Meeting IEEE. Link: https://goo.gl/Qe573w
  9. kmeans. Link: https://goo.gl/my1Rrv
  10. Arthur D, Vassilvitskii S (2007) K-means++: The Advantages of Careful Seeding. In: Proceedings of the Eighteenth Annual ACM-SIAM Symposium on Discrete Algorithms p. 1027–1035. Link: https://goo.gl/iom3qT
  11. Lloyd S (1982) Least Squares Quantization in PCM. IEEE Transactions on Information Theory 28: 129–137. Link: https://goo.gl/jy21tR 
  12. Seber GAF (1984) Multivariate Observations. Hoboken, NJ: John Wiley & Sons, Inc
  13. Spath H (1985) Cluster Dissection and Analysis: Theory, FORTRAN Programs, Examples. Translated by Goldschmidt J. New York: Halsted Press.
  14. Korkmaz SA, Poyraz M (2014) A New Method Based for Diagnosis of Breast Cancer Cells from Microscopic Images: DWEE—JHT. Journal of Medical Systems 38: 92. Link: https://goo.gl/2XsEHQ
  15. Korkmaz SA, Binol H (2017) Analysis of Molecular Structure Images by using ANN, RF, LBP, HOG, and Size Reduction Methods for early Stomach Cancer Detection. Journal of Molecular Structure 1156: 255-263. Link: https://goo.gl/Po4Dkq
  16. Korkmaz SA, Binol H, Akcicek A, Korkmaz MF (2017) A expert system for stomach cancer images with artificial neural network by using HOG features and linear discriminant analysis: HOG_LDA_ANN. Intelligent Systems and Informatics (SISY), IEEE 15th International Symposium on 327-332. Link: https://goo.gl/b1p5hh
  17. Turkoglu I (2007) Hardware implementation of varicap diode's ANN model using PIC microcontrollers. Sensors and Actuators A: Physical 138: 288-293. Link: https://goo.gl/a13cey
  18. Korkmaz SA (2018) LBP Özelliklerine Dayanan Lokasyon Koruyan Projeksiyon (LPP) Boyut Azaltma Metodunun Farklı Sınıflandırıcılar Üzerindeki Performanslarının Karşılaştırılması. Sakarya University Journal of Science 22: 1. Link: https://goo.gl/7ZYRNi
  19. Korkmaz SA, Poyraz M (2014) A New Method Based for Diagnosis of Breast Cancer Cells from Microscopic Images: DWEE--JHT. Journal of Medical Systems 38: 1. Link: https://goo.gl/B8jeFc
  20. Korkmaz SA, Poyraz M (2015) Least square support vector machine and minumum redundacy maximum relavance for diagnosis of breast cancer from breast microscopic images. Procedia-Social and Behavioral Sciences 174:  4026-4031. Link: https://goo.gl/EfrAV5
  21. Korkmaz SA, Korkmaz MF (2015) A new method based cancer detection in mammogram textures by finding feature weights and using Kullback–Leibler measure with kernel estimation. Optik-International Journal for Light and Electron Optics 126: 2576-2583. Link: https://goo.gl/JhDKCS
  22. Korkmaz SA, Korkmaz MF, Poyraz M (2015) Diagnosis of breast cancer in light microscopic and mammographic images textures using relative entropy via kernel estimation. Medical & biological engineering & computing 54: 561-573. Link: https://goo.gl/6WBXMs
  23. Korkmaz SA, Eren H (2013) Cancer detection in mammograms estimating feature weights via Kullback-Leibler measure. In: Image and Signal Processing (CISP). 6th International Congress on. IEEE 2:1035-1040. Link: https://goo.gl/ukfiFj
  24. Korkmaz SA (2017) Detecting cells using image segmentation of the cervical cancer images taken from scanning electron microscope. The Online Journal of Science and Technology 7: Link: https://goo.gl/WS6DqS
  25. Korkmaz SA, Akcicek A, Binol H, Korkmaz MF (2017) Recognition of the stomach cancer images with probabilistic HOG feature vector histograms by using HOG features. Intelligent Systems and Informatics (SISY), IEEE 15th International Symposium on 339-342. Link: https://goo.gl/2JzbW3
  26. Korkmaz SA, Korkmaz MF, Poyraz M, Yauphanoglu F (2016) Diagnosis of breast cancer nano-biomechanics images taken from atomic force microscope. Journal of Nanoelectronics and Optoelectronics 11: 551-559. Link: https://goo.gl/EZzDii
  27. Sengur A (2012) Support vector machine ensembles for intelligent diagnosis of valvular heart disease. Journal of medical systems 36: 2649-2655. Link: https://goo.gl/eNzFZH
  28. Şengür A (2008) An expert system based on principal component analysis, artificial immune system and fuzzy k-NN for diagnostic of valvular heart diseases. Computers in Biology and Medicine 38: 329 - 338. Link: https://goo.gl/Zt4mTu
© 2018 Esmeray F, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.