Technical Evaluation of a PV-Diesel Hybrid System with Energy Storage: Case Study in the Tapajós-Arapiuns Extractive Reserve, Amazon, Brazil

: In 2018 the number of people without access to electricity dropped to less than 1 billion. However, the difﬁculty of serving these people became higher, as the locations are in the most remote areas of the world. Brazil, for example, needs to bring electricity to around 1 million people who, in the vast majority, live within the Amazon region. In this way, hybrid energy systems (HESs) count as an attractive alternative for power generation, especially in remote areas. Therefore, this article analyzes a case study of a hybrid photovoltaic-diesel system installed in the Tapajós-Arapiuns Extractive Reserve in the Brazilian Amazon region. The studied plant is composed of a photovoltaic (PV) system, a lead-acid electrochemical battery bank, a diesel generator, and electro-electronic loads with highly variable demand throughout the year. The HOMER PRO software is used as the simulation tool. The results show that the load following dispatch strategy is the best option, with 85.6% of the load demand being supplied by PV energy and only 14.4% by the diesel generator set. As a result, the system is technically feasible to be replicated as a reliable energy source in other areas of the reserve to supply schools, public health places, and other community services.


Introduction
Universal access to electricity has increased exponentially in recent decades. In the early 1990s, approximately 71% of the global population had access to energy. About 26 years later, in 2016, this number reached over 87%. However, 940 million people in that year still remained without any energy support. The number of people without access to electricity dropped from almost 1 billion in 2017 to 860 million in 2018 [1][2][3]. These data are extremely significant, but the challenge becomes higher as the populations who still need to be served are located in remote areas of the world. The crucial solutions to serve these populations predominate in the use of renewable sources such as solar and wind for the implantation of individual energy systems or hybrid systems.
Among other remote areas of the world, the Amazon imposes significant difficulties to the electricity supply. The Amazon region is characterized by its size, biodiversity, range of natural resources, and its population. Its population is mainly composed of traditional and rural communities. These populations are used to building small communities next to the forests, mainly on river channels that provide the inhabitants with locomotion and food.
(wind, solar, and hydro) to determine the optimal solution corresponding to a minimum annualized cost of capital. Many other authors also present works in the areas of autonomous HESs [6,21,[23][24][25][26].
The main contribution of this research is to highlight the potential for renewable energy in the Amazon region, showing the possibilities of meeting the diverse energy demands and offer renewable energy as a great solution in the form of hybrid systems to mitigate the use of diesel and reduce local energy problems. The sustainable propagation of energy in this region adds economic development to the local population. A consequence of this whole process is the improvement of quality of life. However, in the last few years, this population has faced unrestrained deforestation with higher CO 2 emission rates. However, remote communities in the Amazon are mainly responsible for fighting for sustainable land development, access to energy, and the preservation of the forest with its flora and fauna.

Autonomous Hybrid Energy Systems
The definition of stand-alone hybrid energy systems is their operation off-grid [5,6,27,28]. Generation systems connected to the grid in Brazil present mandatory standards aiming at the proper security of the electricity distribution system and its users. Autonomous systems are also subject to requirements according to the country's normative resolution N • 493/2012, which establishes the procedures and conditions for electricity supply through isolated energy generation and distribution systems or individual generation system of electricity with intermittent sources, also including hybrid systems resulting from the combination of two or more of the following primary sources: solar, wind, biomass, hydro, and diesel [5,29,30].
Autonomous hybrid energy systems can be used with isolated topologies or mini-grids in low or high voltage, single-phase or three-phase. The demand for power and the load to be installed is what governs the system specifications. Low-voltage (single-phase) networks to supply a low-power point demand (using televisions, cell phone chargers, radios, and lighting) dominate isolated systems applications. In contrast, a high voltage/three-phase grid has its application reduced to systems with high power loads such as motors and pumps [31][32][33].
There are two main classifications of system operating topologies: alternating current (AC) and direct current (DC). In the case of the DC topology, the photovoltaic generator and the diesel generator share the function of charging the battery storage. The DC serial topology requires more electronic power resources that can reduce the overall efficiency of the system, as well as the viability of the system. In addition, this topology limits the control of the diesel generator [34,35].
In the parallel topology, the generators have a connection on the AC bus and the storage system (batteries) to the DC bus. The operation occurs with the PV generation system charging the batteries when necessary and feeding the loads simultaneously. The diesel generator in the system has a backup function. The implementation of this system is more complicated due to the operation in parallel, but the advantages provide more significant optimization and efficiency [36][37][38][39].
A third option is the application of mixed topologies, where all sources supply DC energy, which is then converted to AC by power electronics equipment. PV generators naturally produce electrical energy in DC. The AC energy supplied by the diesel generator is rectified to DC. This is the most used topology, which allows the supply of AC energy according to required voltage and frequency standards through DC-AC inverters [5,40].
Despite many advantages, HOMER has limitations such as not considering the depth of discharge (DOD) of the battery bank that affects the calculation of the useful life and the number of batteries used in the system. The software does not operate with busbar voltage variations. It also does not have the possibility of 3D simulations to study shading when using a photovoltaic energy source.
Lambert et al. [66] defines HOMER PRO (Hybrid Optimization of Multiple Energy Resources) as a computational model developed by the United States National Renewable Energy Laboratory (NREL). The software consists of three main tasks: simulation, optimization, and sensitivity analysis.

Calculation of the Output Power of the Photovoltaic Array
The nominal capacity of the photovoltaic generator is the amount of energy produced under standard irradiance test conditions of 1 kW/m 2 and module temperature of 25 • C.
The following equation models the calculation of the output power of the photovoltaic array, taking into account the effect of temperature.
where, Y PV is the nominal capacity of the photovoltaic module under standard test conditions (kW), f PV is the photovoltaic energy reduction factor (%), G T is the solar radiation incident on the photovoltaic panel in real conditions (kW/m 2 ), G T,STC is the incident irradiance under standard test conditions (1 kW/m 2 ), α p is the temperature coefficient of power (%/ • C), T c is the photovoltaic cell temperature in real conditions ( • C), T c,STC is the photovoltaic cell temperature under standard test conditions (25 • C).

KiBaM Battery Model
HOMER incorporates in its simulations the battery model of Manwell, and McGowan [67], who developed an empirical model of lead-acid battery for application in hybrid energy systems, known as Kinetic Battery Model (KiBaM) [68,69].
The KiBaM model presents the battery charge distribution divided into two tanks: one is responsible for the charge available to be transformed into DC electricity, and the other corresponds to the limited charge (a chemical charge that is not yet available for consumption).
The KiBaM model ( Figure 1) shows that each part has a different height, width, and volume. A fraction c of the total capacity represents the available load (y 1 ) and the fraction 1 − c is the limited load (y 2 ). The process occurs with the available charge providing electrons for the load I, while the limited load provides electrons through the k valve for the available load. HOMER calculates the life of the battery bank in years, from: where N batt is the number of batteries, Q li f time the life rate of a single battery, Q tta the annual transfer rate (the total amount of energy that travels through the bank of batteries in a year) and R batt, f is the battery life (the maximum life, regardless of performance).

Diesel Generator Set
The sizing of the diesel generator set considers the complete supply of the need for peak load demand and powering the storage [66]. The main physical properties of the generator are its maximum and minimum electrical power, its expected life in hours of operation, the type of fuel it consumes, and its fuel curve, which relates the amount of fuel consumed to the electrical energy produced. Its characterization is given through fuel consumption in L/h or specific consumption in L/kWh to meet the particular load demand [70][71][72].
The fuel consumption of the diesel generator is described as a straight line with a y-intercept and can be calculated from the equation: where, P DG is the nominal capacity (kW) and PN DG the electrical output of the diesel generator (kW), respectively. The coefficients a 0 and b 0 are the intercept of the fuel curve which is the unladen fuel consumption of the generator divided by its nominal capacity and the slope of the fuel curve representing the marginal fuel consumption the generator. The results are expressed in L/kWh.

Locality Characterization
The Tapajós-Arapiuns Extractive Reserve is a territory of almost 650 thousand hectares, in the state of Pará within the Amazon region with the geographical coordinates 02 • 20' to 03 • 40' South, and 55 • 00' to 56 • 00' West ( Figure 2). It is the most populous reserve in Brazil, with about 22 thousand inhabitants distributed in 74 communities, most of whom are traditional low-income and vulnerable agricultural groups. Access to the reserve is made by boat through the Tapajós and Arapiuns rivers. In the strategic area of the region, close to the Tapajós and Arapiuns rivers, a non governmental organization called the Center of Advanced Social and Environmental Studies (CEAPS) has created a reference pole called "Active Forest" Experimental Center (CEFA). CEFA is a reference pole for training and developing social and environmental projects and technologies replicable to the entire forest. It is a place for seminars, training, and workshops for the local population. Sustainable development at the site brings together the main productive elements in a single integrated field system (nurseries, vegetable gardens, aquaculture, meliponiculture, among other activities). CEFA also hosts experimental renewable energy systems to study the feasibility of applying these systems in the region to foster public policies aimed at the energy development of isolated communities without access to the conventional electrical system.

Load Profile
Determining the typical load curve of a place with such climatic seasonality presents certain obstacles, as these variations directly influence the local habits of electricity consumption [33,73,74].
CEFA is a place with considerable variation in energy consumption and there is a high degree of complexity to accurately estimate its average annual load profile. After estimating the annual use of the loads in 2017, five different consumption profiles ( Figure 3) were found that appear at certain times of the year, and that drastically affect the use of each HES generation component. Realistically, the size and shape of the load profile vary from one day to another. We obtained five load profiles based on the consumption information of past years. With these data HOMER was used to plot random daily synthetic profiles with variability of 80.136%, with a constant consumption peak. The profiles retained the same shape for each day, but were scaled up or down. Thus, daily consumption was 6.16 kWh with a peak demand of 4.08 kW, with July having the highest peak demand.

Solar Energy Potential
The local radiation data taken from the NASA SSE base showed the annual average daily radiation was 4.96 kWh/m 2 /day ( Figure 4). March and May were the months of less radiation (4.37 kWh/m 2 /day) because of the rainy periods in the region and September had the highest daily radiation of 5.93 kWh/m 2 /day.

Assessment of Wind Energy Potential
Wind farms in Brazil behave differently according to geographic regions. The territory of the Tapajós-Arapiuns Extractive Reserve (Resex) occupies part of the Eastern Amazon Basin region. The Eastern Amazon portion extends around the city of Santarém approximately 100 km from the coast between Amapá and Maranhão. In this region, the tropical depression is part of the relief at low altitudes (in some locations below sea level). Light winds mold the wind profile in the central depression area.
The wind potential in this region of the Amazon is weak, with an annual wind average of less than 3.5 m/s. In the area of Resex, this value is rarely exceeded during the year, occurring between the end of July and the beginning of October. The average annual wind speed at CEFA is approximately 1.87 m/s. This low energy potential is due to the location and the high surface friction caused by the dense vegetation's roughness.
The analysis of the wind potential so that it has a technically usable result must present the result of a power density greater than or equal to 500 W/m 2 (E m ), at a standard height of 50 m above the ground, which requires a minimum wind speed of 7 to 8 m/s (V m ) [75].
By observing the wind data from Resex, the likelihood of success at a wind farm in this location is extremely low. Comparing the CEFA wind data with classes defined in Table 1, the result is the classification in class 1 (Vm > 3 m/s), proving the inviability of the wind project.
The HES initially had a 500 W wind turbine with 24 VDC output (18 A), connected to the DC link in parallel with the photovoltaic system for charging the battery bank. Although the wind turbine was part of the hybrid system, the wind potential in the region was scarce, as demonstrated in this topic.
HOMER simulations of the original HES corroborate the infeasibility of wind energy in this place. To reach the total power of 500 W of the wind turbine, the wind speed must be in the range of 15 to 16 m/s, being a wind regime that did not occur on this site.
The contribution of wind energy to the HES energy supply was approximately 1% of the total. Figure 5 shows the output power of generators and the load power during a certain time interval, in which the power of the wind turbine was negligible. Table 1. Definition of wind energy classes. In class 1 the probability of success in the installations is extremely low. Class 4 occurs in environments with greatest wind potential. Classes 2 and 3 have intermediary energy potentials. Source: Adapted from Feitosa [76].

PV-Diesel System Description
The CEFA has a highly variable energy demand throughout the year. The photovoltaic-diesel hybrid system ( Figure 6) feeds part of the loads such as electronics. In the local there are other photovoltaic systems to meet specific cooling and lighting loads. The presented system supplied AC loads in a 127 VAC grid. On the DC side there was a 2.1 kWp PV generator with an array formed by strings of two modules connected in series and several paralleled strings. The charge controller performed the energy management of the panels with a 1100 Ah battery bank. The AC power was provided by an inverter of 1800 W/24 VDC-120 VAC (technical specifications in Table 2) and a diesel generator set to 12 kVA. Table 2. Technical specifications of the Xantrex PROsine 1800 W inverter.

Model Xantrex PROsine
Output power 1800 W Outbreak classification 2900 W Output current (peak) 45 A Output voltage (AC) 120 V Output frequency 60 Hz The photovoltaic system had a power of 2.1 kW composed of four sets containing modules of 260 W and 265 W. The first two strings were composed of four photovoltaic panels of 265 W, with two modules associated in series and connected to charge controllers with a 24 VDC input. The 260 W modules had the same configuration as before and the load controllers were connected in parallel to connect to the inverter outside the grid (Figure 7).   The diesel generator set installed in the hybrid system had a Yanmar series YT22E stationary diesel engine, with a continuous output power of 16.2 HP/1800 RPM (11.9 kW) and intermittent output power 18 HP/1800 RPM (13.2 kW). This was a single-cylinder and four strokes engine, with direct injection and cooling system with a water tank. The starting system was manual, and the engine could support a 15 L fuel tank with consumption of 175 g/HPh (126.5 g/kWh). The generator was a 12.5 kVA Bambozzi synchronous alternator with 10 kW continuous generated power, four poles and 60 Hz-380/220 VAC output. The output voltage could be adjusted in a range of ±15% of the nominal voltage, in addition to operating at 1800 RPM and 16.8 HP traction power.
The battery bank had 10 stationary 220 Ah lead-acid batteries. Each battery had a nominal voltage of 12 V, a standard temperature of 25 • C, and fluctuation voltage in the range of 13.2 V to 13.38 V. The system had a configuration of two batteries in series, providing a nominal voltage of 24 V and a total capacity of 1100 Ah and 26.40 kWh. However, of this capacity, only 21.12 kWh could be used, taking into account the state of charge (SOC) of 80%. The number of cycles for these batteries ranges from 1095 to 10,300, considering that the discharges were close to 80%. The power conditioning system has controllers for solar and wind loads, protection boxes (DC and AC), inverter, and manual switching (Figure 8). These components operate by controlling the power flow between power generators and loads, as well as converting directly to alternating current. The system was configured for a 24 VDC nominal voltage. Therefore, all equipment followed this parameter. The transfer switch was like a switch between the loads and the two sources. If there was a power interruption by the photovoltaic generator at any time, the transfer switch activated the diesel so that it operated while the PV system resumed operation. This type of transition manually caused some human failure, such as detecting the ideal time to turn on each energy source.

Results and Discussion
From the technical infeasibility of the wind system, the analyses continued to consider only the sources with the best performance and the most significant assistance in meeting the loads. The reduced HES, now limited to a photovoltaic-diesel combination, remained capable of responding for 100% of the service to loads, in addition to reducing system losses and excess energy production. Thus, this session presents the results of the performance of this system based only on PV and diesel as energy sources.
HOMER was fed with input data to perform different configurations of the sources and combinations of the components. The result appeared through a list of the best systems. The system's input data were the specifications of a 10 kW diesel generator, 260/265 Wp solar panels, and 220 Ah batteries. Figure 9 shows the configuration adopted for the system, the mixed topology (series/parallel) with the generator set on the side installed on the AC bus for backup, minimizing the consumption of non-renewable fuels.

Analysis of the Daily Variation of the Minimum Load Profile
For much of the year, energy consumption was low, characterized by a daily power of 3.53 kW. This same demand varied throughout the year and increased by up to ±10% in everyday life without variation in the peak load of 660 W. Furthermore, for the optimization of the hybrid system, ideally, the photovoltaic system should meet the demand with a variation of up to ±30%. The use of the diesel generator in these conditions was not efficient, as it was on an empty work regime without operating at full power and with increased fuel consumption.
An analysis of the functioning and performance was carried out based on the daily variation of the minimum load profile of the hybrid photovoltaic-diesel system.
A variability of 10% in the load profile was added in the simulations because, in real cases, the loads did not behave steadily every day. Based on this information, HOMER changed the load profile for each day at a random value, maintaining the same peak demand. Figure 10 shows a month of the year with the load profile with and without variability. The results of the variation of 10% showed the system produced 99.9% of the PV generation, and 0.0840% by the diesel generator.
The maximum annual photovoltaic production was 2973 kWh/year, and diesel was started only once in this simulation for about 3 h with a production of 2.50 kWh/year and consumption of 1.20 L/day of fuel (Figure 11). This isolated event occurred in the morning between 7:00 a.m. and 9:00 a.m. when the charge was high, the batteries reached the minimum charge state, and PV production was starting. Figure 11. Behavior of the photovoltaic-diesel HES when there is low photovoltaic (PV) production with the activation of the diesel generator set.
The storage system allowed a minimum charge state of 40%, for higher battery life and less replacements, operating at 24 VDC. The results show that the battery bank received 1067 kWh of energy annually, having losses of approximately 214 kWh/year. The state of charge of the battery bank for different days of the year and the blue region showed the highest consumption of stored energy in March and the lowest in July.
The photovoltaic system produced 1125 kWh/year of energy in addition to the minimum load needs. It could supply a variation of up to 50% overnight without the need to start the diesel generator. When the load variation option was triggered, the profile changed without affecting its size.
Access to the power system usually caused demand to rise over time. Climatic variation also changed consumption. The communities of Tapajós-Arapiuns Reserve used the place, and the number of people was unstable, causing a more significant variation of the daily load profile. After that, the simulation gained variation of the load profile by 50% randomly during the days, maintaining the minimum daily consumption of 4.27 kWh. However, higher demand peaks of up to 1.25 kW occurred.

Energy System Analysis with Diesel Generator Set
Generally, isolated locations are supplied only by diesel generators, which causes significant losses, gas emissions, and high fuel costs. Besides, diesel transportation is highly dangerous due to the risk of soil and effluent contamination.
Without using renewable energy systems, this facility would be supplied only by diesel generators, which is a widespread routine for electricity supply in isolated areas. Thus, the simulations aim to verify and analyze from the equipment installed in CEFA's hybrid energy system concerning the load supply, the use of the generator set, and the failures that may occur. After simulating the system, it is possible to see the relationship between diesel and loads when they need a power supply. Figure 12 shows that the diesel generator operated with a minimum output power of 2.5 kW even when the load needs was below 0.5 kWh. As a consequence, the system produced 89.9% of excess electricity. That is, the consumption of loads was 2247 kWh annually, and the diesel generator produced approximately 20,039 kWh/year.

Hybrid Photovoltaic-Diesel System
The results obtained show that the hybrid system provided 85.6% of photovoltaic energy and 14.4% of the diesel generator, showing that the system is feasible and that the use of diesel was necessary only in times of peak consumption. The PV system produced an average of 8.15 kWh/day and generates 2973 kWh/year.
For better conservation of the battery bank, the minimum charge state of the batteries was 40%. This choice occurred through the result of sensitivity analysis for SOC min of 40%, 50%, and 60%. When the batteries reach SOC min and the photovoltaic system was not established to supply the battery bank, the diesel generator was activated. The state of charge throughout the year is shown in Figure 13.  The charging and discharging of the battery bank seen in Figure 15, presents the moments of low energy that occurred when there was no PV production, and the battery was in maximum discharge mode. When this situation occurs, the diesel generator started to supply the AC system, and DC was disconnected until its production stabilized ( Figure 16).  The diesel generator size permitted to meet the need for peak load demand and feed the building. The diesel generator operated for 200 h, being activated 156 times throughout the year. Energy production of 500 kWh/year, consuming a total of 239 L, a yield of 21.3%, and specific consumption of 0.478 L/kWh. Months with the highest incidence of activation were May, July, and September.
The results without using the battery bank show that the system was not viable because the renewable penetration by the PV system was only 13.4% and 86.6% of the diesel generator, causing an overproduction of electricity 89.9%. The absence of the battery bank in the system caused stability problems. The option with only the photovoltaic system and batteries is not considered in any simulation by HOMER, as the autonomous system would not be able to meet the need for loads alone in times of significant variation when using the consumption profiles 3, 4 and 5.
Without using the hybrid energy system, the loads were supplied only by the isolated diesel generator, which is a well-known routine for energy supply in remote areas of the Amazon. However, the internal combustion engines generated a series of problems for the health of the population and the environment. Table 4 presents the results of emissions generated in the pure diesel energy systems and the hybrid PV-diesel system. Comparing the data from the two systems, HES presented considerable mitigation in the emission of harmful gases. The harmful effects of exposure of the population, fauna, and flora to the emitted gases are seen in Table 5.  Table 5. Harmful effects on health and the environment caused by the emission of polluting gases [77][78][79][80][81].

Pollutants Effects Caused
Carbon dioxide (CO 2 ) Air pollution, greenhouse effect imbalance, acid rain, cardiovascular and respiratory diseases. Carbon monoxide (CO) Acts in the blood reducing oxygenation, nausea and intoxication.
Unburned hydrocarbons (UHCs) Precursors of the formation of tropospheric ozone and the potential cause of greenhouse effect (methane). Particule matter (PM) Respiratory cancer, lung inflammation and asthma symptoms.
Sulfur dioxide (SO 2 ) In the body they cause respiratory problems like asthma and the environment reacts with water in the atmosphere forming acid rain.
Nitrogen oxides (NO x ) It causes pulmonary edema in the human body and the environment prevents gas exchange, harming the process of photosynthesis.

Conclusions
The contribution of this work is the technical evaluation carried out on the existing local system, in which it shows that solar and diesel sources are viable components for application in HES in the region of Tapajós-Arapiuns Extractive Reserve in the Amazon region. However, the wind potential is not sufficient to provide success in power generation . In conclusion, the most suitable systems for replication in the region are HESs PV-Diesel or individual low-power systems with a solar energy system.
The case study, using the HOMER PRO Microgrid Analysis Tool software (version 3.12.0, Copyright (c) 2017, HOMER Energy LLC, Boulder, CO, USA), evaluated data such as energy balance, renewable generation potential, the behavior of eletrochemical batteries, the performance of the photovoltaic system, and the diesel generator set. This system is also a pilot project to be ideally replicated among communities in this region. From the simulations it was possible to draw a synthetic load profile involving all five consumption profiles built with variation in daily demand. The dependence on diesel fuel is as small as possible. As a result, it was possible to observe that a significant part of the energy used by the loads, about 85.6%, is provided by photovoltaic generation.
Through the sensitivity simulations for analysis of the daily variation of the load profile, it was possible to draw a minimum load profile for daily consumption in which the dependence of the diesel fuel is the smallest possible. With the synthetic load profile, it was possible to simulate variations in daily consumption of up to 50%, avoiding the increase in peak demand. Therefore, changing load variability between days without increasing consumption decreases the hours of operation of the diesel generator. The results for sensitivity analysis also show that from the increase of demand in specific periods and overtime makes efficient management of energy consumption at peak times in March and April necessary.
Another contribution is that the results on the feasibility of using hybrid systems can be used by local entities to demand appropriate public policies for the region's reality. The replication of this HES promotes a solution to expand the project to universalize access to electricity in remote areas of the Amazon. It also adds economic development through the contribution of CEFA to the local population, which generates the quality of life as a result of this whole process.
Thus, the research will continue in other publications bringing real data for monitoring the generation and load systems, economic evaluation, and optimization of the HES.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript: