Techno-Economic Feasibility Analysis through Optimization Strategies and Load Shifting in Isolated Hybrid Microgrids with Renewable Energy for the Non-Interconnected Zone (NIZ) of Colombia

: In developing countries, electriﬁcation in remote areas, where access to energy is limited or null, has been one of the biggest challenges in recent years. Isolated microgrids with renewable generation are an e ﬃ cient alternative for the energy supply in these areas. The objective of this work was to analyse the techno-economic viability of 6 isolated microgrids in di ﬀ erent locations in the non-interconnected zone of Colombia, considering di ﬀ erent climatic conditions, the availability of renewable resources, the current consumption proﬁle, and a modiﬁed proﬁle applying demand-side management. Modelling and simulation were performed considering storage systems based on lithium and lead-acid batteries. The resulting simulations provide the optimal system cost, emissions levels, electricity cost and battery lifetime. This study demonstrates that isolated hybrid microgrids with renewable energy are a feasible alternative to solve access to energy problems, reducing the need for diesel generators and optimizing the use of renewable energies and battery-based storage systems.


Introduction
A modern and reliable electricity supply is crucial for human well-being and for the economic development of a country. Access to energy is vital and allows the provision of drinking water, lighting, heating, food, transportation, and telecommunications [1]. However, approximately 1 billion people still do not have access to electricity and live in areas without connection to the electricity grid [2].
In Colombia, as in the rest of the countries of Latin America, access to electricity in remote areas is very limited due to the lack of infrastructure necessary to bring electricity to these places [3] and due to the precarious economic conditions of potential users, so it is not profitable for electricity companies to invest in these areas. In the so-called non-interconnected zone (NIZ), which corresponds to 52% of the territory of Colombia, approximately 92% of the electrical energy is generated by thermal plants with diesel generators, and the rest is generated by small hydroelectric plants. In these areas, the population lives in places with difficult access, which increases the price of fuel (diesel) in thermal plants, which emit polluting gases and generate noise pollution [4], so this generation system is not the most recommended [5]. In addition to this insufficient and limited electricity supply, autonomous energy resources have not been improved or used [6], which has led to poor social and economic development of the population of these areas.
The Government of Colombia, through Law 697 of 2001, established that the rational use of energy was a matter of social character and of national interest. Law 1715 of 2014 [7] advocated the use optimum system using very low computational times. iHOGA uses advanced models to accurately estimate the lifetime of the batteries, which are generally the most expensive components, with high requirements for costly replacements.
As a first criterion for the selection of microgrids, the availability of renewable resources was taken into account. The input data for the optimization correspond to irradiation and wind speed data for each location [43], as well as the actual daily load profiles, the modified profiles applying demand-side management, and financial data, such as the inflation rate and the interest rate of money. As a result of the optimizations, the sizes of each of the components were obtained, which corresponded to the solution with the lowest LCOE. In addition to the economic results, the total CO 2 emissions of the life cycle and the useful life of the batteries were obtained.

Geographic Location and Climate
The locations selected for this study are located within the NIZ. However, they have very different climatic conditions and renewable resources. The 6 selected locations have an altitude of less than 300 metres. The first three locations are found in tropical forests, Guacamayas is in the tropical savanna, Providencia has a dry tropical climate, and Puerto Estrella has an arid desert climate [44,45]. Figure 1 shows the map locations, and the Table 1 shows the geographic and climate information of the selected locations: Energies 2020, 13, x FOR PEER REVIEW 2 of 22 The Government of Colombia, through Law 697 of 2001, established that the rational use of energy was a matter of social character and of national interest. Law 1715 of 2014 [7] advocated the use of nonconventional sources for energy generation. Programmes such as PROURE [8] (Rational and Efficient Energy Use Programme and unconventional energy sources) promoted the financing of energy generation projects in the NIZ. All these initiatives can foster distributed generation and microgrids since they allow electricity to be generated and supplied near the places where it is consumed. A microgrid can be defined as a system that includes generation sources, loads and smallscale energy storage devices [9]. They are called hybrid microgrids when they combine two or more energy sources, such as diesel generators, renewable energies, fuel cells, etc. [10]. Hybrid microgrids can be an alternative energy supply in remote areas and they have already been used in several countries for the electrification of rural areas or islands [11][12][13][14][15][16][17][18][19][20].
The use of more than one source of electric power generation, of storage systems, and the intermittent nature of solar and wind irradiation complicate the design of hybrid microgrids, since it

Population
The population density of these localities is very low, classified as populated centers according to the classification of the National Administrative Department of Statistics (DANE) [46]. These communities have the particularity that they belong to rural areas with accessibility problems and they lack access Energies 2020, 13, 6146 4 of 20 to the electricity networks belonging to the National Interconnected System (NIS). Table 2 shows the number of households in each locality.

Energy Demand and Current Generation Sources
Most of the energy demand of these communities corresponds to lighting, small appliances and refrigeration equipment. The demand for energy during daylight hours is low because the main activities in these communities are agriculture and fishing, which naturally take place outside of the home. For the preparation of meals, mainly firewood is used, and in some cases liquid propane gas (LPG) [47]. Table 3 shows the current situation of energy demand and generation systems of the 6 locations. In the case of Providencia, the average consumption for an isolated house was selected, and in the case of Puerto Estrella, the consumption for 20 houses was considered, and wind generation was not considered because the wind turbines currently installed were not in service [48]. Figure 2 shows 2 of the systems considered in this work; they are currently operating as hybrid systems in Titumate and Puerto Estrella. communities have the particularity that they belong to rural areas with accessibility problems and they lack access to the electricity networks belonging to the National Interconnected System (NIS). Table 2 shows the number of households in each locality.  Chocó  Titumate  138  Amazon  Tarapacá  205  Nariño  Santa Rosa  173  Guaviare  Guacamayas  205  Guajira Puerto Estrella 20 San Andres Providence 1

Energy Demand and Current Generation Sources
Most of the energy demand of these communities corresponds to lighting, small appliances and refrigeration equipment. The demand for energy during daylight hours is low because the main activities in these communities are agriculture and fishing, which naturally take place outside of the home. For the preparation of meals, mainly firewood is used, and in some cases liquid propane gas (LPG) [47]. Table 3 shows the current situation of energy demand and generation systems of the 6 locations. In the case of Providencia, the average consumption for an isolated house was selected, and in the case of Puerto Estrella, the consumption for 20 houses was considered, and wind generation was not considered because the wind turbines currently installed were not in service [48]. Figure 2 shows 2 of the systems considered in this work; they are currently operating as hybrid systems in Titumate and Puerto Estrella.   The annual profiles of the daily demand for the 6 locations were prepared using data provided by the national monitoring center of the Institute of Planning and Promotion of Energetic Solutions (IPSE) in the Non Interconnected Zones [50]. Figures 3 and 4 show the average daily demand curves for one year, as well as the daily demand profile for the 6 locations. It is important to note that most of the Energies 2020, 13, 6146 5 of 20 energy demand is produced at night, except in the case of Providencia, where consumption is similar during all hours of the day. In some of the simulations, a modified demand curve was used applying demand-side management [51,52].
Energies 2020, 13, x FOR PEER REVIEW 5 of 21 The annual profiles of the daily demand for the 6 locations were prepared using data provided by the national monitoring center of the Institute of Planning and Promotion of Energetic Solutions (IPSE) in the Non Interconnected Zones [50]. Figures 3 and 4 show the average daily demand curves for one year, as well as the daily demand profile for the 6 locations. It is important to note that most of the energy demand is produced at night, except in the case of Providencia, where consumption is similar during all hours of the day. In some of the simulations, a modified demand curve was used applying demand-side management [51,52].

Availability of Renewable Resources
The 6 locations selected for this study have average daily irradiation values of 4.5 kWh/m 2 /day, exceeding the global average value [53]. On the other hand, only in Puerto Estrella and Providencia is generation by wind resource viable, with both having an average wind speed greater than 8 m/s [54]. Figures 5 and 6 show the hourly irradiation and wind speeds over a whole year. Puerto Estrella is located in an area with the highest average wind speeds in South America, making wind power generation viable [55].  The annual profiles of the daily demand for the 6 locations were prepared using data provided by the national monitoring center of the Institute of Planning and Promotion of Energetic Solutions (IPSE) in the Non Interconnected Zones [50]. Figures 3 and 4 show the average daily demand curves for one year, as well as the daily demand profile for the 6 locations. It is important to note that most of the energy demand is produced at night, except in the case of Providencia, where consumption is similar during all hours of the day. In some of the simulations, a modified demand curve was used applying demand-side management [51,52].

Availability of Renewable Resources
The 6 locations selected for this study have average daily irradiation values of 4.5 kWh/m 2 /day, exceeding the global average value [53]. On the other hand, only in Puerto Estrella and Providencia is generation by wind resource viable, with both having an average wind speed greater than 8 m/s [54]. Figures 5 and 6 show the hourly irradiation and wind speeds over a whole year. Puerto Estrella is located in an area with the highest average wind speeds in South America, making wind power generation viable [55].

Availability of Renewable Resources
The 6 locations selected for this study have average daily irradiation values of 4.5 kWh/m 2 /day, exceeding the global average value [53]. On the other hand, only in Puerto Estrella and Providencia is generation by wind resource viable, with both having an average wind speed greater than 8 m/s [54]. Figures 5 and 6 show the hourly irradiation and wind speeds over a whole year. Puerto Estrella is located in an area with the highest average wind speeds in South America, making wind power generation viable [55].      Tables 4-9 show the parameters used in the optimization of the microgrids in the 6 locations. Commercial PV modules, wind turbines, batteries, diesel generators and inverter/chargers were selected. The minimum/maximum number of components in parallel shown in Table 9 were obtained with the pre-sizing calculations of iHOGA, these were needed to limit the search space in the optimization. In all cases, it has been considered that the lifespan of the system coincides with that of the photovoltaic generator (25 years). The models used to estimate battery lifetime were those of Schiffer et al. [56] for lead-acid, and the model by Wang et al. [57] for LiFePO4/graphite lithium iron phosphate batteries. For the diesel generators and the wind and hydraulic turbines, the mathematical models are found in [58]. The calculations of life cycle emissions were based on previous work [59]. The inflation rate of Colombia was applied, which is currently 4% [60], and an interest rate of 7% was used. With these parameters, the iHOGA software was able to obtain the optimal solutions (generation of system configurations for each microgrid) using evolutionary algorithms [61].   In addition to the optimal configurations of the generation systems, iHOGA determined the most appropriate control strategy between the two that were considered. These two strategies are as follows:

Parameters Used in the Optimization
• Load following (LF): In systems that include batteries and a diesel or gasoline generator, when the energy from renewable sources is not sufficient to satisfy the demand, the batteries are responsible for supplying this deficit. In the case that the batteries are not able to supply all the energy demanded, it is the generator that must provide it. • Cycle charging (CC): Differs from the previous strategy in that in the event that the generator is required to operate, it will operate at its nominal power to satisfy the demand and, in addition, to charge the batteries only during that hour. There is a variant of this cycle charging strategy, called the setpoint strategy, in which the diesel generator continues to operate at its nominal power until the battery bank reaches a specific value of state of charge, which by default is 95%.

Simulation of the Current System
For the 6 locations, the current systems were simulated. In the six cases, taking into account that the generation of energy is carried out basically by diesel generators, high generation costs could be expected. However, a very low LCOE was obtained in the town of Guacamayas because in this case, in addition to diesel generation, there is a small hydroelectric plant. Table 10 shows a summary of the simulation results. The most adequate isolated microgrid for the energy demand of the population of Titumate corresponded to a combination of PV-diesel-battery. Figure 7 shows the daily load curve for this location considering the current load and the modified load with demand-side management. The modified load was obtained by changing the timing of some of the consumptions, to coincide with hours of high irradiation. In many cases it is difficult to change the hours of electricity consumption, since it implies a change in the population's habits. However, in this work we want to see the implication of this change in the cost of the optimal system. Energies 2020, 13, x FOR PEER REVIEW 9 of 21 The most adequate isolated microgrid for the energy demand of the population of Titumate corresponded to a combination of PV-diesel-battery. Figure 7 shows the daily load curve for this location considering the current load and the modified load with demand-side management. The modified load was obtained by changing the timing of some of the consumptions, to coincide with hours of high irradiation. In many cases it is difficult to change the hours of electricity consumption, since it implies a change in the population's habits. However, in this work we want to see the implication of this change in the cost of the optimal system.
The results are shown in Table 11, where a considerable reduction of the NPC is observed in systems with lead-acid batteries (52%) and in systems that use lithium batteries (56%). Similarly, the production cost of each kWh is reduced by 75-90% compared to the current system.   The results are shown in Table 11, where a considerable reduction of the NPC is observed in systems with lead-acid batteries (52%) and in systems that use lithium batteries (56%). Similarly, the production cost of each kWh is reduced by 75-90% compared to the current system.  Figure 8 shows the results of the optimization for the first 4 days of the year in the town of Titumate considering the cases of the current load and of the modified load with the best NPCs. It is observed how the SOC of the battery bank increases when using the modified load profile, remaining practically above 60%. This can extend the useful life of the batteries and simultaneously reduce the operation time of the diesel generator.
Energies 2020, 13, x FOR PEER REVIEW 10 of 21 Figure 8 shows the results of the optimization for the first 4 days of the year in the town of Titumate considering the cases of the current load and of the modified load with the best NPCs. It is observed how the SOC of the battery bank increases when using the modified load profile, remaining practically above 60%. This can extend the useful life of the batteries and simultaneously reduce the operation time of the diesel generator.  Figure 9 and Table 12 show, respectively, the load profile and the results for the optimal system configuration. The lowest NPC value is obtained with the modified load profile (1,125,231 €), with an LCOE of 0.14 €/kWh and an emission level of 36,049 kgCO2/year, with a useful life of batteries of 4.83 years.  Figure 9 and Table 12 show, respectively, the load profile and the results for the optimal system configuration. The lowest NPC value is obtained with the modified load profile (1,125,231 €), with an LCOE of 0.14 €/kWh and an emission level of 36,049 kgCO 2 /year, with a useful life of batteries of 4.83 years.

Tarapacá
Energies 2020, 13, x FOR PEER REVIEW 10 of 21 Figure 8 shows the results of the optimization for the first 4 days of the year in the town of Titumate considering the cases of the current load and of the modified load with the best NPCs. It is observed how the SOC of the battery bank increases when using the modified load profile, remaining practically above 60%. This can extend the useful life of the batteries and simultaneously reduce the operation time of the diesel generator.  Figure 9 and Table 12 show, respectively, the load profile and the results for the optimal system configuration. The lowest NPC value is obtained with the modified load profile (1,125,231 €), with an LCOE of 0.14 €/kWh and an emission level of 36,049 kgCO2/year, with a useful life of batteries of 4.83 years.   Figure 10 and Table 13 show, respectively, the load profile and the optimization results for the Santa Rosa locality. The optimization of the hybrid PV-diesel system reduces the NPC by 41% using lead-acid batteries and the current load curve and 34% with lithium batteries if the consumption is concentrated during daylight hours. Similarly, the LCOE is reduced from 1.46 €/kWh (see Table 3) to 0.24 €/kWh when lithium batteries are used and consumption is displaced. These results present a great improvement with respect to the current situation.  Figure 10 and Table 13 show, respectively, the load profile and the optimization results for the Santa Rosa locality. The optimization of the hybrid PV-diesel system reduces the NPC by 41% using lead-acid batteries and the current load curve and 34% with lithium batteries if the consumption is concentrated during daylight hours. Similarly, the LCOE is reduced from 1.46 €/kWh (see Table 3) to 0.24 €/kWh when lithium batteries are used and consumption is displaced. These results present a great improvement with respect to the current situation.

Guacamayas
For this location, the results are shown in Figure 11 and Table 14. The optimal system has an NPC of € 273,133 and an LCOE of € 0.08/kWh, corresponding to a PV-diesel-hydro system with lead-acid batteries and diesel-hydro with lithium batteries, under the same load profile. Having battery storage increases the reliability of the system, mainly against phenomena such as El Niño, in which the level of the rivers drops considerably [62]. lower in the modified case. The advanced lead-acid battery lifetime model used [56] considers many variables to determine the battery degradation, including, for each time step: current (charge and discharge rates), charge throughput, time between full charge, time at low SOC, partial cycling, temperature… A small difference in the load profile can imply low changes in these variables and therefore a small change in the battery lifetime estimation. In this case the modified load profile implies a slightly lower battery lifetime.    It can be seen that the modified profile, with the lead-acid battery optimal system, cost is slightly higher than the actual profile optimal system cost. It happens because the battery lifetime is slightly lower in the modified case. The advanced lead-acid battery lifetime model used [56] considers many variables to determine the battery degradation, including, for each time step: current (charge and discharge rates), charge throughput, time between full charge, time at low SOC, partial cycling, temperature . . . A small difference in the load profile can imply low changes in these variables and therefore a small change in the battery lifetime estimation. In this case the modified load profile implies a slightly lower battery lifetime. Figure 12 shows the load curve, and Table 15 shows the optimization results. The high average wind speed and irradiation values confirm that the optimal hybrid system is PV-wind-diesel. The NPC decreases when considering the modified load profile, 45.3% when using lead-acid batteries and 32.8% when using lithium batteries. A shorter longevity of the useful life of the batteries is observed because the ambient temperature of the locality is 30 • . Figure 12 shows the load curve, and Table 15 shows the optimization results. The high average wind speed and irradiation values confirm that the optimal hybrid system is PV-wind-diesel. The NPC decreases when considering the modified load profile, 45.3% when using lead-acid batteries and 32.8% when using lithium batteries. A shorter longevity of the useful life of the batteries is observed because the ambient temperature of the locality is 30°.   Figure 13 shows the simulation for the first 4 days of the year for the town of Puerto Estrella. With the modified load, the output power of the wind turbine is better used in hours of low radiation, which leads to an increase in the SOC of the battery bank.   Figure 13 shows the simulation for the first 4 days of the year for the town of Puerto Estrella. With the modified load, the output power of the wind turbine is better used in hours of low radiation, which leads to an increase in the SOC of the battery bank.

Providence
For this location, the optimal generation system is PV-wind-diesel, which has an LCOE up to 83% lower than the current system, based only on diesel generators. It also presents a considerable reduction in NPC and LCOE when the modified load profile is considered. Figure 14 and Table 16 show, respectively, the load profile and the optimization results for this location.

Providence
For this location, the optimal generation system is PV-wind-diesel, which has an LCOE up to 83% lower than the current system, based only on diesel generators. It also presents a considerable reduction in NPC and LCOE when the modified load profile is considered. Figure 14 and Table 16 show, respectively, the load profile and the optimization results for this location.

Providence
For this location, the optimal generation system is PV-wind-diesel, which has an LCOE up to 83% lower than the current system, based only on diesel generators. It also presents a considerable reduction in NPC and LCOE when the modified load profile is considered. Figure 14 and Table 16 show, respectively, the load profile and the optimization results for this location.   Figure 15 shows the different energy costs obtained with the current and modified load profiles for the 6 microgrids using lead-acid batteries and the load following strategy. A lower energy price is observed in 5 of the locations using a modified load profile. Figure 16 shows the NPCs of the 6 locations for optimization using lithium batteries and with the load following strategy, observing a decrease in costs in 5 of the 6 microgrids using modified load profiles. In three locations (Titumate, Santa Rosa and Puerto Estrella) the cost reduction is around 50% with the modified load profile.

Discussion
is observed in 5 of the locations using a modified load profile. Figure 16 shows the NPCs of the 6 locations for optimization using lithium batteries and with the load following strategy, observing a decrease in costs in 5 of the 6 microgrids using modified load profiles. In three locations (Titumate, Santa Rosa and Puerto Estrella) the cost reduction is around 50% with the modified load profile.
The level of emissions also decreases in 5 locations, as seen in Figure 17, where the results of the optimizations using lithium batteries with the cycle charging control strategy are presented.   The level of emissions also decreases in 5 locations, as seen in Figure 17, where the results of the optimizations using lithium batteries with the cycle charging control strategy are presented.
The results obtained in the simulation model of the microgrid is performed during several years (usually 20-25 years), the performance is repeated considering all years the same, considering the load to be constant. This is a limitation, as load can change during the years.
Further research should be done for the accurate estimation of the diesel price, considering that diesel cost in the NIZ of Colombia is highly variable due to its drawbacks associated with the transportation in areas of difficult access. Further research could also include sensitivity analysis considering factors such as: load variation, the price of fuel, renewable energy subsidies, interest rates and acquisition cost of components of the system. In addition, the simulations were performed using mono-objective optimization (minimization of NPC), however future studies can address the use of multi-objective optimization including equivalent CO 2 emissions, human development index (HDI) and job creation. All of these features are available in the iHOGA software.
From a technical and economical point of view, this study opens the possibilities for exploring isolated hybrid microgrids in developing countries like Colombia, considering future technological improvements and cost reductions in batteries and PV modules.  The results obtained in the simulation model of the microgrid is performed during several years (usually 20-25 years), the performance is repeated considering all years the same, considering the load to be constant. This is a limitation, as load can change during the years.
Further research should be done for the accurate estimation of the diesel price, considering that diesel cost in the NIZ of Colombia is highly variable due to its drawbacks associated with the transportation in areas of difficult access. Further research could also include sensitivity analysis considering factors such as: load variation, the price of fuel, renewable energy subsidies, interest rates and acquisition cost of components of the system. In addition, the simulations were performed using mono-objective optimization (minimization of NPC), however future studies can address the use of multi-objective optimization including equivalent CO2 emissions, human development index (HDI) and job creation. All of these features are available in the iHOGA software.

Conclusions
This article presents a techno-economic study of isolated microgrids of the NIZ of Colombia. Optimal generation hybrid systems have been obtained for 6 locations, considering the possibility of using diesel generators, solar panels, hydraulic turbines, wind turbines and batteries. The results show that NPC values lower than the current ones (powered mainly with diesel) can be achieved in almost all scenarios thanks to the reduction in the number of operating hours of the diesel generators and the use of demand-side management. However, this demand-side management is limited, to a large extent, by the difficulty of changing the consumption habits of users. The results have also shown that lithium batteries can be a good alternative to lead-acid batteries, considering the useful life and costs of the system.
It is important to note that optimization strategies could include a demand side management program that can reduce operation cost. In addition, the development of microgrids with renewable energies in rural areas also will help to meet the challenge of energy supply of remote zones and will reduce the dependence on fossil fuels. The study findings provide a basis to explore optimization of microgrids with other technologies such as fuel cells and biomass. Nevertheless, the Colombian government will have to play a crucial role for the development of the isolated hybrid microgrids in remote areas. Funding: This work was supported by the Universidad de Zaragoza programme "Proyectos de investigación-proyectos puente-convocatoria 2019", project "Modelos de envejecimiento de baterías de litio para su aplicación en simulación y optimización de sistemas aislados de la red eléctrica" [grant number: UZ2020-TEC03].

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