Renewable energy resources show variabilities at different characteristic time scales. For a given resource and demand pro le, there is an absolute maximum achievable coverage (when limiting the fraction of energy lost during production and delivery to a reasonable value). To reach larger coverage factors, two plausible paths can be taken: a mix of resources with different time variabilities and/or an energy storage system. The case treated in this paper is the electricity supply of an Astronomical Center in the North of Chile. The economical feasibility of both possibilities is explored and compared to a grid connected alternative.
First, data from local weather stations was collected to have a realistic evaluation of the variability of the solar/wind resource at all time scales. Then, we developed a scalable design of a solar/wind plant and a pumped hydro energy storage system. The free parameters of the design are the maximum installed power for each resource and the capacity of the storage system. Finally, the electricity production is calculated to determine the coverage factor and losses for different values of these parameters.
We found that a coverage factor of 64% is economically feasible for systems without storage. The associated total losses are 24%. To reach larger coverage factors is not economically possible and a storage system must be introduced. If this is done, there is a quantum increase of the total cost of about 30%. However, losses are reduced to about 5% and the coverage factor reaches almost 90%. The cost increase is marginally economically feasible, but it has some other advantages: the consumer is independent of the volatility of electricity prices, and is more sustainable.
The time variability of renewable energy resources difficults reaching coverage levels larger than 60%. Energy storage systems are a requirement. Periods of zero net production seem unavoidable unless the renewable energy and storage system are largely overdimensioned. Back up systems based on fossil fuels seem to be unavoidable. Both the energy storage and back up system add an extra cost that has to be paid if such high coverage levels are a requirement.
The case treated in this paper is the electricity supply of an Astronomical Center in the North of Chile. The ESO is the European Organization for Astronomical research in the Southern hemisphere. It operates the VLT (very large telescope), located at Cerro Paranal in the Atacama desert, North of Chile. The E-ELT (European extremely large optical/infrared telescope) in Cerro Armazones (20 km away from Cerro Paranal) is in advanced design phase and will be the largest optical telescope in the world. Finally, the CTA collaboration (Cherenkov Telescope Array) has chosen the Armazones-Paranal site for construction of its Southern Observatory.
When the two new observatories enter in operation, the peak power demand of the Armazones-Paranal site is estimated to be ∼ 8.5 MW and the total annual energy consumption ∼ 70 GWh. Currently, the VLT is generating its own electricity using fossil fuel-based generators.
The two main characteristics of this consumption center are the strong requirements on the stability of the electricity supply, and the relatively large power demanded. Due to these two factors, the use of liquid fossil fuels is economically un-viable. The only two non-renewable solutions plausible are connection to the Chilean national grid or self production of electricity using generators run with natural gas from a nearby pipeline.
The main renewable energy resources available at the site are wind and solar. In this work, we consider a wind-solar PV plant with Pumped Hydro Energy Storage (PHES). We calculate the coverage factor for different values of total Power, Maximum Energy Storage and wind to solar fraction to find energy systems that maximize coverage but with costs below the non-renewable energy solutions. Embedded in this procedure is the fact that renewable energy time variability can be diminished by considering a mixture. An important ingredient of this procedure is the relative cost of each technology. Government estimates are taken when possible.
Additionally, a concentrated solar power (CSP) plant with thermal energy storage is analyzed. This technology is considered separately since the storage system cannot be used by the wind farm.
The design of the systems is not detailed but all sources of inefficiencies are taken into account. The wind and solar input data used is from local weather stations, which provides realistic time series that account for all possible sources of the variability of the resources. Overall, the estimates of electricity production and cost are as realistic as possible so they can be used as a guide if such energy systems are eventually implemented. The total cost of each system includes operation and maintenance over the 25 year lifetime of the astronomical center.
The paper is organized as follows: in “Energy demand” section, the energy demand is described; in “Non-renewable energy systems” section, the non-renewable energy systems and their cost are analyzed; in “Renewable energy resources available in the site: solar and wind” section, the solar and wind data used in our calculations is described; in “Renewable energy systems” section, the methodology to calculate the time series of electricity production for the Wind-Solar PV plant with PHES is presented, together with a modular design of each of the subsystems and their cost; in “Results” section, an algorithm to find the optimum system is presented and compared to the non-renewable energy alternatives. The CSP with thermal storage design and cost are presented in the Appendix.
The energy demand of the VLT is known [1]. The power demand changes from day to night but is rather constant along the year (less than 5% variability). The projected E-ELT (CTA) consumption is taken from ESO estimates [2]. All the sub-systems, including lodging, offices and workshops are included. A simplified model is adopted: a constant power with different day/night values. The start/end for day/night will be calculated using the sunrise and sunset, even though the start/end of astronomical observations is typically later/earlier.
Table 1 shows a summary of the site energy demand. Night consumption is smaller than day for the E-ELT and VLT due to the strict thermal control system.
A 8.5 MW combined cycle gas turbine (CCGT) is considered in this case: it has high efficiencies ∼ 55% and fast time responses. Since there is already a 2.5 MW generator with these characteristics in the site, it will be only necessary to upgrade it with 6 MW more. We consider an investment cost of 1000 e/kW, i.e., a CAPEX of ∼ 6 Me. Natural Gas supplied by Gas Atacama, whose pipeline passes through the middle of the Armazones-Paranal site, can be used to run these generators. The expected connection cost is ∼ 2.5 Me: a gas sub-station, a low capacity (7000 m 3 per day) 5-km pipeline and a low capacity tank for regulation. In total, the CAPEX of the back-up system is 8.5 Me.
The OPEX is mainly due to the purchase of natural gas. The natural gas prices are high in Chile. The projections from the Chilean government are taken to correct the world market prices to the special case of Chile. The following equation is adopted to estimate the time-dependent price of a kWh generated by CCGT:
where C gas is given by (1+f N years )·9, N years is the number of years since 2015 and f takes into account the interannual increase of prices. We consider two values: f=0.01 and f=0.1. This equation yields 0.07 e/kWh for 2015.
Due to the strong requirements on the stability of the supply, this system is also a requirement for all renewable energy systems considered.
The total cost normalized to year 0 is estimated using:
$$ C= CAPEX + \sum_where k is the interest rate, 3%. The lifetime of the observatories and the renewable energy system is taken as 25 years. Table 2 shows the results.
The temperature is also an important factor that determines the performance of solar plants. The weather station temperature time series is used in our calculations.
Wind and speed direction from the VLT meteo mast is used to characterized the wind resource [5]. Measurements at 10 and 30 m from the last 15 years exist. Table 3 shows the average wind speeds at 30 m for the last 10 years. Figure 4 shows the wind speed distribution for the year 2011 at 30 m.
The mentioned software does not provide a time series of the produced electricity. This is a problem for our study: an storage system cannot be dimensioned without them. To overcome this problem, we use the following assumption to characterized the time series:
where P Turbine is the turbine power as a function of air density and wind speed at hub height:
$$ v_where v 30(t) is the measured temporal series of the meteo mast at 30 m, f vertical is a factor to extrapolate measurements to different heights:
$$ f_>=U(z)/U(z_)=(z/z_)^,~ \text~\alpha =0.05^ $$and f horizontal is a factor that takes into account the geographical variations of the wind speed. The value of f horizontal is adjusted so Eq. 8 gives the same duty factor as OpenWind.The PV and Wind plant requires an electricity based storage system that fulfills the following criteria:
The only technology that matches these criteria is the pumped hydro energy storage (PHES). The site is located in the Atacama desert where water is scarce. Due to the proximity to the coast, there is the possibility to use sea water as storage medium. However, due to the size of the facility and plausible technological and environmental problems, it is advised the use of desalinated water either self produced or bought.
The PHES plant consist in an upper and lower water reservoir connected by penstocks, and a system of turbines and pumps than convert gravitational energy into electricity or vice versa. The system is closed, so filling of the reservoirs has to be done only once. A separate turbine and pumping system is planned, so typical elapsed times to go from pumping to full load generation are of the order of minutes. Water evaporation 4 and filtration of water are important and will be taken into account in the design. P \(_^\) is fixed to 14 MW, so hydraulic losses does not severely affect the design.
The hydro power in W is given by:
$$ P_where ρ is the water density in kg/m 3 , g is the gravity acceleration constant in m/s 2 , Q is the water flow rate through the penstocks in m 3 /s, and δ h n is the net height difference given by:
$$ \delta h_where δ h g is the gross height difference and δ h(Q) are the hydraulic losses in the whole system that depend on the flow rate. The electric power in generation mode is given by:
$$ P_where η turb and η gen is the efficiency of the turbine (that depends on load) and the generator. The electric power in storage mode is given by:
$$ P_where η pump and η mot is the efficiency of the pump system (that depends on load) and the motor.
The required value of P e turb / P e pump is 8.5/14 MW.
The design of the system proceeds in two phases:
The site selection implies indirectly choosing two important variables: δ h g and penstock length. The second variable is crucial when determining the hydraulic losses, and is an important contributor to the total cost of the system. As a general rule, larger values of δ h g and smaller penstock length yield smaller investment costs. However, other factors have been analyzed:
Topographic maps have been used to choose four possible sites. All sites have similar availability of water/infrastructures and geological risks. Therefore, the site with larger height difference and the smaller penstock length was chosen. Figure 6 shows a detailed topographic map of the site. It is located in the Coastal Cliff, close to the Wind Farm location.
Our choice for the turbine system is the use of two Pelton turbines with one injector that can work in parallel to provide the maximum power. The Pelton turbines can work up to 10% of the nominal load, have efficiencies around 90% and are adequate for the site height differences and required nominal flows. The turbines will be coupled to two generators with nominal power 5 MW, AC output voltage of 6 kV and 98% efficiency.
Regarding the pumping system configuration, our choice is the use of multistage centrifugal pumps: 6 of 2 MW and 2 of 0.5 MW. To simplify the calculations an efficiency of 90% for all loads is considered. The motors that drive the pumps work at 6 kV with an efficiency of 98%.
Steel penstocks have rugosities of ∼ 0.6 mm. The hydraulic losses are calculated using standard formulas for different pipe diameters. For each case, the nominal flow rate in production and storage mode is calculated by solving iteratively Eq. 13/Eq. 14. The hydraulic losses drop below 5% in both modes at nominal conditions for a tube diameter of 0.85 m. Losses because of other hydraulic components like valves, bypasses, contractions/expansions, etc. are small (10% of Penstock losses) and taken into account. Table 5 gives the final nominal flow rate and hydraulic losses in both modes. Using these calculations the storage efficiencies s 1 and s 2 are calculated.
On the basis of the costs shown in Table 2, we select two target maximum costs: 100 and 130 M e. For each simulated case, the total cost over 25 years is calculated as in “Cost estimation” section. The case with a cost below the target and with maximum coverage is kept. The two cases selected for the two targets are shown in Table 8. The coverage factors are as large as 64 and 88%. It should be mentioned that the losses for the high cost target are driven by the storage efficiency, transport losses, and availability.
Table 8 The selected systems for the two budget thresholdsFinally, the case of a concentrated solar power (CSP) plant with thermal energy storage is analyzed. This technology is considered separately since the storage system cannot be used simultaneously by the Wind farm 6 . The design and costs are presented in Appendix. The total cost is 124 M e and f cover 72.5%. This alternative is within the high cost target, but it has lower coverage factor than the case presented in this section.
1 I 0 is only 12% smaller than the irradiance outside the atmosphere (1370 W/m 2 ), which is an indicator of the quality of the site.
2 The electricity production using the model and the raw data for the reference year agrees within 5%.
3 α=0.08 from the ratio of the measurements at 10 and 30 m. A conservatively smaller value is taken: measurements at 10 m can be affected by the surrounding buildings
4 According to our estimations, it can be severe, reducing the water level by almost 3 m per year.
5 It is calculated assuming: PV system prices will decrease at a rate of 20% over 25 years; PV module degradation is 20% over 25 years.
6 Electricity from the Wind Farm would have to be converted into thermal energy. To convert back to electricity the efficiency is given by the steam turbine, ∼ 32%.
The CSP is a technology that needs to be considered when there is plenty available land, the cloudy fraction is small and the fraction of direct irradiance is high. The dessert characteristics of the site fulfill these three criteria. The technology considered in this work is the parabolic trough collectors (PTC), widely considered in a stage of maturity.
In a CSP plant, an oil is heated in the solar field from 293 o C to 393 o C and sent either to the thermal storage system or to a heat exchanger that produces water vapour at 380 o C and 104 bar. The vapour is then conducted to a steam turbine coupled to a generator. After the turbine, the vapour is taken to a condenser and fed again into the loop. Due to the scarcity of water in the site, aerocondensers are considered. The efficiency to convert thermal energy into electricity depends on the nominal power of the turbine and for a 10 MW steam turbine is ∼ 32%.
The solar field is an array of PTCs. The mirrors have a one axis tracking system (North-South) that ensures that at all moments the solar vector lies within the plane perpendicular to the aperture of the collector. Alignment is a strong requirement in PTCs, and also cleaning.
The PTCs have lengths between 100 and 150 m. The 8 module EuroTrough collector with PTR-70 Schott tubes is selected. N series of these modules are placed in series to form a group. N parallel groups are connected in parallel in central feeding configuration to minimize pipe lengths. The separation between rows of collectors is three times the width of the parabola to ensure that annual shadowing losses are below 1%.
The thermal power captured by the collector is given by:
$$ P_where \(\eta _>~K(\Phi)\phantom \!>\) is a parameterization of the optical and geometrical losses of the collector, A c is the aperture area, I D is the direct irradiance in W/m 2 at the period considered, F e is a factor that takes into account the dirt in the mirrors (0.95), and P losses are the thermal losses parameterized with its dependence on the temperature difference between the fluid and the ambient, as well as on the direct irradiance and incidence angle.
The collected power can also be written as:
where Q m is the fluid mass flow in kg/s, C p is the specific heat in J/K Kg and T in /T out is the start/final temperature of the fluid. The thermal fluid chosen is an oil called Therminol VP1. Its maximum working temperature is 398 o and solidification temperature is 12 o . This fluid has to be pressurized to 10.5 bar so it is not gas phase at the maximum working temperature. The specific heat and density depends on temperature and is taken from a parameterization provided by the manufacturer.
N series of collectors have to rise the fluid temperature from T in =293 o C to T out =393 o C. The necessary value of Q m is calculated iteratively by equating Eqs. 17 and 18 in 1 m intervals.
The fluid must circulate in a regime turbulent enough to avoid thermal gradients between the external/internal face of the tube that can cause fractures. The optimum value of N series is calculated by imposing a condition on the Reynolds number of the circulating fluid for the time of maximum direct irradiance. In our design, N series must be 4.
The hydraulic losses are calculated for each configuration of the system (N series , N parallel ) and time period considered using the oil and tube characteristics and ambient conditions. Losses in the pipes that connect the collectors with the heat exchanger and the losses in the pump are also taken into account. 7 .
The required electrical pumping power is given by:
$$ P_where η m ( ∼ 70%) and η e ( ∼ 99%) are the mechanical and electrical efficiency of the pump.
The electrical power produced by the plant is given by:
where η is the efficiency to convert thermal to electrical energy (32%). The storage efficiencies considered are s 1= s 2=96% (Round trip efficiency of 92%). Transport losses are only applicable to t 3 (2%). The availability is included as described before.
The electricity production described in “Electricity production time series: methodology” section is calculated in 10 min intervals during a period of 48 hours around the summer solstice. N parallel is increased until f cover =100%. The required value of N parallel is 33. E MSC is given by the maximum storage level during the design period (100 MWh).
The storage system must be able to store 100 MWh, i.e., 312 MWth. This capacity is increased by a safety margin of 8%, i.e., 337.5 MWhth. The temperature in the hot/cold tank corresponds to the temperature of the oil before/after the heat exchanger. Nitrate salt (60% by weight NaNO3 and 40% KNO3) is considered as storage medium. The mass required can be calculated using:
which yields 8530 tons of salt to store 337.5 MWth. The corresponding volume of the hot and cold tank is different due to temperature. The volume required for the cold/hot tank is 4471 and 4618 m 3 . Fast fluctuations of the solar resource are easily tracked by the thermal storage system by controlling the flow from the solar field that is diverted to the heat exchanger of the storage system.
The electricity production is then calculated for the whole year. The results are shown in Table 9 together with the main design parameters.
Table 9 CSP design parameters and annual resultsA flat area is necessary to ease installation of the solar field. A possible site has been found 10 km away from Cerro Paranal. The access road to the Cerro Paranal passes by the solar field, so no extra civil works are planned. For electrical infrastructures and their cost, see Table 10.
Table 10 Electrical infrastructures required for the CSP plantThe investment cost (CAPEX) of the CSP plant is estimated to be 58.5 M e. Table 11 shows the breakdown. The OPEX considered is 2% of the CAPEX with a 3% inter annual increase. The total cost after 25 years normalized to year 0 is 124 M e.