Search

Pri korištenju biomase kao sirovine za proizvodnju goriva i toplinske energije, potrebno je odrediti njezina energetska svojstva. Primjenom modela umjetnih neuronskih mreža moguće je s velikom točnošću predvidjeti donju ogrjevnu vrijednost pojedinih energetskih kultura, čime se olakšava pronalaženje optimalnog rješenja. Sposobnost samoorganizacije, optimiziranog učenja i visoke tolerancije na pogreške neuronskim mrežama omogućuje pouzdano i precizno izračunavanje donje ogrjevne vrijednosti biomase. Ispravnom konfiguracijom mreže i prikupljanjem podataka, modeli umjetnih neuronskih mreža mogu se koristiti za predviđanje donje ogrjevne vrijednosti u realnom vremenu.

Kontinuiranim povećanjem broja stanovnika, kao i poboljšanjem standarda dolazi i do povećanja proizvodnje otpada. Stoga ne čudi da sve veće količine otpada uzrokuju zabrinutost zbog ekonomske i ekološke održivosti trenutnog načina gospodarenja komunalnim otpadom. Gospodarenje biorazgradivim komunalnim otpadom od posebnog je značenja zbog njegovog potencijala u proizvodnji komposta ili kao sirovine u bioplinskim postrojenjima za proizvodnju energije i digestata. Izgradnjom bioplinskih postrojenja osigurali bi se uvjeti za provedbu odvojenog prikupljanja biorazgradivog otpada, povećala bi se proizvodnja energije iz obnovljivih izvora i posljedično smanjila količina odloženog otpada.

Kontinuiranim povećanjem broja stanovnika dolazi i do povećanja proizvodnje otpada. Isto tako, uz gospodarski rast i, povezano s time, poboljšavanja životnog standarda, također, dolazi do povećanja proizvodnje otpada. Stoga, ne čudi da sve veće količine otpada koje svake godine nastaju uzrokuju opravdanu zabrinutost zbog ekonomske održivosti i ekološke prihvatljivosti trenutnog načina gospodarenja otpadom. Glavni problem s kojim se suočava stručna i znanstvena javnost je kako predvidjeti količinu otpada koja će nastati u bliskoj budućnosti. Planiranje optimalne regionalne ili nacionalne strategije gospodarenja otpadom usko je povezano s količinom otpada koja će nastati. Za rješavanje navedenih problema pokazuje se potreba za kreiranjem pouzdanog modela za predviđanje količine nastalog otpada. Na temelju dosadašnjih istraživanja, umjetne neuronske mreže pokazuju bolje rezultate kod predviđanja nastanka otpada u usporedbi s drugim matematičkim modelima, stoga u ovom istraživanju koristit će se upravo umjetne neuronske mreže kao alat za razvoj matematičkog modela za predviđanje količina nastalog biorazgradivoga komunalnog otpada na europskoj i nacionalnoj razini. U ovom istraživanju poseban naglasak stavljen je na razvoj modela za predviđanje nastanka biorazgradivoga komunalnog otpada. Proučavanje biorazgradivoga komunalnog otpada od posebnog je interesa jer se upravo kod ove vrste otpada vidi veliki potencijal za njegovo relativno jednostavno i jeftino iskorištavanje, i to u vidu sirovine za dobivanje komposta pogodnog za daljnje korištenje u poljoprivredi ili u vidu ulazne sirovine u bioplinskim postrojenjima. Za kreiranje umjetne neuronske mreže u ovom doktorskom radu ulazne podatke činio je set sociodemografskih, ekonomskih i industrijskih podataka 17 država članica Europske unije za razdoblje od 25 godina. Kreiranim modelom u ovom doktorskom radu željele su se predvidjeti količine promatranih vrsta otpada koje će nastati na području 17 država Europske unije u razdoblju od 2020. do 2025. godine. Uz samo kreiranje mreže za predviđanje količina komponenti biorazgradivog otpada, cilj je istražiti i utjecaj sociodemografskih i ekonomskih pokazatelja na količine biorazgradivoga komunalnog otpada. Prema razvijenom modelu od 2020. do 2025. godine očekuje se da će u 17 država Europske unije nastati 411.351.769 tona miješanoga komunalnog otpada (u sklopu kojeg će nastati i 81.776.732 tona biootpada), 90.280.031 tona papira i kartona, 35.926.182 tona otpadnog drva i 3.511.589 tona tekstilnog otpada. Rezultati ovog istraživanja pokazuju kako na sve četiri promatrane vrste komunalnog otpada pozitivno utječu parametri kao što su broj stanovnika, bruto domaći proizvod po tržišnim cijenama, srednji ekvivalent neto prihoda, turizam, izvoz nafte i naftnih derivata i neto vanjski dug. S druge strane, životni vijek, realni BDP po stanovniku, ukupne obveze financijskog sektora i uvoz roba i usluga negativno utječu na sve četiri vrste otpada. Zaključno se može reći da iako je Europska unija heterogena zajednica i bez obzira na poteškoće u pronalasku što ažurnijih podataka o otpadu, kreiran model pokazao je zadovoljavajuća svojstva i mogućnosti u predviđanju količina miješanoga komunalnog otpada, otpadnog papira, drva i tekstila. Rezultati istraživanja mogu poslužiti kao pomoć pri uspostavi ekonomičnijeg i ekološki prihvatljivijeg načina gospodarenja biorazgradivim otpadom. The increasing amounts of waste generated each year raise legitimate concerns about the economic viability and environmental sustainability of the current way of waste management. The main problem facing professionals and academics is predicting the amounts of waste that will be generated in the near future. Planning an optimal regional or national waste management strategy is closely linked to the amount of waste that will be generated. To solve these problems, a reliable model for predicting the amount of waste needs to be developed. Such a tool should make it possible to select the most appropriate waste management technique. Based on previous research, artificial neural networks show better results in predicting waste generation compared to other mathematical models. Therefore, in this research, artificial neural networks are used as a tool to develop models for predicting the amount of biodegradable municipal waste at European and national level. This doctoral thesis is divided into 5 basic thematic units. It begins with an introduction in which the reader is briefly introduced to the main topics such as waste and artificial neural networks. It also defines the research area and the main objectives and hypotheses. For example, the introduction clearly explains that there are two main objectives of the research. The first objective is to develop a mathematical model for predicting the amounts quantities of components of biodegradable waste using artificial neural networks with the aim of applying it at European and national level. The second objective is to predict the impact of socio-demographic, economic and industrial indicators on the amount of biodegradable waste using artificial neural networks. The introduction is followed by a review of previous literature. In this chapter, the concepts of municipal waste, biodegradable waste and artificial neural networks are explained in more detail. Research conducted by other authors to date is also presented and explained in detail. They refer to the parameters that influence waste generation, as well as research that has dealt with the creation of mathematical models and their success in predicting waste generation. An additional significance of this research lies in the scale of the research. To date, artificial neural networks have been used to make predictions on a smaller local or regional scale. This usually covers the area of a particular city, state or group of closely related states. The aim of this research is to investigate the accuracy of a model with large-scale data (17 countries of the European Union). In the Materials and Methods chapter, the methodological approach and the way of creating a mathematical model are explained. To develop the model, demographic data (population, life expectancy, educational attainment), economic progress data (gross domestic product at market prices, gross domestic product per capita, total financial sector liabilities, net external debt, nominal effective exchange rate, direct investment in the reporting economy, house price index, data on the number of (non-)employed persons (total number of employed persons, unemployment rate, youth unemployment rate), tourism data (arrivals in tourist accommodation facilities, number of nights spent in tourist accommodation facilities), trade data (imports of goods and services, exports of goods and services, exports of oil and petroleum products) and waste data (annual municipal waste generation in thousands of tonnes, municipal waste generation per capita, municipal waste recycling rate, waste disposal) were collected. All data were collected for a period of 25 years for 17 countries of the European Union: Belgium, the Czech Republic, Denmark, Estonia, France, Croatia, Ireland, Italy, Latvia, Lithuania, Luxembourg, Hungary, Malta, the Netherlands, Slovenia, Spain and Sweden. In this research, the Multi-Layer Perceptron (MLP) model was used, which consists of a total of three layers: input, hidden layer and output. Before starting to compute the model, the database of collected data was divided into data for learning (60% of the data), for verification (20%) and for testing the neural network (20 %). Numerical verification of the obtained artificial neural network model was tested using the coefficient of determination (r2), reduced chi-squared (χ2), mean bias error (MBE), root mean square error (RMSE), mean percentage error (MPE), sum of squares error (SSE) and average absolute relative deviation (AARD). The constructed neural network model showed promising generalisation properties for the collected database and could be used to accurately predict waste generation: 20 networks, the maximum values of r2 (during the training cycle, r2 for the output variables (mixed municipal waste, municipal waste paper and cardboard, wood and textiles) were: 0.999, 0.998, 0.997 and 0.998). The results obtained show that artificial neural networks are indeed a reliable tool to create a mathematical model to predict the amount of biodegradable municipal waste at European and national level. The actual accuracy of the results of this research in terms of waste generation in the 17 countries observed will be verified when the data on waste generation for the year 2020 becomes available. In addition to creating a network to predict the quantities of biodegradable waste components, the influence of socio-demographic, economic and industrial indicators on the quantities of municipal biodegradable waste generated was also observed. Of the 28 input data, 10 input factors have a positive influence on all 4 observed waste types, while 4 input factors have a negative influence on all 4 observed waste types. Other observed factors (such as foreign direct investment, annual unemployment rate data, exports of goods and services and education) did not yield results from which a single conclusion could be drawn. From the above, it is clear that the accuracy of predicting the amount of biodegradable waste using artificial neural networks really depends on the choice of socio-demographic, economic and industrial indicators. For further studies to be carried out, it is proposed to use parameters such as population, gross domestic product at market prices, mean net income equivalent, tourism, exports of oil and petroleum products and net foreign debt, as these parameters positively influence all four types of municipal waste observed. On the other hand, life expectancy, real GDP per capita, total financial sector liabilities and imports of goods and services have a negative impact on all four types of waste. The results of this study are in line with the studies conducted so far, according to which the generation of municipal waste is mainly influenced by GDP, tourism, population and wages. Depending on changes in these factors, the amount of waste generated also changes. The model created can help the waste management system behave like a "living organism". In this flexible way, the waste management system could change in parallel with social and economic changes. This would make municipal waste management more efficient and economical, with less impact on the environment.