Detailed analysis of Türkiye's agricultural biomass-based energy potential with machine learning algorithms based on environmental and climatic conditions

Energy is needed at every stage of modern life. Energy consumption is an essential parameter in determining a country's development level. Energy consumption in developed countries is higher than in developing countries (Khanlari et al. 2020; Can 2022; Ozturk et al. 2017). With the development of technology, the energy demand of developing countries is increasing rapidly to meet the needs. Energy use is both a cause and a consequence of economic growth and development. Energy is essential for most economic activities. Fossil fuels are the primary source of energy for worldwide energy needs (Sayin et al. 2005; Jayarathna et al. 2020). These energy sources' production, transportation, and use cause high emissions, damage to the atmosphere, and climate change now and in the future (Pence et al. 2023; Kaygusuz 2010). Climate change has become an essential global problem related to energy, the economy, the environment, and technology. Significant steps must be taken to reduce global greenhouse gas (GHG) emissions (Barbera et al. 2019; D'Adamo et al. 2019; Zheng et al. 2019). For this reason, it is necessary to determine greenhouse gas emission management in countries subject to international commitments, agreements, and national policies. In this context, it is aimed to increase the share of renewable energy resources with some agreements and protocols to prevent emissions worldwide (Zheng et al. 2019; Chang and Hu 2019). In the 2015 Paris Agreement, which Türkiye also signed, it was decided to reduce CO₂ emissions by 45% compared to 2010 and zero them by 2050.

The European Union (EU) aims to increase the share of renewable energy to at least 32% by 2030, reduce greenhouse gas emissions by 55%, and increase energy efficiency by at least 32.5% (Guler et al. 2022; Filipović et al. 2022). Depending on this decision, reducing emissions caused by energy consumption, which is one of the essential parameters in emissions, is crucial. The economy of Türkiye and industry organization in this direction will strengthen the country's adaptation to the European Green Deal (Tanasa et al. 2020; Cekinir et al. 2022). Increasing renewable energy sources in all sectors is vital to reducing greenhouse gas emissions, especially in industry. Increasing the use of energy resources such as hydroelectric, wind, solar, and biomass will be beneficial in preventing emissions by taking steps toward energy independence and security (Tumen Ozdil and Caliskan 2022).

The largest share of electricity in Türkiye is obtained from thermal power plants that consume natural gas, oil, and imported coal. Although lignite and coal have an essential potential, oil and natural gas reserves are almost negligible compared to world reserves (Erat et al. 2021; Telli et al. 2021). While production increased by 45.21% between 2010 and 2020, it increased by 17.20% between 2015 and 2020. Türkiye's electricity production in 2020 is 306,703 GWh. The use rate of renewable energy in electricity generation is approximately a 3.7% annual average increase from 2012 to 2020, 7.8% in hydraulics, 20.50% in wind, 36% in geothermal, and 31% in biomass (Pence et al. 2023; Ocak and Acar 2021). On the other hand, according to Türkiye's renewable energy targets, renewable energy sources are expected to meet 20% of total energy consumption and 30% of electricity production by 2023 (Pence et al. 2023; Rincon et al. 2019).

Türkiye's electricity demand is expected to reach 424 TWh in 2023, and it is predicted that the demand will be approximately five times higher in 2030 compared to 2000 (Melikoglu and Menekse 2020; Yurtkuran 2021). Türkiye has great potential in renewable energy. Türkiye's geographical location has several advantages for the widespread use of most renewable energy sources. Due to these advantages, it is a suitable country for energy production from renewable sources such as wind, solar, hydroelectricity, and biomass. Hydroelectric power plants come to the fore in Türkiye among these energy sources. As of the middle of 2022, there are 750 hydroelectric power plants in Türkiye (Şenol et al. 2030; Bakay and Ağbulut 2021). Türkiye's total installed power as of the end of 2023 is 106,071 MW. About 55.3% of the power plants in operation are power plants that produce electricity from renewable sources. Türkiye's total electricity installed power consists of thermal power plants (44.7%), hydroelectric power plants (29.8%), wind (11%), solar energy (10.6%), biomass (2.3%), and geothermal (1.6%) (TETC 2024).

Biomass is one of the important renewable energy sources with many uses worldwide. Municipal waste, vegetable oil waste, agricultural waste, forest product waste, industrial waste sludge, and sewage sludge are sources of biomass (Guler et al. 2022). This energy source has many advantages over other renewable energy sources. Some of these advantages are that it is easily obtained from organic materials, easy to obtain energy, and has many uses, such as home and industrial sectors that concern a large part of society (Ozturk et al. 2017; Channi et al. 2022; Jayarathna et al. 2022). Biomass can be burned directly or converted into solid, gaseous, and liquid fuels with the help of conversion technologies such as fermentation to produce alcohol, anaerobic digestion to produce biogas, and gasification to produce natural gas substitutes (Samadi et al. 2020). From an economic and technical point of view, it is one of the most sustainable energy sources for the country. This resource is available in a stored form and can increase employment opportunities in rural areas. It can help to reduce the trade deficit by reducing the dependence of developing countries like Türkiye, which imports a large part of its energy needs, on energy imports (Knápek et al. 2020). Due to these advantages, biomass stands out compared to other renewable energy sources, and it is accepted that biomass is more efficient in terms of economic and technical feasibility (Cekinir et al. 2022; Asghar et al. 2022). Biomass energy can be converted into thermal and electrical energy. For these reasons, energy policies and plans that ensure food safety worldwide should be supported in using biomass for energy purposes locally, nationally, and globally (Toklu 2017; Singh 2016).

Aiming to use bioenergy in Türkiye, at the First Agricultural Congress in 1931, it was discussed that the fuels needed for agricultural machinery would be produced with domestic resources instead of imports. It was emphasized that the fuels obtained using local resources benefit the national economy, and the importance of biofuel production was discussed in various dimensions. The first official document regarding biofuels in Türkiye was signed in 1934. Biogas production, one of the bioenergy sources, was initiated by the Soil and Water Research Institute in the 1950s. In the 1960s, pilot facilities were established within the State Production Farms, and eight biogas facilities were established by the Eskişehir Soil Water Research Institute affiliated with the Ministry of Agriculture. However, the work has ended due to a lack of technical personnel and inadequate training of farmers. Especially with the oil crisis in the early 1980s, studies on establishing biogas units increased. After 1980, studies on biogas production in Türkiye gained momentum, and the work that started with establishing a 35 m³ facility in Muş Province expanded to establishing approximately 1000 facilities with state support. Studies in this field in Türkiye continued increasingly after 2000. By the end of 2023, there are a total of 212 bioenergy facilities in Türkiye, 199 of which produce electrical energy from bioenergy, eight of which produce biodiesel, and five of which produce bioethanol (RTMENR 2024; Kumaş et al. 2019; Hatunoğlu 2010).

Biomass residues are directly linked to crop yield at all stages of agricultural production. Excess product production results in excess residues since residues constitute a certain percentage of the crop. During crop and fruit production in agriculture, many residues are obtained from field and horticultural crops. Biomass residues are the stems, straws, stems, leaves, branches, etc., after harvesting the main crop in agriculture, which are the remnants left by cutting and pruning (Tumen Ozdil and Caliskan 2022; Güney and Kantar 2020; Avcıoğlu et al. 2019). Biomass energy potential can be calculated based on these parameters. The potential also depends on environmental factors such as crop yield and biomass residues and their agronomic development, climatic conditions, and soil structure. For this reason, even if the same amount of agricultural product is obtained, different amounts of agricultural product residues used as biomass are obtained in different countries (Zafar et al. 2021; Balsalobre-Lorente et al. 2019; Aydin 2019). Türkiye shows different characteristics depending on its geographical location and landforms. Therefore, it has a wide variety of agricultural products (Senocak and Guner 2022). Different indicators such as climatic data, harvest time, and geographical conditions affect the amount of agricultural products. In recent years, agricultural areas have also been affected by the instantaneous effect of meteorological events, which causes climate change (Avcıoğlu et al. 2019; Zheng and Qiu 2020). In the agricultural sector, the climatic characteristics of that region are primarily considered when making construction and operation plans and production plans.

Climate and meteorological data are the absolute guiding factors in selecting the plant variety to be grown, tillage, planting, pruning, hoeing, irrigation, fertilization, spraying, harvesting, and micro-climatological environment planning. Climatic factors (temperature, relative humidity, dew, fog, precipitation, cloudiness, light, wind, snow, and frost) affect agricultural activities and cause severe problems if various applications are not made according to climatic factors (RTMAF 2022).

In recent years, studies have been carried out in Türkiye on using biomass resources such as hazelnut shells, agricultural waste, wheat straw, tea waste, and olive peel for energy purposes. It is imperative to focus on efficient production to meet the increasing energy demand and use of biomass energy to meet both traditional and modern fuel requirements (Yurtkuran 2021; Balat 2005). Studies have been carried out on the potential of biomass resources in the world and Türkiye. These studies, which investigated different results for calculating agricultural biomass residues and potentials, are as follows.

Singh (2016) developed a method for estimating India's biomass power potential from agricultural waste. It has been determined that 650 Mt of agricultural biomass is obtained annually in the country, and 1/3 of it can be used. It is stated that the equivalent of this biomass has an energy potential of 3.72 EJ, which is approximately equal to 23–35 GW of electrical power (Singh 2016).

On the other hand, Ozturk et al. (2017) compared the renewable energy and biomass potential by showing the future energy scenarios of Türkiye and Malaysia. By evaluating the biomass presence of the countries, their possible contributions to the economy of both countries were evaluated (Ozturk et al. 2017). Toklu (2017) examined Türkiye's biomass potential for different sources in 2010. The study stated that the total biomass energy potential of Türkiye is approximately 33 Mtoe, while the usable biomass potential is approximately 17 Mtoe (Toklu 2017). Bilandzija et al. (2018) examined Croatia's agricultural waste's biomass and energy potential. Biomass potentials of 3050.3 t, 1441.8 t, and 733.68 t were determined for different scenarios. The study calculated that 51.14 PJ, 24.06 PJ, and 12.18 PJ energy potentials could be obtained depending on this potential (Bilandzija et al. 2018). Ma et al. (2018) used the artificial neural network method to model and predict the production and consumption values of biomass and hydroelectric energy resources in the USA. In the study, data between 2009 and 2016, as well as LSTM and RNN estimation algorithms, were used (Ma et al. 2018). Avcıoğlu et al. (2019) investigated agricultural wastes' biomass and energy potential in Türkiye. The total amount of biomass obtainable from field and garden plant wastes was calculated as 9432 kt and 15,652 kt, respectively. The theoretical energy potential based on the amount of biomass was estimated as 908,119 Terajoule (TJ) from field crops and 90,354 TJ from horticultural crops, respectively (Avcıoğlu et al. 2019). Moustakas et al. (2020) investigated the biomass potential of the Thessaly Region, Greece. In the study, the biomass potential from agricultural waste is approximately 707,164 tons/year, as per the data. It has been stated that with the potential of biomass, a maximum of 619 GWh of electricity and 895 GWh of heat can be obtained per year (Moustakas et al. 2020). Knápek et al. (2020) examined the biomass potential of the Czech Republic for different scenarios. The GIS (geographic information system) method was used to determine the potential, and it was stated that the biomass potential would increase by 35% in energy production with the planning of the arable land. Considering the year 2040, it was stated that the biomass potential could increase by 42 PJ in total (Knápek et al. 2020). Samadi et al. (2020) estimated Iran's energy production by gasification technology of agricultural waste. It is stated that the total energy obtained from agricultural residues by gasification is 341.29 TJ, and the amount of electricity and heat is 66,075 and 399,112 TJ, respectively (Samadi et al. 2020). Tumen Ozdil and Caliskan (2022) determined the biomass potential and the associated reproducible electrical energy from agricultural wastes in Türkiye between 2008 and 2018. The total theoretical average biomass potential obtained from its plants was calculated as 522,875 kt and 51,359 kt, respectively. The potential of electricity produced from agricultural field crops is 994 × 10⁹ kWh (Tumen Ozdil and Caliskan 2022). Senocak and Guner (2022) estimated the amount of animal and agricultural waste expected for the coming years and the energy potential for the Acıpayam district of Denizli, Türkiye, using artificial intelligence. In addition, spatial analyses and different scenario evaluations were carried out using GIS (Senocak and Guner 2022).

Due to climate and land conditions, Türkiye has an essential geographical position regarding energy crop cultivation. Although it has a high potential in terms of biomass resources potential and the energy that can be obtained from these resources, the desired levels have not been reached in terms of installed power in the country (Pence et al. 2023; Şenol et al. 2030; Avcıoğlu et al. 2019). Determining the amount and distribution of biomass resources as accurately as possible is crucial in making strategic decisions such as energy management policies (Tumen Ozdil and Caliskan 2022; Toklu 2017; Avcıoğlu et al. 2019). According to the literature studies, generally known basic statistical approaches were used to estimate the renewable energy potential in Türkiye. Adopting a variety of scenario approaches that also take into account uncertain factors can lead to better results. In addition, reliable, analytical, and flexible estimation methods are needed to determine the energy potential of biomass. By adopting such a systematic, integrated approach, energy management decisions can be made more effectively. Unlike the literature, this study proposes a decision support method that enables estimating the energy potential of the resources produced from biomass.

A machine learning algorithm is a subset of artificial intelligence that learns from data to improve performance on a given task without being explicitly programmed. A machine can learn from experience and improve its performance over time by recognizing patterns and making decisions based on them. Machine learning can be classified into supervised, unsupervised, semi-supervised, and reinforcement learning. In supervised learning for regression, a model is trained to predict a continuous numerical output based on a set of input features. In supervised learning, the model is provided with training examples that include both the input features and the corresponding target values, and the goal is to build a mapping function that can accurately predict the target value based on new, unseen input values (Goodfellow et al. 2016).

Image and speech recognition, natural language processing, recommendation systems, and predictive modeling are just a few of the many uses for machine learning algorithms. They are incredibly well suited to manual tasks that are too difficult or time-consuming for humans to complete. Machine learning can be used to develop models that can accurately predict the energy equivalent of the theoretically calculated biomass potential using the waste rate, production amount, average moisture of each crop, sub-calorific value, and percentage availability of each field and orchard crop. These machine learning models can be trained on large datasets of various climate data, agricultural land, and the corresponding biomass potential data, allowing them to learn the relationships between them. The models can save time and resources for direct measurements or calculations by being trained to predict biomass potential in new areas or at various times.

Türkiye is geographically located between the European and Asian continents. Due to its location, the country has climate transitions, different land structures, and diverse agricultural products. No study has been found in the literature that includes long-term and up-to-date data for countries with such characteristics. When the literature is examined, it is noted that there are few studies on agricultural biomass. Generally, only theoretical biomass calculations have been made in the literature, and other factors affecting it have not been examined. This study tried to close this gap in the literature by incorporating climate data, the effects of which have not been examined before, into machine learning models for agricultural biomass potential estimation.

Regression models are used to predict the value of a dependent variable. It is essential to use these algorithms to predict the value of a dependent variable and explain the relationship between variables. It was assumed that the amount of agricultural biomass potential depends on independent variables such as climate data, and modeling with regression algorithms was preferred for this purpose. Among the regression models, Random Forest, K-Nearest Neighbors (KNN), and Gradient Boosting algorithms, known as state-of-the-art in the literature, are widely used. The eXtreme Gradient Boosting Regressor (XGBR), which has become popular with its successful results in recent years, also stands out.

In this study, the agricultural biomass potential of each province in Türkiye for the years 2010–2021 was modeled using various climate data and agricultural land. Random Forest, KNN, Gradient Boosting, and XGBR, popular machine learning algorithms in the literature, were used for modeling.

The novelty and contribution are: (1) Agricultural biomass potential has been modeled with high success by machine learning methods using climate data; (2) with the feature selection, the prediction success was increased by using only the agricultural area, temperature, biomass type, and humidity; and (3) considering many years in terms of agricultural biomass, a model has been established for Türkiye.

This study consists of four parts. The first part examined literature studies in the world and Türkiye. In the second part, the mathematical calculation of the theoretical biomass potential and the methods to be used in estimation have been explained in detail. The findings obtained from the mathematical method and estimation were compared in the third part. Finally, in the fourth part, the results and recommendations were given.

Türkiye is geographically located between 36° and 42° north latitudes and 26°–45° east meridians. It is located at an important point connecting the continents of Europe and Asia like a bridge. The surface area of Türkiye is known as 814,578 km². Türkiye's population reached 84.5 million in 2022 (Pence et al. 2023). Mediterranean, Black Sea, and continental climate types are seen in Türkiye. The summer months are hot and dry, and the winter months are mild and rainy. Mediterranean climate is seen on the Aegean and Mediterranean coasts. In the Black Sea climate, it is rainy in all seasons. The continental climate is observed in the inner parts of Türkiye. Wheat, barley, and corn are the most grown products in Türkiye. In addition, products such as cotton, flax, sesame, and poppy, which have high economic returns, have been grown for a long time. Soybeans are grown in the Mediterranean region. A wide variety of fruits are grown in many parts of Türkiye. Therefore, various agricultural residues are a source of biomass energy (Can 2022; Tumen Ozdil and Caliskan 2022; Şenol et al. 2030). In Türkiye, in the second half of 2022, cereal and other herbal products will increase by approximately 14% compared to 2021, while fruits, beverages, and spice plants will increase by 3%. Production amounts 2022 were approximately 70 million tons in cereals and other herbal products, 32 million tons in vegetables, and 26 million tons in fruits, beverages, and spice plants. Cereal production volumes increased by 21% in 2022 compared to 2021. When the past 2 years are compared, wheat production increased by 12% to 19.8 million tons, and corn production increased by 23% to 8.3 million tons. In addition, barley production increased by 48% to 8.5 million tons, rye production increased by 37% to 273 thousand tons, and oat production increased by 32% to 365 thousand tons. The production of fruits, beverages, and spice plants increased by 3.8% in 2022 compared to the previous year and became approximately 25.8 million tons. Compared to 2021, fruit products increased by 5% in apples, 13% in grapes, 12% in peaches and nectarines, 8% in plums, 8% in strawberries, and 71% in olives. However, there was a decrease of 17% in tangerines, 31% in oranges, and 32% in lemons. There was an increase of 11.8% in hazelnuts, 99% in pistachios, 9% in figs, and 13% in bananas. Bean production decreased by 11.5% to 270 thousand tons, soybean production decreased by 15% to 155 thousand tons, and sunflower production increased by 5% to approximately 2.6 million tons. Tobacco production increased by 15% to 82 thousand tons, and sugar beet production increased by 7% to 19 million tons (TUIK 2022).

In this study, the agricultural biomass potential of each province in Türkiye for the years 2010–2021 was modeled using machine learning algorithms. First, a tenfold cross-validation analysis was performed on the dataset for 2010–2020, and then, the prediction for 2021 was made. For the training of machine learning algorithms, analyses were carried out using all features (the type of biomass, year, agricultural area, temperature, maximum temperature, minimum temperature, humidity, wind speed, precipitation, soil temperature at 5 cm, and sunshine duration) and only four features (the type of biomass, agricultural area, temperature, and humidity) determined during the feature selection stage. Python programming language (Python 3.8) implements machine learning algorithms and data analysis. As one of the machine learning libraries, the Scikit-learn library was preferred because it has a rich environment and is widely used in the data science community. Different values have been tried for hyperparameters for machine learning algorithms: the minimum sample on the leaf for Random Forest and the number of neighbors for KNN. All other parameters are selected as default settings in the relevant libraries.

Regarding the results obtained, the original data and KNN prediction values for 2021 were obtained as a total of 308,888 TJ and 327,122 TJ for field products and a total of 77,002 TJ and 74,395 TJ for garden products, respectively.

For field products, the minimum value of the original data for 2021 was 29 TJ, and the KNN estimate was 39 TJ, while the maximum value was 39,223 TJ in the original data and 39,363 TJ in the prediction.

For garden products, the minimum value of the original data for 2021 was 0.8 TJ, and the KNN estimate was 0.5 TJ, while the maximum value was 8079 TJ in the original data and 6908 TJ in the prediction.

In addition, regarding average values, the original values for 2021 field products are 3813 TJ, the KNN prediction is 4038 TJ, and the margin of error is 5.9%. The average original value for garden products was 950 TJ, the KNN prediction was 918 TJ, and the margin of error was 3.3%. Accordingly, it can be seen that the difference between the predicted values and the original values is slight.

Like this study, Avcıoğlu et al. (2019) obtained the agricultural biomass potential for Türkiye as 298,955 TJ for total field crops and 65,491 TJ for horticultural crops for 2015 (Avcıoğlu et al. 2019). When the original 2021 data calculated in this study was compared with the 2015 data, it was seen that there was an increase of 3.3% for field crops and 17.5% for garden crops. Senocak and Guner Goren (2022) used 16 years of data from agricultural and animal biomass resources for a small selected region in Türkiye and estimated the amount of energy using SVR-GIS. The theoretical total energy in the 3 years was estimated as 27,088.095, 27,165.993, and 26,862.373 TEP/year, respectively (Senocak and Guner 2022). Tumen Ozdil and Caliskan (2022) calculated the theoretical energy potential of 2,825,932 × 10¹² J from field crops and 752,031 × 10¹² J from garden plants, using the biomass potential in Türkiye between 2008 and 2018 (Tumen Ozdil and Caliskan 2022).

Türkiye is a wealthy country in terms of agricultural diversity due to its geographical location and climatic conditions. Along with agricultural diversity, crop residues originating from fields and gardens, which can be a significant energy source, are formed. Agricultural residues are generally used as animal feed in Türkiye or directly incinerated for primary thermal heating. The agricultural sector in Türkiye has developed over the years using modern agricultural methods, and its importance has increased in recent years as the surrounding countries are at war. Many different support mechanisms are established for farmers in Türkiye to develop the sector. Some of them include seed support, fuel support, fertilizer support, energy support, tractor support, appropriate agricultural credit support, and irrigation support. With the sector's development, the amount of agricultural products has increased every year, so the amount of product residue, which is a source of biomass, has also increased. Agricultural residue management is critical in biomass energy production and use in terms of its features, such as an alternative energy source and environmental friendliness. This study calculated the energy production potential obtained from agricultural products for Türkiye between 2010 and 2021. The total theoretical energy potential obtained from field and garden products has been calculated as 222,620 TJ and 61,737 TJ for 2010, respectively, and 308,888 TJ and 77,002 TJ for 2021.

This study determined the most successful features as agricultural area, temperature, biomass type, and humidity with the SGD method. Different models and hyperparameters were tested with cross-validation analysis using four features for 2010–2020, and the best model for agricultural biomass potential was determined. In addition, the biomass potential prediction for 2021 was performed using 11 years of data. Based on the environmental and climate data covering 2010–2020, the agricultural biomass potential for 2021 was estimated using machine learning algorithms. When the results of the tenfold cross-validation and predictions for 2021 were examined, it was concluded that the Random Forest algorithm (leaf = 1) was successful and stable. Accordingly, the Random Forest algorithm (leaf = 1) obtained 0.9211 and 0.9434 R² values for tenfold cross-validation and 2021 prediction using four features (agricultural area, temperature, biomass type, and humidity), respectively. In addition, when four features are used, the KNN algorithm (k = 2) has achieved an R² value of 0.9560 only in the prediction of 2021. When all features are used, the Random Forest algorithm (leaf = 1) reached 0.9328 R² in tenfold cross-validation analysis, while the XGBR method reached 0.9495 R² in 2021 predictions.

The successful results have shown that machine learning methods can be used to model agricultural biomass potential. By using environmental and climatic values instead of too many variables from too many categories that need to be known for the theoretical calculation, it is provided to perform the calculation much more comfortably. In addition, thanks to the feature selection, these variables were reduced to four, making the calculation even easier with a high estimation rate.

With the proposed model, web and mobile applications that can estimate biomass potential can be created using agricultural field information and meteorological data published online. In this way, decision mechanisms can be developed regarding industrial production, updating the installed capacity of the biomass power plants in Türkiye and which regions they should focus on for agricultural waste to be collected. Such applications can also contribute to Türkiye's renewable energy policies and achieve its zero-emission target.

The model developed in this study was created only according to Türkiye's climatic parameters and estimated the agricultural biomass potential. The model can be recreated with data and parameters for another country or other biomass resources.

This study offers a detailed perspective on improving energy management practices in agriculture. Additionally, it is thought that machine learning algorithms will contribute to the literature for institutions and researchers that promote sustainable development in the sector. The efficient use of renewable energy resources in agriculture and artificial intelligence-supported energy efficiency management practices can help achieve sustainable agriculture goals.