Increasing electricity access for health facilities in Ghana through solar powered mini-grids—a GIS-based energy system modelling approach

In off-grid areas, health facilities could serve as the starting point for rural electrification. The aim of this paper is to create an overview of the solar-mini-grid potential to support the reliable operation of health facilities located in off-grid or weak-grid areas in Ghana. Moreover, in off-grid areas, health facilities could serve as the starting point for rural electrification. According to the World Bank, the electrification rate within the country is 85.9% (2020) [1]. Rural communities in particular are among those lacking this access, and with it, the healthcare facility (HCF) in their area. However, many services provided by HCFs require reliable electricity supply: From offering vaccinations and antidote against snake venom, which require a reliable cold chain and thus the operation of refrigerators, to the holistic examination of body parts and functions using x-ray and ultrasound equipment, to monitoring vital functions during surgery. When electricity access is limited or nonexistent, HCFs must limit the services they offer to surrounding communities, torpedoing their access to quality healthcare.

Both challenges—access to electricity and quality health care—are anchored in the United Nations Sustainable Development Goals (SDGs) as SDG 3 (good health and well-being) and SDG 7 (affordable and clean energy). Both targets are also national priorities in Ghana. The Ghana Health Service (GHS) under the Ministry of Health has the mandate to increase access to good quality health services and develops appropriate strategies and sets technical guidelines to achieve Ghana national policy goals and objectives [2]. The Ghanaian National Energy Transition Framework developed by the Ministry of Energy foresees to achieve an electrification rate of more than 95% by 2030 [3]. Thus, political targets are being set.

Despite these efforts being made, the current situation remains difficult for rural communities and their healthcare systems. Investment decisions in the energy sector have long lead times and long-lasting effects, as power plants and grids often last for 40 years or more [4]. Although Ghana has a high averaged daily photovoltaic (PV) output potential of 4.02 kWh kWp⁻¹, the electricity generation mix is still dominated by fossil fuels [5, 6]. This fact underlines the importance of strategic and integrated planning, especially in the energy sector, which is responsible for one-third of the country's CO₂-emissions and impacts many other important areas with vital functions for society (e.g. provision access to quality health care) [5].

The large number of communities situated in remote areas (approx. 10 km distance to grid) makes the challenge to reach the last mile with electricity more difficult. In general, three common ways to electrify communities exist: via small stand-alone systems (e.g. solar-home systems), decentralized mini-grids (run on diesel generators and/or renewable energy technologies and/or battery storage), and power grid extensions. As this paper focuses on the reliable provision of HCFs with electricity in order to guarantee quality health care access to surrounding communities, smaller stand-alone systems are not considered. These systems suit to electrify single households, but the amount of electricity provided is insufficient to power HCFs. In communities connected to the power grid—especially in rural areas—shortages in power generation lead to power outages (also called Dumsor). Connection to the power grid is therefore no guarantee of a reliable electricity supply. Mini-grids, on the other hand, have their own power generation at site and are therefore self-sustaining island grids that supply electricity to their customers. For this reason, we focus our analysis on the electrification of HCFs with mini-grids—and solar-powered mini-grids in particular, given the enormous solar power potential in Ghana as well as the reduction potential for CO₂-emissions.

Given that both—access to electricity and access to quality health care—are critical issues for rural communities in Ghana, it is obvious that integrated health sector and energy system planning is necessary for the countries's sustainable development. In order to do so, an overview of the location and energy access situation of HCFs in Ghana, focusing on those having limited or no electricity access, is necessary. Currently, such an overview is not publicly available: there is data on the type and location of HCFs (e.g. through the Ghana Open Data Initiative (GODI), Open Street Map, the WHO) and power grid data has been estimated, e.g. based on night light imagery by the Global Predictive Mapping Project. However, these datasets from different sectors (energy and health) are barely linked.

In order to address this gap, we formulated the following research questions:

Research Question 1: Which HCFs are most suitable to be electrified with PV mini-grids in order to increase their energy self-sufficiency and reliability of services provided and which communities are within their service area?

• Which HCFs are likely to have no or poor access to electricity supply?
• How can these HCFs be categorised?
• Which settlements are within the catchment area of these HCFs?

Research Question 2: Are PV mini-grids an appropriate option to increase energy self-sufficiency at these HCFs, and do the surrounding communities play a role to stabilise cost and electricity supply?

• What is the electricity demand for identified HCF sites and their surrounding communities?
• Which mini-grid system setup is most suitable to electrify identified HCF sites?
• What is the impact on system sizing and system costs of including the demand of surrounding communities into the energy system modelling?
• How does accounting for allowed electricity shortages from surrounding communities affect the mini-grid economics when included in energy system modelling?
• What is the accumulated solar mini-grid potential to electrify all identified HCF sites?

Ghana's healthcare system includes a variety of HCFs, which can be broadly categorized into private and public entities. Private HCFs are owned and operated by private individuals and organizations, or fall under the private not-for-profit category, where mainly missionary and faith-based organizations directly deliver health services [7]. Public facilities, on the other hand, are operated by the Ghanaian government through the Ministry of Health and currently deliver the largest proportion of health services to the Ghanaian population [8, 9]. The GHS serves here as the operational authority that rolls out the governmental health plans. Ghana's public health system operates on three main levels: primary, secondary and tertiary (figure 1). Primary facilities are the first access point to health care, referred to as Community-based Health Planning and Services, or CHPS [9]. These treat most minor ailments and provide close-to-client preventive care, as well as basic and non-invasive health services, maternal and child health services, and basic tests [10]. Among the basic supplies that CHPSs are required to have are diagnostic sets, sterilizers, thermo- and glucometers, dressing trolleys, at least one adult hospital bed and stretchers, as well as adequate illumination at all times [11]. Other primary level facilities include health centers, as well as poly- and small clinics. Additional service and equipment requirements for these facilities compared to that of CHPSs include minor surgical procedures, a basic laboratory, full resuscitating equipment, as well as delivery trays and beds [12, 13]. At the secondary level we find district and regional hospitals, whereas emerging and established teaching hospitals belong to the tertiary level [9, 10]. The equipment, services and staff requirements for these last levels are the highest within Ghana's health system and include ultrasound, x-ray, MRI, mammography, electroencephalogram, electrocardiogram, and computerized tomography services [14, 15]. Moreover, they are required to offer extensive maternal and child health services, and have fully equipped eye care, ear, nose & throat (ENT), dental and mental health departments, not to mention the appropriate staff to render all services in the fields of specialization [14, 15].

Zoom In Zoom Out Reset image size

The type and scope of services that should be provided by any health facility in Ghana depend on their respective level within the country's healthcare system. As the level of care increases, so does the need for more personnel and equipment, which in turn increases energy requirements. However, according to a 2013 review of data on electricity access from 11 countries in sub-Saharan Africa [17], 26% of HCFs reported no access to electricity on average and only 28% claimed to have reliable access. The Republic of Ghana was one of the countries included in the review, with 31% of its 428 surveyed facilities reporting no access to electricity, at least as of 2002. With the arrival of the COVID-19 pandemic, an additional strain fell on many of Ghana's already under-supplied facilities and on a weakened healthcare system that many healthcare workers considered inadequately prepared for a pandemic response [18]. The lack of reliable, stable and sustainable electricity in many Ghanaian HCFs means that not all of the listed services can be provided without disruption. This presents both a challenge for the provision of universal healthcare, which the Ghanaian government has itself declared as a national priority [19, 20]. But it also represents an opportunity in the advancement towards sustainable electrification, and towards the appropriate utilisation of Ghana's rich renewable (solar) resources to fill the gap in energy provision to power health [21, 22]. Such developments strongly enhance the relevance and importance of this study for sustainable development in the region.

According to the International Energy Agency, the World Bank and others [23, 24], Ghana has increased its electricity provision from 42.6% in 1998 to 83.5% in 2019. However, there is still a noticeable urban-rural divide in this regard. As of 2020, around 95% of Ghana's urban population was reported to be electrified, compared to 74% of the rural population [25, 26]. That is to say that many rural settlements are still located in weak- or off-grid areas [24]. Even if connected to the grid, settlements are not guaranteed to have a stable supply of electricity, as transmission and distribution networks in the country are inefficient, with losses of up to 25% [27]. The Ghana Social Development Outlook (GSDO) [28] recognized that unreliable electricity, inadequate clean energy technologies, and high prices are the main constraints to the country's sustainable development.

Modern energy provision plays a crucial role in building and enabling access to a resilient healthcare system [29]. After all, no modern hospital can properly operate without a stable and safe electricity supply [30]. Many rural HCFs in Ghana are located in weak- and off-grid areas, which means that many of the appliances they need cannot be properly operated due to lack of access to reliable electricity. Although other factors that challenge universal access to healthcare are not to be underestimated (e.g. weak supply chains, financial problems, lack of personnel, lack and cost of transportation between settlements and HCFs), constant electricity supply can be considered a paramount basic requirement which underpins some of the other challenges, regardless of the type or size of HCF [31]. Properly electrified HCFs, for example, could attract and maintain skilled health workers, which would in turn help combat lack of qualified personnel, especially in rural areas [29]. Electrified areas in Nepal have been found to host a higher number of health workers, with an average of 11 per 10 000 inhabitants, compared to 2 per 10 000 people in non-electrified areas [32]. Additionally, case study research in Uganda and Ghana have found that access to energy is associated with improved access to communication technologies, enhanced health worker motivation and increased community satisfaction [33].

Lack of power in the health system becomes a hindrance for the provision of even the most basic equipment and medical services in Ghana and other developing countries: equipment sterilization, child delivery, emergency services, or the above mentioned refrigeration of vaccines and other medicines are just a few examples of how crucial electricity is for a fully functioning HCF [29]. At the same time, however, while improving access to energy in health facilities is absolutely essential, it is certainly not sufficient to strengthen health systems in countries such as Ghana and Uganda [33]. Persistent barriers to the availability of health services—such as drug stock-outs, lack of transport and poor amenities—are primarily health system challenges, even where access to energy is available. With the arrival of the COVID-19 pandemic, an additional strain fell on many of Ghana's already under-supplied facilities and on a weakened healthcare system that many healthcare workers considered inadequately prepared against a pandemic response [18]. With the help of georeferenced data, this study helps identifying those vulnerable HCFs, which at the same time are the most suitable for electrification. It must be added that this study understands electrification as the process of replacing technologies that use fossil fuels (coal, oil, and natural gas) with technologies that use electricity as a source of energy [34], and that the energy in question comes from a renewable source.

It is already beyond doubt that renewable energy technologies are by far a more environmentally friendly and cost-effective alternative to non-renewable sources for electricity generation [35, 36]. With more than half of Ghana's electricity production in 2021 coming from fossil fuels [37], the country has still a considerably long journey towards a sustainable energy transition. All the more reason to consider exclusively renewable energy sources for the electrification of all infrastructure which is still located in weak- or off-grid areas, such as rural HCFs. If we consider Ghana's enormous potential for the generation of solar energy, the capacity of the country to augment its sustainable and reliable energy supply, especially in its health sector, is undoubted [38]. Comparable research on African countries, including Ghana [39], draw similar conclusions, however, aside from top-down approaches there is still a room for more detailed country specific assessments of the most appropriate electricity supply option and system design for health facilities, in particular with the perspective of the isolated health facility supply in comparison to a holistic supply including the surrounding communities.

Given the overarching goal to understand how energy self-sufficiency in HCFs can be increased, it is first necessary to understand the status quo, to locate all the relevant infrastructures and explore the spatial relationships between them. This section provides a guide through the collection, validation, analysis and selection of the spatial data used in this project (figure 3).

Zoom In Zoom Out Reset image size

3.1.1. Available datasets

3.1.2. Data processing, sample design for analysis

Define and extract HCFs to be analysed: The VIIRS data, which provide yearly composites of atmospherically corrected nightlights, has a spatial resolution of about 450 m. This introduces a good proxy, not only for the electrification status of an HCF at a specific point, but also of its surrounding area. Each pixel has an intensity value, also known as digital number (DN). To enhance the interpretability of our 16 bit nightlights image, we decided on a threshold so that the given DNs ranged between 0–100 and rule out potential outliers. After this we converted the image to 8 bit so that the DN range took values from 0–255. A cut-off point separating what we considered electrified from non-electrified areas was decided at a DN value of 50. The decision for this separation resulted from a manual cross-referencing of the dataset compared to daylight satellite imagery with recognizable extents of population settlements. In this comparison, urban centres with higher building and settlement densities showed, unsurprisingly, higher DN values compared to less densely populated rural areas. Using this value, a binary mask of the potentially electrified vs. not electrified areas for all of Ghana was created.

We subsequently used the QGIS NN Join plug-in to observe the minimum distance at which each of the HCFs are separated from a nightlight source or an electrified zone. Based on this distance, we selected a threshold for the selection of HCFs considered to be non-electrified or located in weak-grid areas.

Define and extract surrounding settlements: After HCFs were selected, we created a 1 km buffer surrounding each of them (figure 4). Subsequently, all settlements intersecting were also selected for energy demand and system simulation. However, the relative proximity between some of the selected HCFs meant that their respective catchment areas intersected each other, and thus their corresponding settlements did too. In these cases, we used voronoi polygons to separate households according to their closest proximity to an HCF (figure 5).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Solar-mini-grid systems must be simulated and optimized to supply all identified HCFs and their surrounding settlements. This section gives an overview of applied methods and tools to conduct the energy system modelling and evaluation. The overall methodological procedure of the modelling is illustrated in figure 6.

Zoom In Zoom Out Reset image size

3.2.1. Tool

3.2.2. Input data

Offgridders requires, similar to other energy system modelling tools, (i) component specific inputs (such as initial investment, operation and maintenance cost or efficiency and lifetime), (ii) economic parameters (such as interest rates and fuel prices), (iii) general characteristics (project location and lifetime) as well as (iv) renewable energy potential and (v) demand profiles for each site. These inputs are summarized in a tabular input file and stored in the modelling folder structure. A literature review serves as the basis to identify component-specific and economic input parameters for the Ghanaian context. The project lifetime is set to 25 years as commonly applied for mini-grid projects and the project locations are resulting from the previously described geospatial analysis. As the focus of this paper is the simulation and optimization of solar-mini-grids, only the identification and provision of suitable solar radiation data is necessary, no wind data is required. Assessing the electricity demand of all health facilities and their surrounding communities is complex and described in more detail in the following two sections. Its methodological approach is depicted in figure 7.

Zoom In Zoom Out Reset image size

3.2.2.1. Demand estimation—health facilities

In order to estimate the demand for electricity or electric loads of analysed health facilities, we aligned our assessment with an existing categorisation of HCFs in Ghana. The four different HCF categories that are further assessed are:

• Community-based Health Planning and Services (CHPS),
• Health centre,
• Maternity Home & Reproductive and Child Health (RCH), and
• Clinic.

The level of offered services and respective electrical appliances within these HCFs varies and therefore the demand is assessed for all four HCF-types separately. First, a list of all appliances is compiled based on the governmental Health Facilities Regulatory Agency (HeFRA) requirements for each of these HCFs [47]. The appliance list is complemented by an overview of typical operating hours and power ratings for different health facility types provided by the HOMER Powering Health tool [48]. To create an hourly demand profile for each HCF-type, this appliance list serves as the input to the RAMP tool. RAMP is a bottom-up stochastic model for the generation of load profiles in contexts where only rough information about users' behaviour are obtainable [49].

3.2.2.2. Demand estimation—households

In order to consider and evaluate the effects of integrating the surrounding communities into the energy system modelling and design of the health facilities, the load profiles of all settlements within the catchment area of 1 km have to be determined. Within the underlying research project, no load measurements in settlements are conducted. The load profiles are therefore determined via the following approach:

• Literature review is used to identify typical average household demand figures for remote and semi-urban areas (off-grid and weak-grid areas) and respective load profiles for weekdays and weekends.
• Energy Sector Development Programme (ESMAP) Multi-Tier Framework for Energy Access (MTF) is applied to categorize the average daily load profiles into different TIER levels in order to depict representative low, medium and high demand household categories.
• Demographic and Health Surveys (DHS) data is used to link the household demand categories (low, medium, high) to Ghanaian regions and determine the respective percentage of demand categories for each site. The regional wealth distribution of Ghana is shown in figure 8. The ten regions shown in figure 8 differ from the current 16 political regions. This is due to the fact that a referendum was held at the end of 2018 in which four of the old regions were split into new regions, but the DHS data was recorded before and during the implementation phase and therefore only shows the old regions. However, since there were no general changes to the regional boundaries, the asset distributions of the split regions can be easily transferred to the new regions. The political regions before and after the referendum are illustrated again in table 1.

Table 1. Political regions of Ghana before and after the 2018 referendum [50].

Regions until 2018	Regions today
Ashanti	Ashanti
Brong Ahafo	Bono + Bono East + Ahafo
Central	Central
Eastern	Eastern
Greater Accra	Greater Accra
Northern	Northern + Savannah + North East
Upper East	Upper East
Upper West	Upper West
Volta	Volta + Oti
Western	Western + North Western

Zoom In Zoom Out Reset image size

With this, the acquisition of all required input parameter is concluded.

3.2.3. Scenarios and sensitivity analysis

Different energy system configurations are evaluated to create a holistic understanding of the solar market potential within the health sector of Ghana:

• Solar-diesel-battery system setup (SDB setup), and
• Diesel system setup (D setup).

To get a better understanding of the impact of the integration of surrounding communities to the mini-grid of health facilities, two general scenarios are distinguished and calculated based on the demand created by:

• Health facilities (H demand), and
• Health facilities and their surrounding communities (H-C demand).

Sensitivity with regard to allowed supply shortages of 0%, 5%, and 10% are also calculated and further analysed in order to understand the impact of small shortage allowances on the reductions of system capacities and investment costs. These shortage allowances are calculated only for the H-C demand scenarios (health facility plus surrounding community) because, unlike health facilities, surrounding communities are not considered as critical demands that must be always met. This criterion balances the influence of a reduction in investment costs at the expense of not supplying 100% of demand, and is directly connected to the value for the maximum shortage specified for the optimization.

Figure 9 provides an overview of calculated system configurations (D setup, SDB setup), demand scenarios (H demand and H-C demand) as well as sensitivity cases (0%, 5% and 10% shortage allowances).

Zoom In Zoom Out Reset image size

With this, the methodological approach of this research is fully outlined and explained.

The final number of geo-referenced HCFs within Ghana was estimated to be 2614. Our nearest neighbor analysis in QGIS revealed that the vast majority of these (approx. 74%) lie either within or not further than 1.1 km from an electrified zone. 93.9% do not lie further than 11 km from an electrified area. This paper focuses on those HCFs lying beyond a 10 km distance from a night light source. The reason for this is that HCFs located beyond this threshold are less likely to be connected to the grid than those in immediate vicinity, or are an area affected by frequent power outages, a so-called weak-grid. A total of 188 HCF sites (approx. 7.2% of total) were selected for further investigation regarding PV-based electrification. Four of these were removed from the selected dataset since they were either of a type not considered to provide critical services (e.g. training institutions and nutrition centers) or located in immediate vicinity of another selected HCF. Therefore, the final number of HCFs selected for this study amounts to 184, each of which belongs to either one of the following types: CHPS, Clinic, Health Centre, or Maternity Home & RCH.

To have a better idea of the distribution of all selected HCFs, a heatmap with a quartic kernel and a radius of 50 km was created. Given the low number of HCF dispersed throughout the vast territory of Ghana, a relatively large radius had to be chosen to properly visualise the density distribution from the selected HCF dataset. The heatmap reveals that most of the selected HCF are located in the regions of Central, Oti (previously Volta), Western, Western North, Upper East and Upper West. Indeed, all of these regions, with the exception of Central, had the lowest assumed electricity access rate in the whole country, according to the Ghanaian Ministry of Energy in 2009 [51] (figure 10).

Zoom In Zoom Out Reset image size

Besides the HCFs a total of 655 settlements were identified to meet the set criteria. From these, 405 settlements belonged to the hamlet category, followed in number by the SSA type, with 243 polygons. Lastly, 7 settlements belonged to the BUA type. Not all HCFs had an equal amount of settlements assigned to them, some had a one to one, whereas others a one to many HCF to settlement ratio (figure 11).

Zoom In Zoom Out Reset image size

4.2.1. Summary and analysis of input data

In this section, all relevant input parameters for the energy system modelling and analysis are summarised.

4.2.1.1. Component-specific and economic parameter

Component-specific and economic input parameters for the Ghanaian context are acquired by literature review and summarized in the input template in the attachment including their respective sources (see appendix A).

4.2.1.2. Solar data

According to a paper published within the related research project (Enershelf), SARAH-2 data is the most accurate solar radiation dataset for our study area [52]. However, a detailed analysis of the solar data for all identified project sites revealed inconsistencies. Neighboring project sites (20 m apart) in flat topographies exhibited differences in radiance by a factor of 6. Therefore we used a data set without gaps and with consistent data for all our project sites: the PVGIS-Era-5 dataset provided by the EU Science Hub [53].

4.2.1.3. Health facility demand

As described above, the demand profiles for the four types of healthcare facilities analysed (CHPS, maternity home, health center, and clinic) are modeled stochastically using RAMP. RAMP provides minimum, maximum, and average load profiles as output. Figure 12 shows exemplarily the weekly minimum, maximum and mean load profiles for one health center as calculated by the RAMP tool.

Zoom In Zoom Out Reset image size

The CHPSs have an average hourly demand of 1.3 kW, health centers of 1.6 kW, maternity homes & RCHs of 1.8 kW and clinics of 1.9 kWh. The accumulated average daily demand is for CHPSs 30.1 kWh, for health centers 38.8 kWh, for maternity homes & RCHs 43.3 kWh and for clinics 46.1 kWh.

Since the objective of this work is to model the energy system for HCFs, their specific characteristics must be considered. As mentioned earlier, HCFs usually have high but short demand peaks. These peaks are considered as one of the bottlenecks in energy system modelling. RAMP first models the demand in a minute resolution for one year and afterwards resamples the profile to an hourly resolution. This needs to be done, because Offgridders needs solar- and demand time-series in an hourly resolution as an input file. Therefore, we decided to use the maximum load profile modeled by RAMP in this analysis. Figure 13 shows the representative maximum demand profiles for each health facility type.

Zoom In Zoom Out Reset image size

4.2.1.4. Household demand

To estimate the demand of surrounding communities, the representative average household demands are acquired based on literature review:

• Taale and Kyeremeh (2019) analysed drivers of household's electricity expenditure in Ghana based on data provided by the 2005/2006 and 2012/2013 waves of the Ghana Living Standards Survey that were designed to provide nationally and regionally representative indicators. Based on this study, the average daily household demand in Ghana is 1.58 kWh [54].
• The World Bank is frequently providing average household electricity consumption statistics and based on their analysis, the average daily household demand in Ghana is 4.18 kWh in 2014 (most recent data for Ghana) [55].
• Sakah et al (2019) studied appliance ownership and electricity consumption determinants in Ghana based on data from residential electricity consumption survey (RECS) in combination with electricity end-use monitoring of 60 households in Tema city (Ghana). Their analysis shows an average daily household demand of 8.86 kWh [56].

Other identified studies stated daily average household demands within the range of above listed demand values underlining the suitability of selection (e.g. Diawuo et al [57]). The range of daily household demand accounts for different level and variability of electricity usage within the households. To understand their respective level of electrification, these average daily demand values are linked to the MTF in table 2. The household demands listed above can be classified as TIER 3, 4 and 5 electricity access allowing from up to medium to very high-load appliances. In our analysis we refer to low (TIER 3), medium (TIER 4) and high (TIER 5) demand households and apply these three representative average daily household demands to the energy system modelling. The distribution of load over a day was adopted from Sakah et al (2019) and applied to the three different TIER levels [56].

Table 2. Classification of household consumption in the Multi-TIER system [58].

TIER	Power capacity ratings (at least) kWh−1 d−1 HH−1	Allows for	Representative household demand kWh d−1 HH−1	Source
3	1	up to medium-load appliances	1.58	[54]
4	3.4	up to high-load appliances	4.18	[55]
5	8.2	up to very high-load appliances	8.86	[56]

The variety of daily household demand reveals a non-homogeneous electricity demand distribution amongst all households in Ghana. This is proven by the Energy Commission's report 'Energy Profile of Districts in Ghana' where residential electricity consumption is listed for different districts of Ghana. Low consumption rates are seen in the Savannah or Northern region, while very high consumption patterns are observed in or surrounding urban areas [59]. To reflect this fact in our analysis, we apply different shares of low, medium and high demand households according to their location (region). As detailed information on percentages of low, medium and high demand households within the regions is missing, we determine the shares according to the wealth index in each region. Bridge et al (2016) prove that wealth and electricity consumption are directly linked to each other: the wealthier a household, the more electricity is consumed [60]. The wealth distribution index for Ghanaian regions is taken from The Demographic and Health Surveys (DHS) Program for the most recent year (2019). Table 3 is listing the wealth index distributed in low, medium and high per region as percentage of population.

Table 3. Wealth index distribution across Ghana's population [61].

Region	Wealth index low % of population	Wealth index middle % of population	Wealth index high % of population
Ashanti	20.89	36.35	42.76
Brong Ahafo	45.09	40.03	14.88
Central	24.06	51.31	24.62
Eastern	15.38	43.36	41.26
Greater Accra	1.54	19.73	78.72
Northern	79.38	16.59	4.03
Upper East	82.42	11.66	5.92
Upper West	78.16	14.97	6.87
Volta	42.46	40.62	16.93
Western	22.94	46.10	30.96

The number of households surrounding each health facility site is taken from the geospatial analysis. These households (HHs) are split into the different wealth indices (WI) of their regions—representing their assumed low, medium and high demand behaviour—and multiplied with the three representative daily household demands (TIER 3, 4 and 5) to calculate the respective demand of households surrounding each health facility site (D$_\textrm{HHs}$) according to the function below:

$$\begin{align} \textrm{D}_\textrm{HHs} = \textrm{HHs} * \left(\textrm{WI}_\textrm{low} * \textrm{D}_\textrm{TIER3} + \textrm{WI}_\textrm{medium} * \textrm{D}_\textrm{TIER4} + \textrm{WI}_\textrm{high} * \textrm{D}_\textrm{TIER5}\right) \end{align} \tag{ 1 }$$

Thus, the demand of surrounding households is calculated for each of the 184 health facility locations and added to the demand of the respective health facility (CHPS, maternity home, health center or clinic) to simulate mini-grids supplying both, the health facility and their surrounding communities. Figure 14 shows the weekly load profiles of two representative clinics. The first graph shows the load profile of a clinic surrounded by 1242 households. In this case, the demand peaks of the clinic are covered by the household demand. The lower graph visualises the load profile of a clinic surrounded by 42 households which are not covering the demand peaks of the clinic in their catchment area. For the majority of analysed cases, the base load of surrounding households is covering the demand peaks of HCFs in their catchment area. For 17 out of 48 CHPS, 4 out of 13 maternity homes or RCH, 14 out of 82 health centers and 26 out of 40 clinics, the demand peaks of HCFs are higher than the minimum load expected by the surrounding households. This means that the HCF demand peaks are not lower than the household demand at all times, but it does not account for when these load peaks occur. It could be that these HCF peaks occur in parallel with higher load times of the surrounding households, which then have a balancing effect on the short but high HCF peaks again.

Zoom In Zoom Out Reset image size

In general, balancing the high, but short load spikes of HCFs by including the demand of surrounding households makes sense for cases with low base loads of HCFs and higher numbers of surrounding households. At specific household sizes—low surrounding population—the loads of the HCFs are much higher than the one for the households.

Both electricity demands—for health facility and health facility including their surrounding communities—are calculated as annual profiles to serve as inputs to Offgridders. These files can be found in the appendix C.

4.2.2. Energy system scenarios and sensitivity analysis

The following section presents the results of the energy system modelling and optimisation. The optimal energy system setup for all HCFs including and excluding the demand of their surrounding communities is discussed, followed by the allowable electricity shortages and their impacts.

For all 184 HCFs, the solar, battery, diesel energy system setup shows lower levelised cost of electricity (LCOE) than the diesel setup as shown in table 4. LCOEs represent the minimum electricity price that must be paid to amortize the energy system over the life of the project. Real electricity prices may vary depending on governmental subsidies.

Table 4. Minium and maxium LCOEs for of all simulated and optimised 184 HCF sites and their surrounding communities.

HCF only
Energy system setup	Min. LCOE	Max. LCOE
	(USD kWh−1)	(USD kWh−1)
Solar, battery, diesel	0.30	0.45
Diesel	0.94	1.05
HCF & surrounding communities
Energy system setup	Min. LCOE	Max. LCOE
	(USD kWh−1)	(USD kWh−1)
Solar, battery, diesel	0.23	0.31
Diesel	0.83	0.90

Looking at the HCF only, the range for the PV, battery, diesel system setup has a range between 0.3 USD kWh⁻¹ and 0.45 USD kWh⁻¹. LCOEs for the diesel setup are much higher and within a range of 0.94 USD kWh⁻¹ and 1.05 USD kWh⁻¹. If the surrounding communities and their demand are included into the energy system modelling and optimisation, even lower LCOE (0.23 USD kWh⁻¹) can be achieved for the hybrid system setup. Both, the minimum and maximum values for the diesel setup are lower for the HCF plus community scenario than the HCF only scenario. The higher base load of the combined demand may be a reason for this. Therefore, the diesel generator can run more time in its optimal range (full load hours).

Figure 15 shows the LCOEs for all calculated scenarios (energy system setup), demands (HCF only and HCF plus community), and sensitivity analyses (shortages allowed). The blue circles represent the diesel only system setup, the green ones the hybrid energy system setup.

Zoom In Zoom Out Reset image size

The diagram again illustrates that the diesel energy system setups leads to almost three times as high LCOE as the hybrid system. Again, for both energy system configurations, HCF only demand indicates the highest LCOE. When looking at the shortage allowances, it is becomes obvious that the higher the allowed shortages, the lower the LCOEs for both energy system configurations. In addition, the difference in LCOE between no shortage (0%) and 5% shortage allowance is more significant than the difference between 5% and 10% shortage allowance. Minimal and maximal LCOEs dropped from 0.23 USD kWh⁻¹ and 0.31 USD kWh⁻¹ to 0.21 USD kWh⁻¹ and 0.28 USD kWh⁻¹ for a shortage allowance of 5% of the total household demand and 0.20 USD kWh⁻¹ and 0.28 USD kWh⁻¹ for a shortage allowance of 10% of the total household demand.

Figure 16 is taking a closer look at the LCOEs of the solar, battery, diesel system setup. The range of LCOE for different locations is more evident in the HCF only demand simulations. Further analysis of the data reveals that CHPSs have on average the highest LCOEs of around 0.43 USD kWh⁻¹, followed by health centers as well as maternity homes and reproductive child health institutions with a LCOE of around 0.37 USD kWh⁻¹. According to our analysis, clinic have the lowest average LCOE of around 0.33 USD kWh⁻¹. This fact is also visible in figure 16. LCOEs for hospital-only demand sites appear to be at three different cost levels. This is explained by the demand structures of the four different HCF types. Clinics have the highest demand. Thus, general investment cost are distributed over a higher share of electricity sales (higher demand).

Zoom In Zoom Out Reset image size

Figure 17 visualises the first investment necessary to install an energy system on the 184 simulated sites. The first investment only considers project and system component cost at the beginning of the project. The different scales of both diagrams reveal a significantly higher first investment cost for the PV-diesel-storage mini-grids. This is explained by higher system component cost for PV and especially storage technology when compared to diesel generators. Sites with very high numbers of surrounding households (e.g. site No. 64 'Ekye Health Centre' with 5678 surrounding households or site No. 58 'Damanko Health Centre' with 2896 surrounding households) create high first investment cost outliers. These energy systems need to supply the HCF including a high number of households resulting in high installed capacities and increased cost when compared to sites with less surrounding households. However, the above analysis of LCOEs shows that hybrid systems pay off in the long run, as electricity costs are significantly lower than for diesel setups.

Zoom In Zoom Out Reset image size

The net present value (NPV) does not only reflect the first investment necessary, but also includes future operation, maintenance and replacement cost to operate the energy system over a project duration of 25 years and is shown for all simulated cases in figure 18. The NPV gives an indication on the necessary cash flow to maintain and operate the system. The NPV divided by the electricity produced by the energy system (demand) equals the LCOE. In contrast to first investments, the NPV is significantly lower for the hybrid system setup than for the diesel setup. This is explained by the high diesel fuel cost over project duration. The amount of diesel fuel consumed for the diesel only setups has a range between 13 751 l/a and 3668 274 l/a compared to the hybrid system setup of 387 l/a and 96 512 l/a (0% shortage and HCF plus surrounding community scenario). If evaluated over the whole project duration, the NPV consolidates that hybrid system setups are favorable for the electrification of HCFs and their surrounding communities.

Zoom In Zoom Out Reset image size

In addition to the long term economic benefits of renewable energy (RE) hybrid system setups compared to 100% diesel systems, the higher independence of fuel supply and the reduction of CO₂ emissions through substituting diesel supply with RE supply are other important benefits. The difference in annual diesel fuel consumption for both system setups is significant (reduction of approximately 97%) as outlined above. Figure 19 shows the RE share for all simulated cases. In case of 0% shortage allowance and supplying the HCFs including surrounding households, the minimum RE share is 97% and the maximum 99%. The graph also shows the tendency that increased shortage allowances lead to an increased RE share. If 10% shortage is allowed, the RE share is even at 100%. This can be explained by the fact that the diesel generator in a hybrid system is mainly used as a backup supply due to its high operating costs. When shortages are allowed in times of high demand, this backup by a diesel generator is redundant, which leads to the high RE share.

Zoom In Zoom Out Reset image size

Finally, table 5 summarizes the market potential to reliably supply the analysed HCFs and their surrounding households with electricity via PV-mini-grids. The sum of the first investment, NPV and PV capacity necessary to implement PV-battery-diesel-mini-grids for all 184 sites is calculated and presented. The figures underscore that most of the investment and PV capacity comes from supplying surrounding households, if they are taken into account. It becomes obvious that there is a large market potential for the electrification of HCFs situated in off- and weak-grid areas of Ghana with PV-based mini-grids.

Table 5. Country-wide market potential for PV-mini-grids for HCF in Ghana.

Parameter	Unit	HCF only	HCF & surrounding households
First investment	(Mio. USD)	13.2	243.8
Net present value	(Mio. USD)	17.0	393.2
PV capacity	(MWp)	4.5	152.8

In summary, three key conclusions can be drawn from the results: First, PV-battery-diesel mini-grids are more economical than purely diesel-based electricity provision. Second, taking surrounding households into account in the design of these mini-grids has several advantages, as they balance high but short peak demand peaks of HCFs, making the implementation and operation of these systems more feasible. Third, if supply shortages to surrounding households are accommodated, investment costs and thus LCOE can be massively reduced. To make this results more widely available to various stakeholders they have been published online on a user-friendly, interactive webmap ¹ .

Research Question 1: Which HCFs are most suitable to be electrified with PV mini-grids in order to increase their energy self-sufficiency and reliability of services provided?

• Which HCFs are likely to have no or poor access to electricity supply?Our analysis suggests that a total of 184 HCF sites are the most suitable to be electrified with PV mini-grids in order to increase the reliability and self-sufficiency of the services they provide. This result takes mainly into account the distance between each of the 2614 HCFs in Ghana and what we considered an electrified zone thanks to nightlight imagery. Those HCFs which lied beyond 10 km from a nightlight source were deemed to have no or poor access to electricity supply and thus taken into our selection for further analysis.
• How can these HCFs be categorised?Each of these facilities belongs to either one of the following types: CHPS, Clinic, Health Center, or Maternity Home & RCH. Initially, two more facility types were included in the selection (training institutions and nutrition centers), but were removed since they were not considered to provide critical services or were located in immediate vicinity of another selected HCF. All of the selected HCF types belong to the primary level of care, according to Ghana's healthcare system.
• Which settlements are within the catchment area of these HCFs?A radius of 1 km defined the catchment area for each of the selected HCFs. A total of 655 settlements intersected with this radii across Ghana's territory and were thus also selected to be part of the electricity system simulation section of this paper. The settlements in question belong to all agglomeration classes (BUAs, SSAs, and hamlets) in the GRID3 data used in this study [62]. It was to some extent surprising to find BUA-type settlements in this selection, given the assumption that they have a comparatively developed infrastructure and are thus expected to have better access to the electricity grid than the other two types. Finally, not all HCFs had an equal amount of settlements assigned to them, some had a one to one, whereas others a one to many HCF to settlement ratio (figure 11).

In summary: Out of a total of 2614 geo-referenced HCFs in Ghana, 184 are estimated to be most suitable for electrification with PV mini-grids in order to increase their energy self-sufficiency and the reliability of the services provided. A study by Moner-Girona et al [39] estimated a similar total number of HCFs for Ghana. However, on closer inspection, and given the use of different sources in both papers, there is little overlap between datasets in their geo-referenced HCF vector points. In addition, Moner-Girona et al's [39] cover a much larger area for their study—at a continental level—so differences in accuracy are to be expected compared to a much smaller area of analysis such as ours.

The selection of these HCFs was based on their distance from an electrified zone, with those beyond 10 km from a nightlight source deemed to have no or poor access to electricity supply. The selected HCFs belong to four types of HCFs: CHPS, Clinic, Health Center, or Maternity Home & RCH, all of which belong to the primary level of care in Ghana's healthcare system. The catchment area for each of the selected HCFs was defined as a radius of 1 km, and a total of 655 settlements intersected with this radius across Ghana's territory and were thus also selected to be part of the electricity system simulation. These settlements belong to all agglomeration classes (BUAs, SSAs and hamlets) in the GRID3 data used in the study. Moner-Girona et al [39] estimate a total of 646 HCFs with a high probability of not having access to reliable electricity in Ghana. However, their methodology differs from ours, including e.g. a 5 km buffer around an electrification infrastructure which is also incompatible with our electricity infrastructure dataset.

Research Question 2: Are PV mini-grids an appropriate option to increase energy self-sufficiency at these HCFs, and do the surrounding communities play a role to stabilise cost and electricity supply?

• What is the electricity demand for identified HCF sites and their surrounding communities?As stated above, our analysis included four different types of HCF which are situated in off- and weak grid areas, namely CHPS, maternity homes & RCH, health centers and clinics. These different HCF institutions have different electricity demands depending on the level of services they provide: The CHPSs have an average hourly demand of 1.3 kW, health centers of 1.6 kW, maternity homes & RCHs of 1.8 kW and clinics of 1.9 kWh. The accumulated average daily demand is for CHPSs 30.1 kWh, for health centers 38.8 kWh, for maternity homes & RCHs 43.3 kWh and for clinics 46.1 kWh. The HCFs have high, but short demand peaks due to the necessary operation of some electricity-intensive appliances such as x-ray machines. Thus, maximum values of modeled demands have to be taken into account for preprocessing the data as inputs for optimising and simulating the energy systems.
• Which mini-grid system setup is most suitable to electrify identified HCF sites?Our analysis showed that RE hybrid mini-grids are the most promising solution to electrify HCF that are located in off- and weak-grid areas. Over a project duration of 25 years, lower LCOEs and NPVs showed that they are economically more feasible than diesel-system setups. In addition, they come with the benefit of less diesel fuel needs and thus creating a higher independence in operating the energy systems as well as causing less CO₂-emissions.
• What is the impact of including the demand of surrounding communities into the energy system modelling?The above mentioned findings apply both to the HCF-only scenarios and to the case where demand from surrounding households is considered in the optimisation and simulation of the energy systems. In addition, the inclusion of the surrounding households helps to lower the LCOE of the optimised PV-battery-diesel mini-grids and in most analysed cases as well the household demand is balancing the high but short demand peaks of HCFs. The ratio between HCF demand and the demand of surrounding households depends heavily on (i) the type of HCF and (ii) the number of households situated in a catchment area of the HCFs.
• How does accounting for allowed electricity shortages from surrounding communities affect the mini-grid economics when included in energy system modelling?Electricity shortages are allowed for the combined demand of surrounding households and HCFs and are calculated as sensitivity cases in the energy system modelling of the HCFs and surrounding household scenario. When looking at the most favorable energy system setup (PV-battery-diesel mini-grids), results showed a significant decrease in LCOE when shortage is allowed: Minimal and maximal LCOEs dropped from 0.23 USD kWh⁻¹ and 0.31 USD kWh⁻¹ to 0.21 USD kWh⁻¹ and 0.28 USD kWh⁻¹ for a shortage allowance of 5% of the total household demand and 0.20 USD kWh⁻¹ and 0.28 USD kWh⁻¹ for a shortage allowance of 10% of the total household demand. The difference in LCOE is much higher between the no-shortage and 5% shortage sensitivity cases than between both shortage allowance sensitivity cases (5% and 10%). The diesel generators usually supply the demand in times of high loads or low PV yield. If shortages are allowed, the diesel generator becomes redundant in many cases as the overall RE share of the PV-battery-diesel mini-grids simulated for this study are already at 97% or more. Thus, the RE share is 100% for both shortage allowance sensitivity cases and all simulated sites. It becomes obvious that it is more important to allow for shortages at all than to allow for higher degrees of shortages if LCOEs and NPVs shall be lowered and RE shares shall be increased.
• What is the accumulated solar mini-grid potential to electrify all identified HCF sites?To electrify all identified HCFs with electricity via PV-battery-diesel mini-grids, an accumulated PV capacity of 4.5 MWp, first investment of 13.2 Mio. USD and NPV of 17 Mio. USD is necessary. If the surrounding households are taken into account for the energy system modelling, the necessary PV capacity is 152.8 MWp, first investment is 243.8 Mio. USD and NPV is 393.2 Mio. USD.

According to our analysis, RE hybrid mini-grids are the most promising solution to electrify HCFs located in off- and weak-grid areas. They are economically feasible, have less CO2₂-emissions, and including surrounding households helps lower LCOEs and balance demand peaks. Similar results have been obtained in other studies for the region [39, 63]. In the former study, the LCOEs of PV systems (including battery storage) installed in areas without access to electricity in sub-Saharan Africa averaged 0.4 EUR kWh⁻¹ (0.39 EUR kWh⁻¹ for Ghana). According to the authors, these values are lower than those reported for mini-grid systems in Africa before 2021 [39]. In the latter, a life cycle assessment for rural electrification in 15 ECOWAS countries concluded that 100% PV and hybrid PV-diesel mini-grids are the most attractive technologies in terms of lowest environmental impact compared to both diesel mini-grids and the region's national grid electricity [64]. Our findings are further corroborated by case studies such as Olatomiwa et al [63], where most PV/wind/diesel hybrid system configurations performed better, polluted less, and were more cost-effective compared to diesel-only systems for the electrification of rural HCFs in six geopolitical zones of Nigeria.

Allowing for shortages is more important than the degree of shortages. To electrify all identified HCFs with electricity via PV-battery-diesel mini-grids, an accumulated PV capacity of 4.5 MWp, first investment of 13.2 Mio. USD and NPV of 17 Mio. USD is necessary. If the surrounding households are taken into account for the energy system modelling, the necessary PV capacity is 152.8 MWp, first investment is 243.8 Mio. USD and NPV is 393.2 Mio. USD.

In summary, the results show that a large number of HCFs and their surrounding communities in Ghana are likely to face challenges of poor and unreliable electricity supply. While diesel generators are known to be a quick fix to fill this supply gap, our analysis highlights the long term economic and environmental benefits of providing electricity through PV-battery diesel mini-grids, especially with the integration of surrounding communities into the mini-grids. However, the financing of these systems is another important challenge that needs to be addressed. While Ghana's HCFs are either state-owned or operated by private companies that usually have the financial means to pay for their electricity supply, some households in the surrounding communities might face an affordability issue. The calculated LCOE for PV-battery diesel mini-grids (between 0.23 USD kWh⁻¹ and 31 USD kWh⁻¹ if no shortage is allowed and 0.20 USD kWh⁻¹ and 0.28 USD kWh⁻¹ if shortage is allowed) and are significantly higher than the current electricity tariffs in Ghana (e.g. 0.04 USD kWh⁻¹ for lifeline customers and on average 0.9 USD kWh⁻¹ for residential customers—depending on their demand structure [65]). According to a study by Twerefou (2014), residential customers were willing to pay 1.5 times higher prices if the electricity supply was reliable [66]. The quality of service that can be improved by properly designed PV battery diesel mini-grids compared to the current unreliable grid supply (where applicable) is therefore a key willingness to pay criterion for Ghanaian households. In addition, another study conducted extensive willingness-to-pay surveys in neighbouring rural Nigeria and found that households expressed a preference for the mini-grid or solar option over the commonly used diesel, with an increased willingness to pay for these options [67]. As the Ghanaian government is committed to the electrification of its people, one solution to bridge the still existing gap between people's willingness to pay and the LCOE is to apply the existing public subsidies for grid electricity to the PV-battery diesel mini-grids for HCFs and their surrounding communities. In addition, studies should be carried out to assess the situation in more detail for each specific community, and connection to the mini-grid should always be optional for each individual household.

When implementing mini-grids, a key question is always what happens when the main grid arrives, as government electrification plans usually include the possibility of grid extension. However, grid extension and mini-grid implementation are not inherently competing concepts, especially in countries with weak and unreliable grid supply. As noted in the introduction to this paper, the unreliability of the grid in Ghana is a major threat to the smooth operation and delivery of all health services. For this reason, HCFs in Ghana usually require on-site backup generation. In many cases, this is provided by backup diesel generators. But there is also evidence that off-grid HCFs are covering their needs with their own (grid-integrated) PV mini-grids to create greater independence from unreliable grid supply [68]. Thus, the operation of grid-integrated PV mini-grids for HCFs in Ghana even in on-grid areas is actually a great opportunity to ensure reliable and uninterrupted health services, creating greater independence from grid shortages (blackouts and brownouts) and diesel supply. In addition, these PV battery diesel mini-grids have the potential to feed back into the grid and thus stabilise the grid supply, if appropriate monitoring and control measures are integrated and regulations are formulated.

Our findings create a Ghana-specific overview with manually cross-validated location data of health care facilities (by satellite data) and their surrounding communities. The openly published input and output datasets can inform further research and thereby extend available research findings such as published in the Clean Energy Access tool ² .

The methodology used provides many valuable insights to better understand the challenges and potentials of providing Ghanaian HCFs with reliable electricity via decentralized mini-grids. However, this methodology and its derivations are an initial approach to the topic and there are limitations that need to be discussed.

5.3.1. Geospatial analysis

5.3.2. Energy system modelling

The integration of health and energy sectors in planning and development has been recognized as a promising strategy for improving access to high-quality health services. Solar mini-grids, in particular, are emerging as a reliable and sustainable source of energy that could play a significant role in ensuring a reliable power supply for HCFs. These mini-grids have the potential to address the energy deficit in HCFs and improve energy self-sufficiency, thereby providing better healthcare services. However, it is essential to note that findings from one type of HCF cannot be transferred to another type. Each HCF has its unique characteristics, such as the size, demand, and the types of services provided, which require a context-specific approach in planning and implementation. Thus, the implementation of solar mini-grids in HCFs must be tailored to meet the specific needs of the facility. To balance the energy flow and stabilize the energy system, it is essential to consider both the energy demand of HCFs and the surrounding communities. The combination of these demands presents a promising solution to improve the overall energy system's stability. It is necessary to assess the electricity demand carefully based on context-specific characteristics such as the type of HCF and the number of households considered. This will help to ensure that the implementation of solar mini-grids is done in a manner that is appropriate and efficient.

In conclusion, the integration of health and energy sectors in planning and development has the potential to improve access to high-quality health services. Solar mini-grids have emerged as a reliable and sustainable source of energy that could play a significant role in ensuring reliable power supply for HCFs. However, a context-specific approach is crucial to ensure that each HCF's unique characteristics are taken into consideration. The combination of HCFs' and surrounding communities' energy demands presents a promising solution to balance the energy flow and stabilize the energy system. Finally, careful assessment of the electricity demand based on context-specific characteristics is essential to ensure the efficient implementation of solar mini-grids in HCFs.

Interdisciplinary research has proven to be critical in exploring the health/energy nexus, and this approach must be applied in further analyses. The results of such research should be applicable to other regions and cases of interest, leading to further research opportunities. Energy system modeling for HCFs as critical demand and shortages allowed on household loads presents a promising solution to decrease the needed system capacity and investment costs. This approach can help optimize the use of available energy resources and ensure a reliable energy supply for HCFs. Further research should also be conducted on other nexus topics, such as energy, health, water, and agriculture, to ensure a more holistic and integrated approach. Interdisciplinary research is critical in addressing complex challenges in these areas. As a climate change mitigation strategy, replacing diesel generators in mini-grids with higher storage capacities or green hydrogen technologies is recommended and need to be investigated in detail. This can significantly reduce greenhouse gas emissions and promote sustainable development. Finally, integrating climate resilience considerations into analysis is essential to ensure that the energy system remains stable and reliable in the face of climate change impacts. By considering these recommendations, we can achieve a sustainable and resilient energy system that meets the healthcare needs of communities while addressing global challenges such as climate change.