ACTIVATED CARBON FROM WINEMAKING WASTE: THERMOECONOMIC ANALYSIS FOR LARGE-SCALE PRODUCTION

An activated carbon manufacturing process using winemaking waste is analyzed and designed at industrial scale. Starting from experimental research, the chemical transformations and thermodynamics during pruning wood conversion are studied as a basis for plant design. In this way, mass and energy balances of hydrothermal carbonization and physical activation are fulfilled and a thermoeconomic methodology is applied to develop an energy-integrated plant. To achieve this target, a network of heat exchangers is allocated to minimize heat consumption and supply hot domestic water, while a cogeneration cycle is designed to provide electricity and satisfy the remaining heat demand. Furthermore, a sensitivity analysis is carried out to determine the influence of the production scale and other operation parameters, such as annual workload, service life, and capital and feedstock costs, on the economic viability of the plant. The energy balance of the plant indicates that the energy integration design manages to provide 48.9% of the overall process energy demand by crossing hot and cold streams and recovering heat from residual flue gas. On the other hand, the exergy cost analysis identifies the combustion of pruning wood used to provide heat demands as the main source of exergy destruction, confirming the suitability of integration to improve the thermodynamic performance. Including activated carbon production, electricity, and hot domestic water, the exergy efficiency of the plant stands at 11.5%.


Introduction
Concepts related to sustainable development stand out among the most influential topics at present, such as replacement of non-renewable resources, reduction of wastes and greenhouse emissions, and the improvement of energy efficiency. In this regard, the European Union launched its political strategy in 2015 to impulse its socioeconomic transition to a new model through the document "Closing the loop: An EU action plan for the Circular Economy" [1]. The plan includes several investments to encourage innovative and efficient companies and is considered by the Brussels authorities to provide the opportunity to reinvent the European economy and keep it in vanguard [2]. Looking at its economic activity, the agri-food industry appears as a chance to research new processes to make consumer goods from its wastes. The wine industry, which generates an abundant amount of organic wastes, such as vine shoots and grape marc and seeds, is an appropriate example. The Spanish wine sector has a strong presence in global production, with an process, including a cogeneration system fed by the same wood residues used raw, is proposed and analyzed, in order to foster energy self-sustainable production. Finally, a sensitivity analysis is applied to determine the plant profitability, based on different production and economic scenarios.

Experimental data
Vineyard pruning wood was received from the Biological Mission of Galicia (CSIC, northwest Spain) and crushed to a grain size of <2 mm using an SK 100 Cross Beater Mill to increase the surface area and reactivity. Then, the pruning wood was mixed with water (74 grams of solid per liter of water) and introduced into a Berghof BR-100 reactor. The carbonization pressure was set at 50 bar, while different times (1 to 7 hours) and temperatures (150 to 250 ºC) were tested. Supplementary Fig.  S1 shows the solid yield of HTC of vineyard pruning wood under the different mentioned experimental conditions, as well as the variation of elemental composition after 1 hour of treatment. The increase of temperature showed a stronger influence, leading to lower solid yields but achieving hydrochar with higher carbon contents; a characteristic that is desirable for the subsequent production of activated carbon. Thus, hydrothermal cracking was established at 250 ºC for 1 hour. Under these conditions, the reaction mass balance is shown in Supplementary Table S1.
Total organic carbon (TOC) of the liquid phase was determined using a combustion/non-dispersive infrared gas analyzer (model TOC-V Shimadzu). Afterwards, the liquid phase was separated into two fractions by evaporation in a rotary evaporator (Buchi Mod. R300), in order to determine its chemical composition.
The condensable phase consisted of an azeotropic mixture of water and minor contents of acetic acid, which were estimated through acid-base titration using a 1 M NaOH solution. On the other hand, the composition of the organic non-condensable phase was determined by gas chromatography-mass spectrometry (GC-MS). Analyzed samples of the non-condensable phase were diluted in dichloromethane at concentrations between 40 and 42 mg/ml and were injected immediately into the chromatograph. This analysis was done with an Agilent 7890A equipped with an Agilent MS5975C mass spectrometer and a 30 m long HP-5ms (5% phenyl-methylpolysiloxane) column. The chemical composition and elemental C, H, and O proportions of this organic phase are noted in Supplementary Tables S2 and S3. In order to make analysis of the process viable, the liquid phase composition was simplified by selecting chemical compounds among those detected in the GC-MS, based on the bibliographic disposability of thermodynamic data [42,43]. Their mass percentages were recalculated to maintain real carbon, hydrogen, and oxygen proportions. Gaseous products could not be determined experimentally, and their compositions were estimated through elemental C, H, and O mass balances, considering other biomass HTC studies published in the literature. As mentioned in the introduction, it is reasonable to expect a gaseous phase composed fundamentally of CO2 and minor contents of CO, CH4, and H2 [11], [16], [32,33]. The formation of other gaseous products was disregarded. Tables S4 and S5 note the estimated mass balance for the vineyard pruning wood hydrothermal carbonization.
Once HTC was completed, 4 g of hydrochar was physically activated in a rotary kiln by using a steam flow of 0.5 ml/min at 900 ºC for 1 h. During activation, an important carbon fraction is removed as CO, giving place to a microporous network (Fig. S2a) that increases the adsorption capacity [44]. Due to high temperatures, a slight coalification of the char was still generated and other gaseous products, such as CO2, H2, and H2O may have formed [45]. Activation burn-off gives a first approach to activation degree and further carbonization (Fig. S2b). Its value was calculated through Eq. 1 and reached 38.88%: where w1 and w2 are the hydrochar and activated carbon masses, respectively. Activation gas composition was estimated through elemental mass balances, as done with the HTC gas phase. In this regard, significant condensable phases from hydrochar decomposition were not observed. The activation mass balance is shown in Tables S6 and S7.

Thermodynamic analysis of hydrothermal carbonization and physical activation
Carbonization and activation heat were theoretically determined through the difference of heats of formation of the products and reagents at their respective reaction temperatures. Except for pruning wood, hydrochar, and activated carbon, all standard enthalpies of formation and specific heat capacities can be found in the bibliography. The heats of formation of these three compounds were estimated through their heats of combustion, which were experimentally measured using a calorimetric bomb. Through modelling a generic combustion reaction (Eq. 2) [16], these enthalpies can be calculated as the difference between the heat of formation of combustion products (CO2 and H2O) and the higher heating value (HHV) of solids: . ( To solve the energy balances of carbonization and activation, solid heat capacities must be estimated by applying previous results for wood thermal properties [44,45]: where u is the moisture content of the material, is the fiber saturation point (near to 25%), and T is the temperature. It should be considered that the hydrochar moisture varied from 25% after hydrothermal cracking and filtering to 7.5% after the drying step (see Table 1). The energy balance revealed that hydrothermal carbonization is an exothermic process with a heat of reaction of -823 Joules per gram of pruning wood (Fig. 1a). This value fit closely with the experimental measurement of -760 J/g obtained by Funke and Ziegler for wood HTC at 240 ºC [46] and the recent assessment done by Pecchi et al. [47] for cellulose and wood. It is pertinent to mark the importance of executing meticulous elemental mass balances based on experimental measurements, in order to prevent unexpected deviations; for example, simplifying the analysis by supposing a gas phase composed entirely of CO2 would have led to a 35% overestimation of the carbonization heat in the present research.
Otherwise, the physical activation of hydrochar is an endothermic process ( Fig. 1.b) and the heat of reaction stands at 2,396 J/g. The strongly endothermic generation of CO and H2 from the carbon-steam reaction is barely compensated by CO2 formation by the decomposition of anhydride and carboxyl groups.  Figure 2 shows the industrial-scale continuous process designed for activated carbon production from vineyard pruning wood. This process was established after a previous thermoeconomic study that identified the optimum points for energy integration. Vineyard pruning wood is crushed in a crusher to homogenize the particle sizes and increase their surface area and reactivity. Then, part of the wood is mixed with water and pumped into the HTC reactor (Fig. 2a). Simultaneously, compressed nitrogen is introduced into the batch, in order to maintain the pressure at the operation set point (Fig. 2b).

System description
It is worth mentioning that a discontinuous reactor model was selected. Instead of the use of pilot projects with continuous or semi-continuous models, the preference to select current standard equipment and to follow a conservative approach for the cost analysis prevailed. To create a continuous flow, a battery of three batches working in parallel was chosen, in order to avoid the carbonization lapse. The reactors are heated with thermal oil, which takes the heat demand from the flue gas of burned raw material in exchanger E-6.
After HTC reaction, the hydrochar suspension is depressurized to 10 and 3.2 bars in flash tanks F-1 and F-2, respectively. As consequence of both expansions, two steam flows are generated. These flows are used to pre-heat the reactor feed in exchangers E-2 and E-3 and, downstream, as an activating agent into the rotary kiln. Flash pressures are established to warm the HTC inlet below the cellulose hydrolysis point (180 ºC), avoiding fogging in heat exchanger E-3 and premature carbonization. The outgoing slurry from flash tank F-2 is used to pre-heat the inlet water and conduced to a press filter, where the hydrochar is removed from the aqueous phase. This wastewater is pumped to exchanger E-7, in order to transfer its heat and supply a hot domestic water stream. Meanwhile, the char is dried and carried to the rotary kiln to accomplish activation. The activation gas heat is partially recovered in exchanger E-4, in which the condensed flow coming from E-3 is evaporated again.
Otherwise, the part of the crushed pruning wood that is not used to produce activated carbon (stream c) is burned to generate a compressed flue gas, in order to provide the remaining heat demand. To recover the sensible heat of the outgoing flue gas and the excess of heat emitted by the nitrogen compressor C-2, a Rankine cycle is coupled. This steam power cycle works at 60 bar and 480 ºC at the turbine inlet, and generates the electricity necessary for the operation of the entire plant.

Incidence matrix
The incidence matrix A (m x n) represents the physical structure of the process, connecting the m streams and n sub-systems. This matrix ought to be amplified by adding sub-system exergy flows or economic process structure, depending on the thermoeconomic data needed.
To define A (m x n), each element aij of the matrix should be valued +1 if stream j enters into sub-system i, -1 if it leaves, or 0 if there does not exist a direct physical connection between them [36]. Under steady-state operation, it is possible to describe mass, energy, and exergy balances using the incidence matrix: A x M = 0, A x E = 0, A x B = Bd, where M, E, B, and Bd are vectors of dimension [n] whose elements correspond to the mass, energy, exergy, and destroyed exergy flows.

Exergy vector
To perform the thermoeconomic analysis, the exergy of each process stream was calculated using the physical (bp) and chemical (bc) exergies of the contained substances (Eqs. 5-7): , , .
Pruning wood, hydrochar, and activated carbon chemical exergies were estimated using their Lower Heating Values (LHV) in Eq. 7 [48]: , where is a correlation that depends on the chemical composition of the solid [46]: , where zH, zC, and zO are the hydrogen, carbon, and oxygen mass fractions of the solids, respectively.
Finally, the exergy vector is completed by adding the exergy flows in each plant device. By multiplying an amplified incidence matrix and the exergy vector, the exergy destruction at each process stage can be determined.

Exergy costs
The exergy cost balance is defined by Equation 10 [36]: , where B* is the exergy cost vector, is the inverse of the amplified incidence matrix A (m x n) with economic structure matrix , and Ω is the vector of imputed exergy costs. All elements of α are null, except for the following cases: -Resources: a value of +1 is assigned to the αir elements corresponding to input resources; -Products and by-products: a value of +1 is attributed to the αip elements belonging to process products or by-products; -Wastes: a value of +1 is assigned to the αiw elements corresponding waste streams; and -Bifurcations: in those sub-systems with various exiting streams, values of + and are assigned to the corresponding j and k flows.
The vector Ω is defined by a series of null elements that correspond to the m rows of the process physical structure, while the other n-m elements are established as follows: -Resources: The values of , corresponding to resources exergy flows, are imputed; and -Wastes and bifurcations: null values are assigned.
By solving the exergy cost balance, the rational yield (τ) is obtained. This parameter denotes the process overall efficiency, relating the exergy contained in products with the exergy needed to obtain them: .

Economic costs
The economic cost balance is calculated using Eq. 12. It expresses the cost of the outputs through the cost of inputs and fixed costs (e.g., depreciation, maintenance, operations, and general plant expenses): .
The execution of the economic balance is quite similar to the exergetic cost balance. The same matrix is used, corresponding to imputed thermoeconomic costs. Φ is composed of the m sub-systems and the n-m costs of resources, by-products, wastes, and bifurcations. Otherwise, is the thermoeconomic costs vector, which indicates how expensive the manufactured products are.
The economic balance is based on equipment investments. Equipment prices were estimated by bibliographic procedures [49][50][51] and actualized according to the Chemical Engineering Price Cost Index [52]. Through this spending (C), the amount of capital needed to create the plant was deduced using the Lang method [53]. Once the plant investment (PI) is estimated, the annualized fixed costs flow (FCA) can be calculated by Eq. 13: , where A/Pi,n is the capital recovery factor and fRM is the repair and maintenance factor. The capital recovery factor is determined by cost of capital (i) and plant lifetime (n) in years: , , , (16) where E and D are the amounts of equity and debt, ke is the rate of return, kd is the cost of debt, and t represents income taxes. The costs of maintenance stand at 8% of the equipment spent, and 2 workers per shift were considered necessary to keep the plant running, with individual annual salary costs estimated at € 40,000. Finally, to define the vector and solve the economic balance, the annualized fixed costs flow is divided proportionally between the m subsystems of the plant: .

Energy and exergy cost analysis
The designed plant was analyzed at three different production scales, based on standard reactor capacities for hydrothermal carbonization [54]: 0.5 ton/h, 1 ton/h, and 2.5 ton/h of pruning wood for activated carbon. Additional mass flows of 1 ton/h, 2.1 ton/h, and 5.2 ton/h of winemaking waste were burned, respectively, to supply the energy process demands. To complete the energy analysis, the size of the cogeneration cycle, as well as electric and heat equipment consumptions were estimated. Table 2 summarizes the overall electric and heat demand and other parameters considered in the analysis, while Fig. 3 shows main energy consumptions broken down by equipment and operation. To describe the process thoroughly, Fig. 4 notes the exergy flow balances of the plant designed to carbonize 500 kg/h of pruning wood.  Compressor C-1, which compresses the air used to burn the wood and provide heat demands, comprises half of the total electricity consumption. Looking at the rest of equipment, the main consumers of electricity are the nitrogen compressor (C-2), the HTC reactor stirrers, the HTC reactor feeding pump (P-3), and the crusher. The remaining equipment represent minor electric consumptions, adding together to 5.2% of the overall.
The energy savings achieved by energy integration between hot and cold streams throughout the plant in heat exchangers E-1, E-2, E-3, E-4, and E-7 are remarkable, covering more than a third of the entire heat demand. The other 65% was provided by pruning wood combustion. Activation of hydrochar uses most of this heat. Thus, vaporization and heating up to 900 ºC in heat exchangers E-5 and E-6, and the endothermic character of the activation, hoard 22.7% of the plant heat flows.
Meanwhile, the heating of stream 11 up to 250 ºC in the hydrothermal carbonization reactors uses 11.9% and the drying of the hydrochar just 0.4%. Finally, the attached cogeneration cycle recovers more than half of the sensible heat of the outgoing flue gas and converts it into electricity, such that only 16.1% of the heat is lost in exiting exhaust gases, wastewater, and reaction waste gas.    As shown in Sankey diagram of Fig. 5a, more significant exergy losses happen due to chemical transformations, while thermodynamic equipment irreversibilities and heat loss represent minor percentages. Considering activated carbon, electricity, and hot domestic water as products, the overall exergy costs of the plant descend to 8.72 units per exergy unit produced, corresponding to an exergy efficiency of 11.5% (Fig. 5b). In this way, the combustion of pruning wood in the burner causes a major destruction of exergy, followed by activation and hydrothermal carbonization (Fig.  5c). These results emphasize the importance of the energy integration achieved by the heat exchangers network.

Economic analysis
Economic investments into equipment, the plant, and process operations were estimated according to the procedure described in chapter 2.4.4 and noted in Table 3, with other economic parameters used as reference.
Costs of equity and debt give a cost of capital of 6.25%. These values are based on a financial analysis of chemical industry [61] and financial reports of the EU [55]; the profit tax rate was set at 25%, according to the current Spanish taxation. Operation of the plant was estimated as 15 years with 4 shifts of 2,000 hours each and 8,000 annual working hours. Pruning wood cost was estimated through an actualization of the prices of this kind of biomass; process water and wastewater costs through the public tariffs of a local water company; and electricity cost according to the Eurostat statistics for Spanish industrial consumers [56].
Figs. 6 and 7 show the influence of production scales on economic costs. The increase of scale is crucial to ensure the viability of the plant. At the smaller scale, the investments and fixed costs associated with the Rankine cycle have an excessive weight and make the plant unprofitable under market scenarios with electricity costs below 0.115 €/kWh. For the higher scales of 1 ton/h and 2.5 ton/h, the incorporation of the cogeneration cycle leads to adequate results (Fig. 6a). In contrast, the cost of hot domestic water supply does not differ significantly, being only affected by the price of mains water and the expenses relating to heat exchanger E-7 (Fig. 6b).
As expected, the increase of the scale notably reduced the weight of the fixed costs on total economic spending (Fig. 7a). Wood price stood out among the variable costs, supposing more than 30% of total investments during the plant service life. This is an important aspect in the economic analysis, as long as the wood used comes from winemaking waste and its price could be cut down easily [32], [57]. Fig. 7b compares the impact of the most expensive equipment on the fixed costs. Hydrothermal carbonization reactors accumulate the highest expenses. This result agrees with other economic analyses carried out for hydrochar production from hydrothermal carbonization of biomass [32], [34,35]. The relatively high spending dedicated to acquiring the batches and the nitrogen compressor indicates that the development of new semi-continuous or continuous industrial reactors might be a point of consideration to increase the competitivity of carbon product synthesis from biomass and wastes.

Sensitivity analysis
A sensitivity analysis was carried out to determine the influence of economic and operational parameters on plant profitability. Figs. 8 and 9 show the activated carbon production costs variation with annual working hours, service life, cost of capital, and cost of pruning wood. To determine adequate economic scenarios, production costs of activated carbon obtained from wood (2.0 €/kg) and fossil carbon (1.10 €/kg) were set as targets [57,58].
The scale of production was demonstrated to be a fundamental parameter in economic plant viability and only at the highest scale was the plant systematically capable of achieving production costs which were competitive with wood activated carbon prices. Shifts and annual working hours also impacted the plant profitability significantly. Fig. 9a shows that it would be reasonable to reduce the original 8,000 working hours to 6,000 working hours for the highest production scale.
Further reductions on operational time could limit the economic feasibility, if other parameters were not improved. In this way, the cogeneration system clearly favors the plant competitivity under scenarios with higher workloads and service life.
Capital costs influenced the activated carbon prices moderately. Its increase caused a growth in plant investments, which marked an adverse tendency against the economic profits given by electricity cogeneration. In any case, the advantage of this system at high production scales is obvious and, even under other adverse scenarios, it manages to keep the production costs low, compared with one that takes the electricity from the grid.
Finally, pruning wood cost is the other great impact economic parameter, which is often the point of focus of projects working with wastes and low-cost raw materials. If the pruning wood cost stands at biomass market prices, the plant only would be able to compete with other biomass activated carbon process, but it is important to note that the wood used in the present research is a waste discarded by the wine industry and does not have any use as a consumer good. Besides, the accumulation of these residues can cause phytosanitary troubles and fires.
Legal regulation forces their proper treatment and elimination, such that its price may even be  a) b) Figure 9. Activated carbon production costs variation in function of the cost of capital (a) and pruning wood (b).

Conclusions
Conversion of waste vineyard pruning wood into activated carbon was studied through experimental research to determine the products formed and the thermodynamics involved. The experimental results and thermoeconomic principles were used to design and model a sustainable production process for activated carbons using these wastes. The analysis focused on improving the efficiency of the plant through energy integration and the coupling of a cogeneration cycle which recovers the residual heat to provide the electric demands. Furthermore, economic and sensitivity analyses were performed to identify the competitive weakness of the plant and to consider the impacts of different parameters on its feasibility. According to the results, the following conclusions can be stated: 1. Pruning wood combustion used to supply heat demand causes the major destruction of exergy and, so, efficiency improvements should necessarily be focused on avoiding heat losses and reducing heat demands through recovery from remaining sources.
2. Energy integration and cogeneration are capable of covering 48.9% of the energy demands of the plant, which demonstrates the utility of the thermoeconomic method to identify the improvement opportunities in process efficiency.
3. High scales of production are fundamental to ensure the economic competitivity of the plant and to reduce the impact of the fixed costs on the overall balance.
4. At the highest scale of 2.5 ton/h of treated pruning wood, the plant achieves production costs which are competitive with activated carbon made from wood (2.0 €/kg). In this regard, the sensitivity analysis revealed the convenience of maintaining a workload above 6,000 working hours per year. 5. Pruning wood cost was the most important parameter in the economic sensitivity analysis. Obtaining reduced costs, due to its waste condition, could allow for ecological activated carbon generation which is competitive with bituminous products, which have production costs near 1.10 €/kg.