Determination of Biomass Power Plant Location Using GIS-Based Heuristic Methods: A Case Study in Turkey

6 Biomass conversion to bioenergy has always been necessary to ensure the most efficient use of 7 the limited biomass resource and enable economic viability. Evaluating biomass transportation 8 cost, electricity transmission cost and heat transferring cost between power plant location/s and 9 supply/demand points and selection of an optimum power plant capacity is an important issue for 10 a robust supply chain design. In this study, we employed designing optimum biomass to the 11 bioenergy supply chain for agricultural activities using Geographic Information System and 12 Simulated Annealing algorithm to overcome a real-world problem in Bismil District of 13 Diyarbakır /Turkey. Our goal is to define a potential investment location/s on the trigeneration 14 system by comparing the trade-offs between the raw material/end-product transportation costs 15 and facility/s and pipeline installation costs. To determine possible locations for power plants, 16 distance matrices were retrieved from suitable candidate power plant locations and agricultural 17 parcel, settlement and the nearest high voltage electricity line from the Geographic Information 18 System. The results showed that establishing one power plant is feasible. The net present value 19 of a potential investment is almost 260 million Euros and the re-payment period is 1.33 years. 20


29
Biomass can be defined as the total mass of living organisms that belong to a society consists of 30 species or consist of several species. Biomass is the most widely used renewable energy source 31 in the world today. It is contributing to energy used in power generation, generating electricity, 32 heating homes, fueling vehicles, heat for industry and buildings, and transport (IEA 2017). 33 Today, the key notion in biomass utilization is excluding the traditional use of biomass. Biomass 34 is commonly converted to energy through traditional methods that lead to inefficiency, health 35 problems, and environmental pollution (Akgül and Seçkiner 2019). Efficiency varies from 5% to 36 15% in traditional methods; however, it reaches up to 90% in modern bioenergy facilities. 37 Modern bioenergy term is generally used to refer to heat or electricity or transport biofuels and 38 exclude the traditional use of biomass. these incentives by governments. The high unit installation cost of bioenergy facility compared 95 to fossil fuel-based facility is compensated with these subventions. Despite the incentives, 96 biomass' geographically scattered structure, high transportation cost, seasonal fluctuation and 97 uncertainty in quantity may make a potential investment infeasible at first glance. The collection 98 and transportation cost of biomass accounts for 33-50% of the total biomass-to-bioenergy supply 99 chain cost (Kumar et al. 2006) Drawbacks in the biomass-to-bioenergy supply chain can be 100 eliminated through the scientific approach. Spatial analysis tools and optimization technics are 101 convenient methods for designing a robust energy supply chain. 102 The remainder of this paper is organized as follows. Section 2 presents a literature review about 103 other works closely related to the biomass to bio-energy supply chain problems which are solved 104 through meta-heuristic methods. In section 3, heuristic method, problem description and 105 geographic information system and spatial analysis tools for the problem dealt with are presented 106 in materials and methods section. The real-world problem is addressed in Section 4 to illustrate 107 the performance of the proposed heuristic technic. Finally, we conclude the paper and indicate 108 future research directions in Section 5. Bioenergy is derived from biomass which is a carbon-based biological material. Biomass is 113 converted into energy via three main methods (Turkenburg 2000). Thermochemical, biochemical 114 and extraction. Biochemical conversion is generally suitable for biomass which has a moisture 115 content higher than 60% like livestock residues and wastewaters, while thermochemical methods 116 are suitable for the biomass whose organic dry matter is higher than 60% like lignocellulosic  Gasification, which is a thermochemical method, is considered as the biomass to gas conversion 121 process since it is a suitable technology for lignocellulosic biomass. At the end of this process, 122 syngas is produced which has a high calorific value and mainly composed of hydrogen, carbon 123 monoxide, water, and fewer undesired contaminants. The syngas can run through gas turbines or 124 other power conversion technology (prime mover) to produce electricity. Integrating some 125 auxiliary equipment to prime mover allows generating cooling, heating, and power 126 simultaneously. This is called trigeneration (combined cooling, heating, and power (CCHP)). In 127 a trigeneration system, total system efficiency can reach up to 85% at full load, almost 40% of 128 which is accounted for electric efficiency, remaining is the thermal efficiency (Akgül and 129 Seçkiner 2019). The load level affects the efficiency of prime mover positively. The district 130 heating and cooling system (DHCS) must be integrated into a trigeneration system to transfer hot 131 and cold heat demand points. Some parameters including pipe diameter, the distance between 132 heat station and demand points, the temperature difference between supply and return water, 133 pressure loss, outdoor temperature have an impact on heat losses. Trade-offs between these 134 parameters must be analyzed to minimize heat loss.

135
The biomass supply chain comprises nine main consecutive steps: Cultivation, harvesting, 136 loading, raw material transportation, unloading, warehousing, pretreatment, conversion to energy 137 and end-product transportation. The baling and shredding (size reduction) sub-processes can be 138 included to supply chain according to a given case. All processes except for cultivation, 139 harvesting, loading, raw material transportation can take place in a trigeneration system.

140
Minimizing any kind of losses between these successive processes calls for optimization 141 technics.

142
Decisions on biomass to energy supply chain have been dealt with strategic, tactic and 143 operational levels. The strategic level includes long-term decisions like the selection of site, 144 capacity, technology, transportation node, biomass suppliers, and preprocessing facilities. Once a 145 strategic decision is made, it is very unlikely to be altered in the short term (Yue and You 2016).

146
Geographic Information Systems (GIS) has effective tools for all three levels, particularly at the 147 strategic level.

148
Within the concert of renewable energy technologies, bioenergy can play a decisive role during 149 the next decades, when smartly designed and applied under favorable conditions. In this respect, 150 an efficient and effective supply chain and logistics management represents one key parameter 151 (Turkenburg 2000). Efficient supply chain management and optimization is a very complex 152 problem. To solve the sustainable biomass supply chain management problem, mathematical 153 optimization modeling, heuristic or meta-heuristic solution technics can be analyzed with 154 understanding the concept of biomass to bioenergy routes. Unlike mathematical modeling, heuristic or meta-heuristics do not guarantee a globally optimal 162 solution, however, they are very fast in complex problems such as NP-hard and NP-complete 163 problems. They can optimize problems with non-linear, non-convex and non-continuous 164 functions in a reasonable amount of time. Also, meta-heuristics, which are generally inspired by 165 natural and physical events, is known as the most sophisticated heuristics. A heuristic is specific 166 to the problem, while meta-heuristic can be applied to all kinds of problems. Heuristics may stick 167 in local optimums and cannot be used in various types of problems, whereas meta-heuristic has 168 ways out of local optimums and can be used in a wide range of problems (Ghaderi et al. 2016). 169 As the size of the problem increases in combinational problems, the computational time of     3.3. Geographic information system (Spatial analysis) 213 In this section, a suitability model that was developed for the biomass to bioenergy supply chain The centroid points of the agricultural parcels were considered as a raw material supply point.

252
The points (centroid coordinate of the polygon) revealed as a result of land suitability analysis 253 were considered as candidate power plant points and residential areas whose household is greater 254 than 40 were taken as the hot and cold heat demand points. The electricity generated in the can be a supply point as cotton and corn can be grown in these parcels.      There are 14,391 parcels and 50 settlements whose population is greater than 50 based on the 371 data retrieved from GIS. Some parcels have a quite small surface area. Therefore, parcels whose surface area is greater than 20 da were considered not to reduce the efficiency of raw material  becomes smaller. So, the chance of accepting worse solutions is decreased. This also means that 407 the system becomes stable towards to end. While the temperature is high, the algorithm can jump 408 out of any local optimums. The higher an initial temperature is determined, the better exploration 409 in the entire search space. However, a high initial temperature increases the computation time.  Pseudocode for simulated annealing 433 Step 1: Set X0, T0, Tf, a, Nn and 

434
Step 2: Start at a random point in the search space

436
Step 4: Move to another location by using the move operator, 

437
Step 5: Look neighborhood points around the current solution point and move to one of these 438 points, Step Step 7: If not, generate a random number between 0 and 1 and check the random number is Step 7.1: If less, take it as the current solution, Xi = Xi+1

445
Step 7.2: If greater, don't take it and stay where you are and look around the current 446 point 447 Step 7.3: n = n + 1 448 Step 8: Repeat steps 5 to 7 while n ≤ N 449 Step 9: Tt+1 = a x Tt 450 Step 10: Repeat steps 5 to 9 while Tt ≤ Tf

451
Step 11: Return Xb  In this study, the average amount of agricultural residue of two years was considered for every 489 year in the project lifetime. Thus, there will be the same amount of income and outcome (cash second assumption is that a settlement is allowed to receive heat from just one power plant. as j, k and l indices respectively in this study. The parameters are exactly as same as the parameters given in that cited study. We invite readers to read our previous study to eliminate the 504 repetition of too many similar parameters in this study. Apart from those parameters specific to 505 our previous study, some extra parameters were determined in this study due to the first 506 assumption mentioned in the section of 6.2.1. Additional parameters were indicated as follows; Here, the low heating value of the agricultural products (LHV), the yield of those products per 529 km 2 (YPSK), organic dry matter rate of those products (ODMR) and unit raw material price of 530 those products (URMP) were given in Table 3. This table can   Approximately an installed power of 29 MW can be produced from corn and cotton residues in 538 the field based on the real data. Distribution of the whole power to candidate power plants is and 10 given as follows: determined based on the formulas presented in a published study (Seçkiner 2005).

586
Initial temperature (T0) was calculated in Equation 21 given as follows: In the beginning, the initial temperature is taken as a high value and a few hundred moves are 596 carried out at this temperature (Kirkpatrick 1983). The probability score (Pi), which is given by 597 the user at the beginning, is compared to the ratio of accepted moves to all attempted moves. If

598
Pi is greater than the ratio, this temperature is considered initial temperature; otherwise, 599 multiplying two increases initial temperature value. This approach was employed for the 600 calculation of the initial temperature in the proposed meta-heuristic. The final temperature can 601 be calculated similarly.  respectively. The number of moves for each temperature was taken as sufficiently large value to 636 search the entire space. In the first SA algorithm, it was taken as 1000.

637
The control parameters of the first approach were presented as follows.   Based on the result, the second algorithm got the same result just as the first algorithm.

664
Narrowing the search space reduced the computational time by nearly 90%.