ELECTRIFICATION PLANNING DECISION SUPPORT TOOL D I Banks, F Mocke, E C Jonck, E Labuschagne, R Eberhard, Rural Area Power Solutions, Mockes Business Consultants, Megasub Software Development, Cape Associates ABSTRACT This paper describes a GIS based model developed to facilitate electrification planning. The model uses demographic and other data from GIS data sets and a score sheet to quantify the ‘assumed benefit’ of electrification of all unelectrified settlements in the target region. The costs of different electrification options (grid, mini-grid and solar home systems) are then derived for each settlement using experience based look-up tables. The system then prioritises projects and technologies, based on the ratio of 'assumed benefit points' and cost. The model is designed to operate as a first pass toolfacilitating long-range strategic level planning for entire regions (including 50 to 2000 settlements). It may therefore assist detailed engineering planning, but does not replace detailed design work. The GIS model is linked to a macro-level financial and economic analysis, which provides regional and national level forecasting of the economic impacts of the micro-level technology and prioritisation decisions. The two systems together comprise a powerful information and scenario analysis tool to assist policy makers and electrification implementation agencies in the process of electrification technology, budgeting and prioritisation decision-making. Abbreviations ADMD Callow Ccont Ckm Cline Cretic Famax or RPP hh Ic Pgrid Pmini PSHS R(km)

After Diversity Maximum Demand Allowable exp. on electrification for community ‘Other’ contributions to elec. resources Cost per km of line extension from existing grid to settlement Cost of the line from the grid to the settlement Cost of local reticulation (excludes Cline) Maximum allowable expenditure per benefit point from national funds Households Iteration cycle Total no. ‘benefit points’ allocated to community assuming that it would be grid electrified As above - assuming mini-grid As above – assuming SHS elec. Max length of grid extension ‘benefit points’ can ‘purchase’.

1. INTRODUCTION Rural electrification typically involves several phases, including: • Identification and prioritisation of development nodes for electrification • Successive stages of settlement identification and preliminary cost estimate development

•

Financial and economic modelling of prospective projects, to assess viability in terms of programme objectives.

In South Africa, where the electrification programme has been very extensive, and heavily cross subsidised, planning decisions have been of crucial importance to communities, as they make the difference between getting no benefits at all from the programme, or receiving an effective per household subsidy of the order of R3000 to R5000 per household. Those that do benefit receive a prepayment meter based system. The connection fee is relatively low (R100 or about 17 US$), there is no minimum monthly charge, and the tariff is 33.7 SA cents (about 5.5 US cents) per kWh. To date, communities that do not receive the grid have not been provided with any alternative energy supply enhancement. Furthermore, accurate information on the probability and timing of possible grid electrification has been difficult to obtain (whether known and held confidential by the national utility Eskom, or simply not yet determined). Several important shifts are currently taking place in the National Electrification Programme [1]: • Responsibility for funding of the non-recoverable portion of the programme is shifting from the national utility Eskom to the national fiscus through the establishment of a National Electrification Fund. • With the shift in funding location, comes a corresponding shift in decision making processes, and the need for supportive planning processes. • There is increasing realisation that grid technologies are too expensive to serve more remote households, and a commitment from government to provide universal access to basic electricity services, using off-grid technologies where necessary. • There are significant efforts underway to get largescale off-grid electrification underway [2]. Given the above elements, there is a critical need in South Africa to develop a comprehensive picture of the projected expansion of the grid network over the next few years. This picture needs to: • Indicate the most appropriate technologies for different regions/settlements (grid or off-grid) utilising an agreed decision making process. • Provide an estimate of the financial costs of the planned programme (informing national and regional budget allocations). • Provide information on the sequencing and prioritisation of projects. • Provide some indication of the macro-level economic implications of the programme.

Paper prepared for the Domestic Use of Energy Conference, Cape Town, April 2000

2. MODEL OVERVIEW The planning tool described here seeks to meet the above objectives, but has an important additional objective. It is designed to be highly flexible, such that a range of planning scenarios can be rapidly produced for a given region. This allows decision makers and planners to rapidly explore the implications of policy level decisions (such as budget allocation rates, increased prioritisation of schools, technology cost changes, etc.) 2.1 REGIONAL PLANNING TOOL The planning tool developed comprises two main elements, a regional planning tool, as illustrated in Figure 1, and a macro level financial and economic analysis tool. The regional planning tool is discussed first, as this generates the information for later analysis using the macro-economic tool. In summary, the model illustrated in figure 1: (1) Determines a weighted benefit of electrification of each settlement. (2) Finds out whether this ‘benefit’ can justify the financial cost of extending the grid (taking into account the need for progressive growth of the grid). (3) If (2) is yes, allocates the settlement to the grid programme, with a priority based on logical grid growth and the ratio of the weighted benefit to cost. (4) If the answer to (2) is no, the decision tool makes a first pass assessment of whether mini-grid1 technologies or stand alone (Solar Home System2) off-grid electrification would be more cost effective. Key elements of the above are discussed in more detail below. (a)

Points scoring as a proxy for socio-economic analysis: There are two approaches which could be modelled to assist in electrification planning project prioritisation and technology choices. In an ideal world, with full access to detailed information, a thorough economic and financial analysis (on a lifecycle basis) can be carried out for each electrification project. Provided that the economic analysis takes good account of most of the factors important to decisions (and provided that it is complemented by a good qualitative report which considers social and environmental issues adequately), this provides a useful method of directly comparing the priority of undertaking different projects. The usual rules are to maximise the net present value (NPV) of the different projects, to achieve a minimum economic rate or return, or to seek a high benefit/cost ration. (See for example Davis and Horvei [3] or Banks [4].)

1

Mini-grid systems use a local isolated grid distribution system to deliver power from a central generating plant (typically solar, wind, diesel, micro-hydro, biomass or a combination) to the customers. 2 Solar Home Systems (SHS) are photovoltaic powered units placed at each household, that generate and store sufficient energy to power lights and other low power appliances such as televisions and radios.

An alternative approach is to utilise a points scoring system. Potential beneficiary’s such as households, schools, health centres etc can be allocated benefit points – with the total number of ‘electrification benefit points’ for a community or project being added together, and then modified by general weighting factors which take account of for example: average income levels, road infrastructure, etc. This approach is easier to apply to automated models. Furthermore, it attractive to committees and planners, as it allows representative committees and experts in different fields to establish a uniform scoring and weighting system, which is relatively transparent and familiar to most decision makers. In order to calculate the ‘electrification benefit score’ for each community within the planning area, the model thus requires data on the different fields of the points scoring table. In South Africa, we are fortunate, in that a number of core datasets exist in GIS format which allow us to determine reasonable ‘electrification benefit scores’ for each settlement. Data availability is however likely to be the key constraint to use of the model for rapid assessments in other regions of Africa. The above described points benefits system was used, in the model to generate the ‘electrification benefit score’ (Pgrid) for each settlement. Factors included in the implemented scoring system are: households (1 point each), schools (range of point values depending on level), clinics (15 points) and police stations. Provision for an income based weighting factor was made, although for the results presented here this was not used. A significantly more comprehensive points table was however derived during the project, and could be utilised as data availability improves [5]. In the case of mini-grid or SHS electrification, modified tables (which account for the lower level of service) could be used to determine off-grid ‘electrification benefit scores’, (PSHS or Pmini). (b) Grid prioritised planning As described by Banks in [6], it seems at first sight appropriate to directly compare costs and benefits of grid and off-grid electrification solutions for settlements, as part of the technology selection process. Banks however presents an argument against such comparisons for first pass decisions, as the level of analysis required to make robust decisions is high. Grid and off-grid technologies offer very different levels of service, and direct comparison is difficult. The alternative route is to make national level decisions about budget constraints to grid electrification, and then to try and maximize the benefits of grid extension within these constraints, i.e.: take the grid as far as you can, and then the remainder is allocated to off-grid. This approach is realised in the planning model by defining a factor (Famax), the maximum allowable Cost per ‘electrification benefit point’. Prior determination of Famax is not critical, as users of the model can quickly home in on reasonable values through an iterative process.

Data input Demographic, infrastructure, income, planning

Calculate a total benefit index for each settlement or GIS density area for different electrification options: Pgrid, PSHS, Pmini

Score sheet with weights and agreed allocations

Calculate maximum allowable grid expenditure for each settlement Callow = Famax * Pgrid + Ccont.

Famax Allowable exp. per ‘point’

Calculate local reticulation cost Cretic Cretic< Callow

ADMD(Income) Density, size

SHS candidate

No Calculate cost/km of line ext (Ckm) required to serve settlement

Yes Allowable line extension. R(km) = (Callow-Cretic)/Ckm

Total load (ADMD, no. hh)

‘Existing’ grid layout Are there settlements which have R(km) great enough to reach the ‘existing’ grid

Iterate until no more settlements can connect to the grid Yes

No

Update ‘existing’ grid plan and note the iteration cycle (Ic) of settlement connection to grid

Use ‘least cost path’ algorithm's such as Kruskal’s to connect these to the grid

For those settlements which could not reach the grid – compare costs of SHS vs. Mini-grid supply and choose least cost option For those that could – calculate final costs

Grid candidates

Mini-grid candidates

SHS candidates

Cost estimate Benefits points Priority (Pgrid/Cost) Sequencing (Ic)

Cost estimate Benefits points Priority (Pmini/Cost)

Cost estimate Benefits points Priority (PSHS/Cost)

Figure 1:

Regional planning tool overview

Given the above determined ‘electrification benefit score’ (Pgrid) for each settlement, one can then make an immediate assessment of the maximum allowable expenditure to achieve grid electrification for a particular community or cluster of communities: Callow = Famax × Pgrid + CcontCcont is included above to allow for the transparent inclusion of additional resources from funds outside of the normal funding channels (for example from community resources) in the decision making process. (c) Costs of electrification – look up tables The capital costs of grid electrification can be conveniently split into: • the local reticulation costs (supply from main transformers in the settlement to each household): Cretic • and the line extension costs (extension from the existing higher voltage grid to the settlement): Cline = Ckm × Distance. In the approach adopted here, we have assumed that local reticulation costs are primarily a function of: the After Diversity Maximum Demand (ADMD), the proximity of households to each other (households per km2) and project size. Based on the work of Dekenah and others (7), we have assumed that average income levels in communities provide a reasonable indication of the expected ADMD (and future consumption growth for households). Income data is available from the South African Central Statistical Services3. In consultation with electrification practitioners, we then generated a simple series of look up tables, which provide an estimate of the local reticulation cost based on the above parameters. Look up tables were chosen, as they allow quick updating of costs, and they provide a readily accessed method for users of the model to check assumptions directly against their experience in the field. Line extension costs are primarily a function of the extension distance (which is determined below), and the total Maximum Demand load for the settlement (or cluster of settlements) fed by a line. Again, simple look up tables were generated based on electrification experience. (d) Decisions, decisions, decisions… If the cost of local reticulation is higher than the allowable budget (Cretic > Callow), the model assumes that grid electrification is not feasible, and that a mini-grid system (which also requires local reticulation) will not be cost effective. These settlements are allocated directly to the Solar Home System category of projects.

3

For the plans produced as part of this report, income data was not available at the time of preparation, and we therefore utilised an figure of R500/hh/month for the entire planning region.

For the remaining communities, the budget remaining after covering the local reticulation costs is then divided by the cost per km of a feeder line of sufficient capacity. The result is an estimate of the maximum line extension length that the settlement could justify (if it were to use the entire provisional budget available): R(km) = (Callow – Cretic)/Ckm If this line length is sufficient to reach the existing grid, the model assumes that the settlement will be grid electrified in due course. As the grid is gradually extended to these communities, the ‘existing grid’ database of course needs to be updated. This may mean that some more remote villages are now within reach of the extended grid. An iterative process is thus followed to capture all the settlements that have sufficient line extension budget available to reach either the existing or developed grid infrastructure. For those settlements that do not get allocated to the grid electrification group, a second level decision can be made as to whether mini-grid electrification or SHS electrification is most appropriate. The modelling tool undertakes a provisional decision process, by comparing either the total capital costs, or the estimated lifecycle costs of the different technologies (noting that the levels of service offered are similar, and thus a cost based decision is appropriate). Costing is done is a relatively simple manner. For mini-grid costing, information on existing diesel and PV systems was used to generate a look up table relating capital and lifecycle costs to the daily load requirement. In the case of SHS systems, a 200 Wh per day system was costed, with some provision made for regional variation in costs dues to solar radiation differences. (e) Line routing Choosing the best route to serve a number of settlements with a grid line is a complex task, often done quite successfully using intuitive or iterative methods by humans, but more difficult to translate into rapidly executable algorithms for computer based planning approaches. Space does not permit a full analysis of the options here. Suffice then to say that our thinking has been informed by the branch of mathematics called graph theory, and the work of Bauer [8] and [9]. The model identifies potential connection nodes on the existing grid (those closest to settlements), and seeks to join these to nodes representing the centroids of settlements, using the well-known Kruskal algorithm. The Kruskal algorithm minimises the total line length required to connect all nodes together in a tree like structure. A limitation of our approach (and others that we have seen) is that it does not adequately take into account the gradual increase in line capacity required (and hence in line cost) as one approaches the trunk of the tree. This is not a serious limitation for relatively small networks (ten’s

of km in length), but will become more severe if complex extensions serving many settlements over distances of 50 to 200 km are modelled.

close-up look at a sub region, and illustrates the technology choices for electrification of each settlement in the region.

(f) Outputs from the regional model As indicated in figure 1, the main outputs from the regional planning model are: • First pass grid extension plans • First pass technology decisions for those communities not reached by the grid • Prioritisation of projects according to: o (Socio-economic benefits score)/cost o Progressive growth of the grid (grid projects only) • Preliminary cost estimates for all projects By re-running the model for different input variables (such as the Famax factor, weightings for particular benefits or infrastructure characteristics, or different cost assumptions), several scenarios can be rapidly generated. Collation of results from different scenarios allows further analysis of trends to be undertaken.

Figure 2: Existing grid infrastructure in region

2.2 MACRO-LEVEL TOOL The macro level spreadsheet based tool is designed to receive outputs from the regional level model, and analyse these from a cash flow perspective, either using direct financial values, or using modified ‘shadow’ cost and benefit valuations to generate an economic analysis. The economics analysis approach is described in reference [10] and in the spreadsheet tool itself. Suffice to say here that model takes into account estimates of revenue growth, willingness to pay, capital requirements, operating expenditure, and revenue losses and produces cash flow analyses for the programme. It also generates Net Present Value assessments of the financial and economic costs and returns. The model can undertake sensitivity analysis of the impact of different input parameters and assumptions.

Figure 3: Grid extension possible for Famax = R3500/point

3. RESULTS 3.1 REGIONAL LEVEL TOOL The results presented below were generated using data from the Northern Transkei region of the Eastern Cape. This area has about 217 000 unelectrified households. Please note: These results are presented for illustrative purposes only, as the model has yet to be verified. Furthermore, there are significant planned and approved bulk grid line extensions scheduled for the region. These were not included in the data supplied. Figure 2 shows the existing grid layout for the region. The light grey patches represent settlements, while the black lines represent the medium and low voltage distribution lines of the Eskom grid. Figure 3 illustrates the extensions to the grid that could be achieved given a ‘Rands per point’ maximum allocation of R3500 (Famax). Figure 4 is a

Figure 4: Technology choices for all settlements.

The above results can of course only be rough indications of the optimum planning solution. It would be unrealistic to expect a computer programme, using limited data sets, to produce detailed electrification plans. However, we think that the tool will be sufficiently accurate to: • Identify particular sub regions requiring more detailed analysis in the short term. • Facilitate local level discussions around electrification planning • Generate data for macro level analysis which allows one to explore the implications of budgetary and policy level decisions. The model execution time is very rapid (about 20 minutes to generate a plan for an entire region such as that illustrated above). It is thus feasible to run numerous scenarios, exploring the effects of policy level decisions on prioritisation (the benefits scores and weightings). One can also rapidly explore the effects of different costing assumptions, or of different budgetary constraints (the Famax factor). For example, Figure 5, illustrates the changes in the percentage of settlements allocated to different technologies as the Famax factor is varied. Northern Transkei technology trade-off 100

% households

80 60 40 20 0

2500

3000

3500

4000

4500

5000

Rands per point

grid

mini-grid

off-grid

Figure 5: Changes in the technology selections as the Rands per point factor (Famax) is increased Tale 1 indicates the summary results of the macroeconomic analysis for three different scenarios, again for the Northern Transkei region. As in all results presented in this paper, these are unverified sample outputs to indicate the capabilities of the model. They should not be used to inform decision-making. It is however of interest to note that for the cost and revenue assumptions used here, it would be preferable from both a financial and economic perspective to electrify the majority of settlements (Famax = R4250/point), even though the capital requirement is higher.

Table 1 Summary of typical macro level outputs Famax (Rands per benefit point) 3250 3750 4250 Targets (% of Grid 8% 51% 73% house holds to be equipped with indicated technology by year 10)

Minigrid SHS

Total no. connections CAPEX (total) R mill. Unit capex / conn. (R) NPV of the cash flow Fin. opex (A) (R/conn) Fin capex and opex (R/conn) Econ. opex (A) (R/conn) Econ capex and opex (R/conn) 4.

LIMITATIONS REQUIRED

23%

6%

0%

69%

42%

27%

217 000 866 3991

217 000 869 4005

217 000 885 4078

(52) (3471)

336 (3103)

532 (2968)

2078

2172

2263

(878)

(750)

(712)

AND

FURTHER

WORK

The model generates detailed electrification plans for all settlements. It is however crucial to remember that it is a first pass planning tool, and will probably recommend numerous incorrect decisions for a percentage of settlements and grid routes. It is crucial that project decision makers undertake more conventional planning process (including engineering, environmental and social issues) prior to project implementation. There are a number of areas that should be further developed to improve accuracy of decisions. These include: • Data availability and quality – the model relies heavily on comprehensive data sets. Ensuring good data availability and quality is essential if the results generated are to be meaningful. • Verification and calibration of results through detailed comparison with the outputs of more conventional planning process. • The grid routing approach described in section 2.1.(e) above seems fair, but there is significant room for improvement. • The look-up tables utilised to generate the range of cost estimates required need to be developed and expanded to improve the quality of cost estimates. 5. CLOSURE As the above results indicate, the electrification modelling tools described here allows extremely rapid ‘first pass’ electrification planning to be under taken for a region (provided of course the that necessary data is available in a digital format). At the regional/local level, we think that

this could facilitate a more transparent, objective, and public electrification decision-making process. This would help to unlock investment and local level resources currently constrained through lack of information about the grid development.

and SALGA. This direct engagement with the electrification practitioners and decision makers was stimulating. For much of the project period, Douglas Banks was employed by the Energy and Development Research Centre, UCT. The contribution of this organisation and in particular of Mark Davis to the model conceptualisation is gratefully acknowledged.

At national and policy level, the tools allow rapid assessment of the implications (on the ground and at a macro level) of policy level decisions on factors such as the potential for grid coverage, capital budget requirements, economic return, and of course the split between grid and off-grid electrification technologies. The tool could also be utilised to help explore the implications of the electrification programme on proposed regional electricity distributors (RED’s).

Principal Author: Douglas Banks, BSc.Eng (UCT), PhD (Wits), has worked for several years with the Energy and Development Research Centre (UCT) on a range of rural and renewable energy research and consultancy activities. He has recently moved to RAPS, where he continues the above interests, but with an added focus on implementation and project development. His address is: Rural Area Power Solutions Pty Ltd P O Box 34921, Glenstantia ,0010 Email: [email protected]

REFERENCES [1] Department of Minerals and Energy: “White paper on energy policy for Republic of South Africa”, December 1998. [2] Banks, D.I., Willemse, J., Willemse, M.: “Rural energy services: sustainable public-private partnership based delivery?” Proceedings of the Domestic Use of Energy Conference, Cape Technikon, April 2000. [3] Davis, M. & Horvei, T.: “Handbook for the economic analysis of energy projects”. Development Bank of Southern Africa, 1995. [4] Banks, D.I.: “Criteria to support project identification in the context of integrated grid and off-grid electrification planning” Energy and Development Research Centre, University of Cape Town, 1998. [5] Banks, D.I.: “Electrification modelling for South Africa: discussion of the methodology employed”, Prepared by Rural Area Power Solutions for Development Bank of Southern Africa. [6] Banks, D.I., “Grid or off-grid settlement electrification? Decision criteria and processes” In Proc. Domestic Use of Electrical Energy Conference, Cape Town 1999, pp 59-64. [7] Dekenah, M., DT Pre Electrification Tool Ver 1999.3, copyright Eskom and Markus Dekenah Consulting, 1999. [8] Bauer, J.: Bestimmung von Versorgungsstrukturen für die Elektrifizierung ländlicher Gebiete in Entwicklungsländern, Universität Stuttgart, Institut für Energiewirtschaft und Rationelle Energieanwendung, Draft PhD Thesis, not yet published, 1999. [9] Lambert T.W.: "Optimization of Autonomous Village Electrification System By Simulated Annealing", Department of Mechanical Engineering, Colorado State University, Spring 1997 [10] Eberhard, R., “Economic Costs and Benefits of Electrification: A brief review of the South African experience with a view to determining the implications for the electrification planning model”, Prepared by Cape Associates for Development Bank of Southern Africa, 1999 ACKNOWLEDGEMENTS The work reported here was carried out on behalf of the National Electrification Coordinating Committee, and has been funded by the Department of Minerals and Energy, Eskom, and the Development Bank of Southern Africa. Their support is gratefully acknowledged. Furthermore, several discussions were held with representatives of the above organisations as well as other interested parties such as Netplan, ECON, Feather Energy

Co-authors: Hardy Jonck, BSc.Eng (UP), has worked for several years in the GIS modelling and software development domains. Hardy and Etienne Labuschagne, BSc.Eng (UP), currently work together on developing simulation, modelling and visualization software solutions integrated with GIS database systems. Their address is: Megasub Software Development P O Box 73382, Lynwood Ridge, 0040. Email: [email protected], [email protected] Ferdie Mocke MDP (UNISA), AEP (UNISA), MD (Henley) has worked in the electrification industry for several years, and was involved in the early part of the South African grid electrification programme planning. He was instrumental in setting up the HELP GIS database system (the core dataset used by this model). He is currently involved in a range of business-to-business marketing activities as well as business and project consultancy. His address is: Link2Market Pty Ltd P O Box 15354, Lyttelton, 0140 Email: [email protected] Rolfe Eberhard, BSc.Eng,(UCT) BA.Econ (UNISA) is has worked for several years in development economics, and has expertise in economics and policy aspects of a range of sectors including the water supply and energy. He currently works from New York, and is in the process of completing a PhD. His address is: 237 West 10th Street, #8 New York, NY 10014 USA [email protected] Presenter: The paper is presented by Douglas Banks