Impacts of grid architectures on temporal diffusion of PV-based communal grids (community owned minigrids or microgrids) in a rural developing community are modelled and simulated using MATLAB/Simulink and a survey-informed agent-based model (ABM) developed in NetLogo. Results show that decentralised control architectures stimulate minigrid formations and connections by allowing easy expansions of the minigrids as each decentralised PV system within a minigrid is treated equally and determines its own real and reactive power, eliminating the need for communication links. This also reduces the cost of implementing such a system; fewer connections are realized with centralised controls as such systems require high speed communication links which make them both difficult to expand and expensive to implement. Results also show that multi-master operation modes lead to more communal grid connections compared to single-master operation modes because in the former, all distributed PV systems within a communal grid have the same rank and can act as masters or can be operated as combinations of master generators (VSIs) and PQ inverters, allowing for more design flexibility and easy connections from potential customers.
Modern energy services are fundamental to all three pillars of sustainable development, i.e. social, economic, and environmental. Most energy developments must be implemented in line with all aspects of the development process, e.g. energy and communication, energy and health, energy and schools, energy and roads, etc. Energy is therefore a complementary factor to socio-economic development. Electricity, the main form of modern energy, is crucial to industrialization and easy access to it is an indicator of a nation’s standard of living. It is estimated that there are about 632 million people in sub-Saharan Africa without access to electricity today and that 84% of these people reside in rural areas, and that this number is projected to rapidly increase, mainly due to population growth, unless sustained annual investments of about US$19.1 billion in electricity generation and distribution are made [
There are two possible routes to rural electrification i.e., through extensions of existing national utility grids from large central power generation systems or through decentralized systems namely communal grids and stand-alone microgeneration systems. National grid extensions are recommended where the load demands are high enough to make the costs of such investments reasonable. Communal grids are recommended for condensed villages or markets far from national grid lines but with potential economic values to warrant such installations. Stand-alone systems are recommended for isolated homesteads or installations far from existing grid lines, other homesteads, or other installations. The International Energy Agency (IEA) estimates that only 30% of the people currently without access to electricity in sub-Saharan Africa can be cost-effectively served through national grid extensions due to the sparseness of the rural populations, rough terrains, low economic activities, and low load densities. The remaining 70% would be most cost-effectively served through decentralized systems, i.e. communal grids (52.5%) or stand-alone systems (17.5%) [
Region | Population without electricity millions | Electrification rate % | Urban electrification rate % | Rural electrification rate % |
---|---|---|---|---|
Developing countries | 1185 | 79% | 92% | 67% |
Africa | 634 | 45% | 71% | 28% |
North Africa | 1 | 99% | 100% | 99% |
Sub-Saharan Africa | 632 | 35% | 63% | 19% |
Developing Asia | 512 | 86% | 96% | 79% |
China | 0 | 100% | 100% | 100% |
India | 244 | 81% | 96% | 74% |
Latin America | 22 | 95% | 98% | 85% |
Middle East | 18 | 92% | 98% | 78% |
Transition economies & OECD | 1 | 100% | 100% | 100% |
WORLD | 1186 | 84% | 95% | 71% |
project implementations, enabling households and communities to initiate the modest power generation programs, and to modify their systems according to their changing energy needs. Depending on local resources, capacities, designs and technologies used, microgeneration systems could provide the final solutions to rural electrification in many developing nations and entrench green economies in the process. In fact, it is estimated that there will be almost 400 TWh of installed microgeneration capacity by 2030, about 40% of new installed capacities towards universal electrification in developing nations [
Many factors influence choices of technologies used in rural electrification, the main ones being availability of resources, demand, investment costs, and local socio-political and cultural environments. Different modelling tools and techniques have been applied in planning rural electrification paths in many countries. However, these often view this problem as a question of expansion of grid coverage through extensions of existing transmission and distribution lines from central power generation stations and seldom address the unique and regionally-specific challenges presented by many developing nations [
Development of electricity delivery infrastructures are path-dependent, meaning, each development decision and step affects subsequent steps, and the final outcome. Human actors are therefore the most important variables in any energy development plan as their decisions affect the way a system evolves. Proper policy-planning tools are therefore required to guide decision-makers on least-cost rural electrification topologies. In this work an agent-based model (ABM) is developed as tool for evaluating impacts of network architectures on temporal diffusion of PV-based communal grids in a rural developing community. The model takes into account the complexities and limitations of solar electricity microgeneration technologies, decisions by human actors, geographical factors, and interactions between the three factors in order to capture the overall macro-effects of different micro-decisions in a virtual world; ABMs seek to model individual entities within a complex system and the rules that govern the interactions of the entities within the system, to capture the overall effect of such interactions. The novelty of the model developed in this work is that it simultaneously captures how technical and socio-economic factors affect temporal diffusion of PV-based communal grids in a typical rural developing community. Specifically, the model simulates how different communal grid architectures affect costs, ease of connections, and scalability.
A PV-based communal gridis defined as a locally confined and independently controlled electric power grid in which a distribution architecture integrates distributed loads and distributed energy resources―i.e. local distributed PV systems and energy storage devices―which allows the communal grid to operate in both islanded and utility grid-connected modes.Communal grids are operated in islanded mode when they are situated far from existing utility transmission networks, or if the owners of the grid prefer independence of their own power source. On the other hand, communal grids can be utility-grid connected to improve system economics, to improve operations, and to improve stability, while the utility grid gets to enjoy improved availability, improved stability, and reduced conductor sizes. For a successful interfacing of communal grids to utility grids, converters/inverters should allow net metering (bi-directional flow of power), peak shaving, advanced communications and control, and fast demand response.
Inverters are used to interface communal grids with utility grids and can be classified according to modes of operation as PQ or V-f (also known as voltage source inverter (VSIs)). A PQ inverter controls the real (P) and reactive (Q) power by adjusting the magnitude of the output real and reactive current. It therefore operates as a voltage controlled current source [
P ( f ) = P 0 − ( f s e t − f ) k f (1)
Q ( V ) = Q 0 − ( V s e t − V ) k v (2)
where P 0 is the power delivered by the inverter at set point frequency f s e t , k f is the gradient of the droop, which determines how much the active power P will change in response to a change in frequency f, Q 0 is the reactive power delivered/consumed by the inverter at set point voltage V s e t and k v is the gradient of the droop, which determines how much the reactive power Q will change in response to a change in voltage V.
The two general control structures used in communal grids are centralised control and decentralised controls [
In master slave mode of operation, all the inverters in the network operate in PQ mode when the communal grid is connected to the utility grid, since the latter controls the voltage and frequency. In islanded mode, a single master inverter switches to voltage source (VSI) mode to provide the voltage and frequency references while the remaining distributed PV systems continue to operate in PQ mode [
In decentralised configuration every decentralised PV system, including the master has a built-in generation profile determined by P-f and Q-V droop curves. Each unit therefore determines its own real and reactive power. A single pilot machine is used to set a pilot waveform. The master VSI sets the voltage and frequency based on its droop and the PQ inverters determine the active power from the system frequency set by the master VSI and the reactive power from the local voltage measurements [
In multi-master mode of operation, all distributed PV systems within a communal grid can act as masters or can be operated as combinations of master generators (VSIs) and PQ inverters [
Decentralised configuration for multi master is almost identical to the single master mode. All distributed PV systems can regulate their power outputs in grid-connected and islanded modes by using individual droop controls. Here also, a single pilot machine is used to set a pilot waveform. Multi master pure droop systems are well suited to systems with multiple integrated, renewable energy/storage generators spread throughout a communal grid. When implemented in this way the load will be shared equally across all distributed PV systems [
Different communal grid architectures, revolving about energy storage, are modelled and simulated in MATLAB/Simulink to determine the most cost-effective option for a given transmission and distribution network. Each system modelled below comprises of 4 PV arrays, each rate at 100 kW at 1000 W/m2
radiation and 25˚C; each array consists of 66 parallel strings, each comprising 5 PV 330 SunPower (SPR-305-WHT-D) modules connected in series (66 × 5 × 305.2 = 100.7 kW). Each array is connected to a 5 kHz boost DC-DC converter with maximum power point tracking (MPPT) and charge-controlling capabilities. Each converter uses the perturb-and-observe method to extract maximum voltage (273V DC) from each array and then boosts it to 500 V DC. For utility grid interfacing, a DC-AC inverter inverts the DC bus voltage to 260 V AC, while keeping a unity power factor. The inverter uses two control loops: one which regulates DC link voltage to +/−250 V and an internal control loop which regulates active and reactive current. Active current reference is the output of the DC voltage external controller while reactive current reference is set to zero to maintain unity power factor. A 40-kvar capacitor bank is used to filter harmonics produced by the inverter. The inverter is connected to the utility grid through a 400-kVA 260 V/25 kV step-up transformer.
Communal grids can be classified as ether DC-coupled or AC-coupled depending on the common bus voltage. A grid using a DC bus avoids many of the power conversion steps required when using an AC bus leading to higher energy efficiency and cost-saving. Traditionally, AC enabled efficient voltage transformation and high-voltage power transmission over long distances. Recently however, technology advances have led to highly efficient AC/DC and DC/DC converters, making high-voltage DC long-distance bulk power transmission more efficient [
・ Voltage and frequency values need to both be regulated through active and reactive power control;
・ If a decentralised power source is a traditional synchronous generator with an AC output, and is connected directly to the utility grid without power electronic interfaces, stability is controlled through the machine shaft’s torque and speed control.
In DC systems, there are no reactive power interactions which suggest that there are few stability issues; system control seems to be oriented towards voltage regulation only.
In these systems generated DC voltage by the PV systems is distributed throughout the communal grid in the DC form. Inverters are therefore not required for common bus connections. Each DC-DC converter for each PV system is necessary for maximum power point tracking (MPPT) and for charge controlling of the storage system.
Majority of appliances require AC power to operate and AC is the main form of power transmitted and distributed around the world. It is therefore important to model AC coupled-networks, and compare them to DC-coupled networks.
network. Due to decentralised storage, each household has its own DC-AC inverter for connection to the common AC bus. A decentralised control system is used to set the network voltage amplitude and frequency. The system works by using a virtual communication system to independently determine the operating points of each decentralised PV system and load within the communal grid network, effectively eliminating the need for communication links and thus increasing system reliability as well as reducing cost [
controlling capabilities is used to connect the central energy storage system to the common AC bus. Since the central inverter is large compared to the individual array inverters, it also acts as master VSI for setting up reference line voltage amplitude and frequency while other inverters function as PQ inverters. Future connections to the communal grid are complicated by the fact that each time the storage capacity is increased to meet the increased demand, a new appropriately sized AC-DC inverter must be purchased for the network.
Simulated data from the Simulink models are fed into an agent based model (ABM)developed in NetLogo, to simulate effects of control architectures on temporal diffusion of PV-based communal grids in a rural developing community, with the main driving factor being the cost of setting up and connecting to a given control architecture. NetLogo is a multi-agent programmable language that was authored by Uri Wilenskey and developed at Northwestern University [
The following agents are created in the model:
1) A representation of the environment and the solar potential in it,
2) The populations in it that require electricity,
3) PV seeds that would use the environment to produce electricity,
4) Links which are used by households with PV to connect to communal grids, and
5) A central observer or stakeholder who determines the strategies and preferences for PV diffusions.
Through these agents and the rules created for their interactions, the model is used to simulate the way decisions and preferences by human actors affect PV diffusion. Kendu Bay survey data are used to inform the model [
A household without PV would consider installing PV if
L U C E P V < C A / k W h (3)
where C A / k W h is avoided cost per kWh, i.e., the prevailing national grid electricity cost per kWh and equals $0.20/kWh as per current Kendu Bay rates [
L U C E P V = A L C C P V W p × E H F S × 365 × C U F (4)
where W p is the rated peak Watt capacity of the PV module and is based on a household’s activity profile and power demand [
Parameter | Value and Unit | Description |
---|---|---|
C A / k W h | $0.20/kWh | Avoided Cost |
EHFS | 8 Hrs | Equivalent hours of full sunshine per day |
CUF | 0.9 | Capacity utilization factor |
C 0 P V | $1.40/Wp | Capital cost of the PV module |
C 0 b a t t | $0.95/Ah | Capital cost of the battery |
C 0 c c | $28.50 | Capital cost of the charge controller |
C 0 a p p l | $75/household | Capital cost of appliances |
C O & M | $2.50/year | Operations and maintenance cost |
i | 12% | Discount rate |
C 0 P V C G | $1,400/kWp | Capital cost of the PV array |
C R F p c u | 0.2774 | Capital recovery factor of the power conditioning unit |
C 0 d n | $2500/km | Capital cost of the power distribution network |
C 0 s c | $125/household | Capital cost of service connections including internal wiring and appliances |
L | 1 km | Length of the distribution network |
C R F p d n | 0.1770 | Capital recovery factor of the distribution network |
R | $1400/kWp | Benchmark unit cost of the communal grid |
b | 0.95 | Scale factor for incorporating cost reduction in overall cost of the communal grid |
reasons and equals 0.9 based on Kendu Bay survey data [
A L C C P V = ( C 0 P V × C R F P V ) + ( C 0 b a t t × C R F b a t t ) + ( C 0 c c × C R F c c ) + ( C 0 a p p l × C R F a p p l ) + C O & M (5)
where C 0 P V is the capital cost of the PV module and is set at $1.40/Wp based on Kenya’s PV costs, C 0 b a t t is the capital cost of the battery and is set at $0.95/Ah based on the price of a typical 48V 100 Ah battery in Kenya, C 0 c c is the capital cost of the charge controller and is set at $28.50 as per Kenyan rates, and C 0 a p p l is the capital cost of appliances and is averaged at $75 per household based on typical load profiles of 95% of households with PV systems in Kendu Bay [
Capital recovery factor (CRF) is calculated using the formula
C R F = i ( 1 + i ) n ( 1 + i ) n − 1 (6)
where i is the discount rate and is set at 12% based on average Kenya’s central bank rates over the past 12 months, while n is the life of the particular component being considered, i.e. 25 years for PV modules/arrays, 5 years for batteries, 5 years for converters/inverters/charge controller, and 10 years for appliances (cumulative load). Based on these figures, C R F P V = 0.1339 , C R F b a t t = 0.2774 , C R F c c = 0.2774 , and C R F c c = 0.1770 .
For a communal grid consisting of N households, levelised unit cost of delivered electricity ( L U C E C G ) is given by
L U C E C G = A L C C C G P V C G × E H F S × 365 × C U F (7)
where P V C G is the capacity of the communal grid (PV array) in kWp and is based on the activity profiles of all households connected to the communal grid, EHFS is the equivalent hours of full sunshine per day and equals 8, CUF is the capacity utilization factor which incorporates non-utilization and outages of systems due to various reasons and is maintained at 0.9, and A L C C C G is the annualized life cycle cost of the communal grid and is given by
A L C C C G = [ ( C 0 P V C G × C R F P V ) + ( C 0 b a t t C G × C R F b a t t ) + ( C 0 p c u × C R F p c u ) ] × [ P V C G × R ] b + [ ( C 0 d n × L ) + ( C 0 s c × N ) ] × C R F p d n + C O & M C G (8)
where C 0 P V C G is the capital cost of the PV array and associated mounting structures and is set at $1400/kWp, C 0 C G r i d _ b a t t is the capital cost of the battery bank and associated structures and equals $0.95/Ah × N, C 0 p c u is the capital cost of the power conditioning unit (power electronics) and equals $28.50 × N, C R F p c u is the capital recovery factor of the power conditioning unit and equals 0.2774, C 0 d n is the capital cost of the power distribution network per km and equals $2500 for Kendu Bay, C 0 s c is the capital cost of service connections including internal wiring and appliances per household serviced and equals $125 based on similar networks in Kenya, L is the length of the distribution network in km and is initially set at 1 km for a neighbourhood radius of 500 m, C R F p d n is the capital recovery factor of the distribution network including the service connections and is calculated to be 0.1770, C O & M C G is the cost of operating and maintaining the communal grid and is estimated to be $2.50 × N, R is the benchmark unit cost of the communal grid and is set at $1400/kWp, and b is a scale factor for incorporating cost reduction in overall cost of the communal grid, without the power distribution network (pdn), due to bulk purchasing of the components used in the grid [
Once cost has been determined, PV is installed by a household if
H P V / S R H T o t a l / S R × 100 > T S R (9)
where H P V / S R is the number of households with PV within a given sensing radius (SR) or neighbourhood, H T o t a l / S R is the total number of households within the same sensing radius, and T S R is the neighbourhood threshold. For Kendu Bay, the initial sensing radius is set at 500 m based on population distribution; a sensing radius below 500 m limits impacts of neighbourhood influence on a household to one’s own clan, based on homestead and farm distributions [
The size of the installed PV system by the household is given by
P V ( W p ) = ( P × h η i n v ( P V ) × η b a t t ( P V ) × η c c × ( 1 − f t e m p ) × ( 1 − f d u s t ) × ( 1 − f m i s m a t c h ) × E H F S ) (10)
where P is the household’s power load, h is the number of hours the load is operated per day, η c c is the charge controller efficiency, f t e m p , f d u s t , and f m i s m a t c h are the losses in PV module due to cell temperature, dust, and mismatch amongst several cells due to shadow and other factors, and EHFS is the equivalent hours of full sunshine. P is calculated from a list of stochastically allocated appliances based on active occupancy, time of the day, and day of the week [
The required battery capacity for the household is given by
B a t t ( A ⋅ h ) = ( P × h η i n v ( P V ) × V × M D o D × η b a t t ( P V ) ) × D (11)
where V is the operating voltage of the battery, η i n v ( P V ) is the efficiency of inverter, MDoD is the maximum depth of battery discharge, η b a t t ( P V ) is the charging/discharging efficiency of the battery, and D is the days of autonomy.
To be allowed to join a communal grid, a household must have electricity needs beyond lighting and must demonstrate this by having installed PV of a given minimum capacity (power-threshold), which is set at 2 KWp. They must also be within a given sensing radius of other houses with PV that meet the power-threshold. A household will join a communal grid if
H C G / S R H T o t a l / S R × 100 > T S R (12)
where H C G / S R is the number of households with PV within the sensing radius that meet the power-threshold and H T o t a l / S R is the total number of households within the same sensing radius.
For a communal grid consisting of N households, the total PV capacity is given by
P V C G ( kW p ) = ( P × h η i n v ( C G ) × η b a t t ( C G ) × η c c × ( 1 − f t e m p ) × ( 1 − f d u s t ) × ( 1 − f m i s m a t c h ) × D F × E H F S ) (13)
where η i n v ( C G ) is the cumulative efficiency of inverters used in the communal grid, η b a t t ( C G ) is the cumulative charging/discharging efficiency of the batteries used in the communal grid, and DF is the diversity factor, ratio of the sum of all individual peak loads to the maximum load of the entire communal grid.
The total battery capacity required for the communal grid is given by
B a t t C G ( A ⋅ h ) = N × ( P × h η i n v ( C G ) × V × M D o D × η b a t t ( C G ) × D F ) × D (14)
Generally, P V C G ( kW p ) should equal N × P V ( W p ) , while B a t t C G ( A ⋅ h ) should equal N × B a t t ( A ⋅ h ) ; the deficit should be evenly distributed between the households forming the communal grid. i.e., they may need to increase the sizes of their installed systems.
all simulations, the temperature is kept constant at 25˚C.
As with current outputs, power outputs from the 4 PV arrays also mirror the irradiances hitting the arrays as shown in
Additional Power Electronics | Centralised | Decentralised | ||
---|---|---|---|---|
DC | AC | DC | AC | |
VSI | 1 | 1 | 1 | 1 |
Central Charge Controller | 1 | 1 | 0 | 0 |
DC-AC Inverter | 0 | 4 | 0 | 4 |
Transformer | 1 | 1 | 1 | 1 |
Filter | 1 | 1 | 1 | 1 |
Total Additional Cost ($) | 250 + GC | 950 + GC | GC | 700 + GC |
networks than AC-coupled networks. This is due to additional costs incurred in DC-AC inverter purchases.
In this work, simulation results from different communal grid architecture
Time (Years) | DC-Coupled | AC-Coupled | ||
---|---|---|---|---|
Decentralised Storage | Centralised Storage | Decentralised Storage | Centralised Storage | |
0 | 0 | 0 | 0 | 0 |
5 | 414 | 301 | 383 | 249 |
10 | 842 | 609 | 744 | 493 |
15 | 1276 | 928 | 1126 | 748 |
20 | 1679 | 1249 | 1497 | 1012 |
25 | 2103 | 1571 | 1887 | 1286 |
models developed in MATLAB are fed into an agent-based model developed in NetLogo to determine the most cost-effective option. The novel aspect of this
paper is that it combines a socio-economic agent-based model with a technical model developed to simulate the most cost-effective communal grid control architecture.
Generally, DC-coupled networks seem to fair better than AC-coupled networks in all categories. This is mainly due to cost and ease of set-up of such networks. As explained above, DC systems tend to be more modular and scalable than AC systems because DC converters are easier to control and to parallel. This allows for more flexibility in system design and expansion, and thus more effective capital investment management. In addition, DC system components tend to be more compact than equivalent AC components because of higher efficiency and by not being frequency dependent. This therefore leads to lower capital costs due to fewer electronic components being used. Moreover, most distributed energy sources and storage devices have inherently DC outputs, making DC architectures more natural options for their integrations. Due to lower power and energy rating, stability issues are more prevalent in communal grids than in utility grids. However, in DC systems where there are no reactive power interactions, there are few stability issues; system control seems to be oriented towards voltage regulation only.
Results show DC-coupled networks with decentralised storage stimulate communal grid connections by allowing easy expansions of the communal grids as each decentralised PV system within a communal grid is treated equally and determines its own real and reactive power, eliminating the need for communication links. This also reduces the cost of implementing such a system; fewer connections are realized with centralised controls as such systems require high speed communication links which make them both difficult to expand and expensive to implement.
Results also show that multi-master operation modes lead to more communal grid connections compared to single-master operation modes because in the former, all distributed PV systems within a communal grid have the same rank and can act as masters or can be operated as combinations of master generators (VSIs) and PQ inverters, allowing for more design flexibility and easy connections from potential customers.
In conclusion, for cost-effective communal grids, DC-coupled networks with decentralised storage systems are recommended as the most-cost effective architectures for rural communal grids. Where a communal grid must be AC-coupled communal grids, systems with decentralised storage and multi-master operation modes are recommended for cost-effectiveness.
This research was funded by Leeds International Research Scholarship.
・ Decentralised control architectures lead to more connections than centralised controlled;
・ DC-coupled communal grids are more cost-effective than AC-coupled ones;
・ Combination of DC-coupling and decentralised storage leads to increased PV installations;
・ Multi-master operation modes lead to more connections than single-master modes.
The authors declare no conflicts of interest regarding the publication of this paper.
Opiyo, N.N. (2018) Different Storage-Focused PV-Based Mini-Grid Architectures for Rural Developing Communities. Smart Grid and Renewable Energy, 9, 75-99. https://doi.org/10.4236/sgre.2018.95006