Random forest yöntemi kullanarak polimer elektrolit membran (PEM) yakıt hücrelerinin ömrünün belirlenmesi
Determining life span in polymer electrolyte membrane (PEM) fuel cell using random forest method
- Tez No: 900461
- Danışmanlar: PROF. DR. HANZADE AÇMA, PROF. DR. SERDAR YAMAN, DOÇ. DR. HALİT EREN FİGEN
- Tez Türü: Yüksek Lisans
- Konular: Kimya Mühendisliği, Chemical Engineering
- Anahtar Kelimeler: Belirtilmemiş.
- Yıl: 2024
- Dil: Türkçe
- Üniversite: İstanbul Teknik Üniversitesi
- Enstitü: Lisansüstü Eğitim Enstitüsü
- Ana Bilim Dalı: Kimya Mühendisliği Ana Bilim Dalı
- Bilim Dalı: Kimya Mühendisliği Bilim Dalı
- Sayfa Sayısı: 157
Özet
Günümüzde petrol kaynaklarımız hızla tükenmektedir. Ayrıca, otomobillerin egzozlarından yayılan gazların neden olduğu hava kirliliği ve petrol kullanımının beraberinde getirdiği diğer çevresel sorunlar da bulunmaktadır. Geleneksel içten yanmalı motorların yerine geçme vaadi sunan hidrojen yakıt hücreleri, temiz ve çevreyi kirletmeyen bir enerji kaynağı olmaları nedeniyle 21. yüzyılda iyi bir seçenek haline gelmiştir. Yakıt hücreleri, hidrojen (yakıt) ve oksijenin (hava) kimyasal enerjisini doğrudan elektrik ve ısı enerjisine dönüştüren güç elemanlarıdır. Bu reaksiyon sonucunda yan ürün olarak elektrik, su ve ısı açığa çıkar. Yakıtın enerjiye dönüşümü yanma ile değil, elektrokimyasal bir süreçle gerçekleşir. Bu süreç oldukça temiz, sessizdir ve yakıt yanmasından 2-3 kat daha verimlidir. Yakıt hücrelerinin avantajları arasında sadece su üretmeleri, yüksek enerji dönüşüm verimliliğine sahip olmaları ve umut verici bir süreç olmaları yer almaktadır. Yakıt hücreleriyle çalışan elektrikli araçların geliştirilmesi, karbon monoksit (CO) emisyonlarını azaltmanın en büyük nedenlerinden biridir. Yakıt hücrelerinin yeniden şarj edilmesine gerek yoktur. Diğer pil türlerinin aksine, yakıt hücreleri sadece sürekli bir hidrojen yakıtı tedariğine ihtiyaç duyar. Birçok yakıt hücresi türü bulunmaktadır. Bu çalışmada, başlıca hedef, maliyetli ve uzun süreli deneyler yapılmaksızın Polimer Elektrolit Membran (PEM) yakıt hücresinin ömrünü giriş parametreleri girerek tahmin etmektir. Hedef doğrultusunda Random Forest yöntemi seçilmiştir. Literatürden elde edilen deneysel verilerden bir set oluşturulmuş ve deneyin girdileri ile çıktıları yapay zekaya tanıtılmıştır. Sonuç olarak, herhangi bir veri setinde PEM yakıt hücresinin ömrü bir çıktı olarak tahmin edilebilir. Sonuç olarak, bu tahmin modelinden elde edilen sonuçlar, PEM yakıt hücrelerinin performansını ve dayanıklılığını artırmaya yönelik gelecekteki araştırma ve geliştirme çalışmalarına önemli katkılar sağlayabilir. Makine öğrenimi gibi ileri düzey hesaplama tekniklerinden yararlanarak, yakıt hücreleri gibi temiz enerji teknolojilerinin benimsenmesinde ilerlemeyi hızlandırabiliriz.
Özet (Çeviri)
Today, our petroleum resources are rapidly depleting. Additionally, there are environmental problems associated with the air pollution caused by the gases emitted from car exhaust and the use of petroleum. Hydrogen fuel cells, which promise to replace traditional internal combustion engines, have become a promising option in the 21st century due to their clean and environmentally friendly energy potential. Hydrogen is used in fuel cells. Like petroleum, hydrogen is not directly found on Earth but must be produced using other sources. Currently, a significant portion of hydrogen is produced through the steam methane reforming (SMR) process. Another method of hydrogen production is water electrolysis; if electricity generated from renewable energy technologies is used in this process, hydrogen can be produced entirely from renewable energy sources. The main issues with hydrogen production are: (i) it is an energy-intensive process, and (ii) to make a real environmental contribution, it must rely on renewable energy sources instead of the SMR process. There is broad consensus that producing hydrogen from renewable energy sources (solar, wind, etc.) holds great promise for the sustainable development of the world. Hydrogen is not directly available in nature but can be produced using other sources. Like any product, hydrogen must be packaged, transported, stored, and transferred to bring it from production to end use. The primary technological challenge for a functional hydrogen economy is storage, and finding a cost-effective way to store hydrogen remains a challenge. To be useful for transportation, hydrogen's energy density must be increased. However, solutions to the hydrogen storage problem are rapidly emerging. The scientific community and commercial organizations are working intensively on technologies with the highest hydrogen storage potential per unit volume. The main hydrogen storage technologies are compressed hydrogen gas storage, liquefied hydrogen storage, solid-state hydrogen storage, and chemical hydrogen storage. Compared to other gases used as fuel (natural gas, LPG), much smaller amounts of hydrogen can be stored using the same storage techniques. In other words, a larger volume is required to store the same amount of hydrogen. Hydrogen storage technologies are critical for hydrogen-powered energy systems. While traditional technologies store hydrogen as compressed gas and cryogenic liquid, underground storage is emerging as the preferred method for large-scale applications. The fuel cell was first developed by Sir William Grove (1811-1896), an English lawyer and amateur scientist. Fuel cells are power devices that convert the chemical energy of hydrogen (fuel) and oxygen (air) directly into electricity and heat. As a result of this reaction, electricity, water, and heat are released as by-products. The conversion of fuel into energy occurs not through combustion, but through an electrochemical process. This process is very clean, quiet, and 2-3 times more efficient than fuel combustion. The advantages of fuel cells include producing only water as a byproduct, having high energy conversion efficiency, and being a promising technology. The development of fuel cell-powered electric vehicles is one of the main reasons for reducing carbon monoxide (CO) emissions. Fuel cells do not need to be recharged. Unlike other types of batteries, fuel cells only require a continuous supply of hydrogen fuel. Various types of fuel cells have been thoroughly researched and studied to increase production potential and make them as accessible and usable by as many people as possible. The type of catalyst is very important as it affects the likelihood of undesirable side reactions. The catalyst must have high activity and selectivity over a wide temperature range. A fuel cell consists of three main components: two electrodes (anode and cathode), two catalyst layers, and one electrolyte (membrane). The two electrodes surround the electrolyte, and the catalyst layers are placed between the electrodes and the electrolyte. The process begins with the oxidation of a hydrogen molecule at the anode. Through oxidation, two electrons (2e- ) and two hydrogen ions (2H+ ) are released. These ions are separated using a polymer electrolyte membrane (PEM). This membrane allows hydrogen ions to pass directly to the cathode while blocking electrons, forcing them to pass through an external circuit. This flow of electrons generates the produced energy. Finally, when both hydrogen ions and electrons reach the cathode, they combine with oxygen and are reduced to water (H2O). A fuel cell is two to three times more efficient than a gasoline-powered internal combustion engine. There are many different types of fuel cells available for a wide range of applications. These applications include Mobility/Transportation, Power Systems, Energy Storage, Consumer Electronics, and Unmanned Aerial Vehicles. There are various fuel cell architectures, all of which generally consist of an anode and cathode electrode where reactions occur, electrons are conducted, and an electrolyte that provides ion conductivity, connecting the two electrodes. All types of fuel cells are based on the same electrochemical principles, but they use different materials, temperatures, and performance characteristics. Different fuel cell types are distinguished by the type of electrolyte material used. Fuel cells are divided into five types based on the electrolyte material used: Polymer Electrolyte Membrane Fuel Cell (PEMFC), Molten Carbonate Fuel Cell (MCFC), Solid Oxide Fuel Cell (SOFC), Alkaline Fuel Cell (AFC), and Phosphoric Acid Fuel Cell (PAFC). PEMFC is a scalable power source, meaning the same technology can be used in different applications. Due to its advantages, it is often preferred as a fuel source in mobile and stationary devices and the automotive industry. Additionally, its popularity is increasing in the aviation industry, and numerous projects aim to use PEMFCs in commercial passenger aircraft. In applications such as drones, where flight endurance and continuous environmental monitoring are required, lithium-ion batteries may not be beneficial. Due to their low energy densities, they cannot provide sufficient power for long flight durations. In contrast, PEMFCs can be used for longer flight durations due to their high energy density and fast refueling times. However, their current usage is not widespread due to commercial challenges. There are still many obstacles to making this technology commercially viable. The main ones include high-cost fuel cell systems, high maintenance costs, and short lifespans. The lifetime of a fuel cell operating under regular conditions can be very long, up to 26,300 hours; however, during accelerated durability tests, the voltage degradation rate is very high, especially during cold starts, where voltage degradation can reach 22.5 mV. Fuel cell lifespan requirements range from 4,000 hours for vehicle applications to 20,000 hours for buses and 40,000 hours for stationary applications. The longest known PEMFC test was a lifetime test reported by GE in 1979, which lasted 60,000 hours. Due to the absence of moving parts, fuel cells are naturally reliable systems. However, reactants, various materials including the catalyst, significant electrical potential and flux density, temperature, and pressure ranges cause material degradation under various operating conditions. Additionally, factors such as non-ideal water management, catalyst degradation, and fuel insufficiency can reduce the lifespan of a PEMFC and even lead to voltage drop. Managing the lifespan of the fuel cell stack depends on how these components and interacting conditions are designed and managed. In this study, the main goal is to predict the lifespan of PEM fuel cells by inputting parameters without conducting costly and long-term experiments. To achieve this goal, the Random Forest method was selected. The Random Forest (RF) algorithm, proposed by L. Breiman in 2001, has been extremely successful as a general-purpose classification and regression method. This approach, which combines several random decision trees and aggregates their predictions by averaging them, has shown excellent performance in situations where the number of variables is much larger than the number of observations. Additionally, it is flexible enough to be applied to large-scale problems, can easily be adapted to various specific learning tasks, and provides measures of variable importance. This supervised learning method is inspired by the early works of Amit and Geman (1997), Ho (1998), and Dietterich (2000) and operates based on a simple yet effective“divide and conquer”principle. It samples fractions of the data, grows a random tree predictor on each small part, and then aggregates and combines these predictors. What has contributed significantly to the popularity of forests is their applicability to a variety of prediction problems and their having few tuning parameters. For lifespan prediction, experimental data obtained from research was first used to create different datasets. The RF application was initially tested using four basic parameters. It is known that the number of parameters can vary depending on the type of target variable, and therefore, it is common to start with as many parameters as possible and then reduce them until an optimal prediction is achieved. Thus, the initial parameters were gradually reduced. The Python programming language was used for the Random Forest algorithm. To validate the progress of the algorithm, 80% of the data was used to train the model, and 20% was used to test the model. The training dataset needs to be larger than the test dataset to avoid overfitting. Predictions were evaluated by considering the R² and RMSE (Root Mean Square Error) values. Outliers were eliminated during the calculation process. As a result, the outcomes obtained from this prediction model can make significant contributions to future research and development efforts aimed at improving the performance and durability of PEM fuel cells. By utilizing advanced computational techniques such as machine learning, we can accelerate the adoption of clean energy technologies like fuel cells.
Benzer Tezler
- Effect of lignin, extractive matter, holocellulose, and alpha cellulose of biomass on calorific value
Biyokütlenin içeriğindeki lignin, ekstraktif madde, holoselüloz ve alfa selülozun kalorifik değer üzerindeki etkisi
ÖZLEM ECEM KAYNAR
Yüksek Lisans
İngilizce
2022
Kimya Mühendisliğiİstanbul Teknik ÜniversitesiKimya Mühendisliği Ana Bilim Dalı
PROF. DR. SERDAR YAMAN
- Wi-fi tabanlı parmak izi yöntemi kullanarak iç ortam konumlandırma
Indoor localization using wi-fi fingerprinting technique
IŞIL KARABEY
Yüksek Lisans
Türkçe
2015
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolAtatürk ÜniversitesiBilgisayar Mühendisliği Ana Bilim Dalı
YRD. DOÇ. DR. LEVENT BAYINDIR
- Intrusion detection system in internet of things networks using machine learning techniques
Nesnelerin internet ağlarında makine öğrenme teknikleri kullanarak saldırı tespit sistemi
MUHANAD BADEE MUHAMMED AL-DOORI
Yüksek Lisans
İngilizce
2023
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolAltınbaş ÜniversitesiBilgi Teknolojileri Ana Bilim Dalı
PROF. DR. SEFER KURNAZ
- Dengesiz bir diyabet veri setinde makine öğrenmesi yöntemlerini kullanarak diyabet hastalığının teşhisi
Diagnosis of diabetes disease using machine learning methods in an imbalanced diabetes dataset
İSMAİL BUĞRA BÖLÜKBAŞI
Yüksek Lisans
Türkçe
2023
Endüstri ve Endüstri MühendisliğiBursa Uludağ ÜniversitesiEndüstri Mühendisliği Ana Bilim Dalı
PROF. DR. BETÜL YAĞMAHAN
- İş ilanlarında doğal dil işleme ile duygu analizi
Sentiment analysis with natural language processing in job postings
ŞEYMA SARIGİL
Yüksek Lisans
Türkçe
2023
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolSelçuk ÜniversitesiBilgisayar Mühendisliği Ana Bilim Dalı
DR. ÖĞR. ÜYESİ MURAT KÖKLÜ