1. Introduction
In 2016, the UK received 24.5% of its electricity generation from renewable technologies, with further 46.2% coming from low carbon systems [
1]. Most electricity in the UK grid comes from large producers located far from customers, e.g., large offshore wind farms or isolated nuclear plants. This leads to enormous volumes of electricity wasted through transmission losses and step-up/step-down transformers [
2].
This reality has raised the incentive to research different ways to provide more economical and environmentally friendly electricity production. Many authors have generated various MILP models to demonstrate the potential of distributed energy resource (DER) systems to provide both domestic and commercial buildings with greater local, self-generated energy from renewable or low carbon assets. This will be able to mitigate much of these transmission losses and, consequently, reduce carbon emissions and energy bills [
3,
4,
5,
6,
7].
Much work into this area exploits the fact that many governments are willing to pay users to consume self-generated, low-carbon electricity and heat, through Feed-in tariffs (FITs), and renewable heat incentives (RHIs). They will also pay for these units to export excess electricity to the wider national grid [
8,
9], a good case for homeowners and businesses to invest in DER technologies. However, while such investment would help meet national CO
2 reduction targets, accompanied by money saving for customers, large-scale investment into domestic DERs without sensible controls could destabilize national grid structures. The reason for this being that many already deal with various large-scale renewable assets intermittently feeding into it, not necessarily at times of demand [
10]. This forces grid operators to ask large non-renewable plants to start-up and shut-down operation without much warning, makes grid balancing difficult and expensive, and leads to increased costs for the consumers.
Accordingly, to make residential DERs compatible with the national grid, a control strategy is needed which encourages renewable asset owners to generate, consume, store and sell energy at times which are mutually beneficial to them and ensure the stability of the national grids.
The IoT concept is defined as a group of infrastructures interconnecting connected objects and allowing their management, data mining and the access to the data they generate [
11]. The IoT technologies have the potential to better control and optimize DERs in conjunction with the needs of the national electricity grids, by providing a sustainable solution for the dynamic management of the system. This can be done through internet connected controllers which utilize live data on energy prices, generation, consumption, and asset failures, in a manner aligned with national grid strategies. They also present the capability to optimize the design and operation of local renewable resources to cope with predicted changes in weather or disruptive events.
One strategy which has been proposed to help stabilize national grids is the use of dynamic pricing, as discussed in Finn and Fitzpatrick [
12]. Dynamic pricing is a strategy in which national grids publish, either in real time, or a few days ahead, variable prices per kWh of electricity, within a given time period. For example, one might make 06:00–08:00 P.M. on a winter’s day an expensive time in which to consume energy, because this is a time of high demand, when many homeowners cook dinner and/or watch television. The effect of this would be that large-scale consumers, for example chemical processing plants, might cut back operation during this period to save money, and recover the lost production when electricity demand is much lower and the consumption is encouraged by the grid through a lower cost per unit of electricity, for example at 02:00 A.M.
Without automated optimization of such variable operation, only a few large industrial users will likely modify their usage to accommodate such a strategy. It has been suggested that dynamic pricing could be employed in smart homes [
13], where, for example, electric vehicles would be charged overnight when electricity is cheap and dish washers would only turn on in the early hours of the morning, when the IoT controller tells them to, in accordance with user pre-set requirements.
This strategy could also be flipped on its head and used to encourage renewable resources within an IoT controlled residential DER network (e.g., wind turbines (WTs), solar cells, battery storage etc.) to sell electricity to the grid during peak national demand, when the sale price is high, and to store renewable power, in battery units, when the sale price is low [
10]. These two routines would help flatten the demand profile throughout the day (or perhaps even the year) and simplify the balancing of demand and supply for the grid operators, and could even reduce the voltage used on national grids since peak volumes of energy could be reduced.
Dynamic pricing could be used to reflect the needs of the grid, in a process known as Demand Side Management [
12], which can be a function of the time of day, week and year, as well as weather conditions and public demand. Pricing variations of imports and exports could also reflect the failure of large assets, e.g., emergency shut-down of a large coal or nuclear asset, or changes in legislation of FIT and RHI pricing.
This could better enable grid actors to deal with blackouts and possibly minimize the size of the affected areas during such events. The IoT controllers could also use big data capabilities to learn more about end-user’s operation, which would better inform the grid on the generation and demand profiles, and enable the integration of multiple local grids that support each other, possibly even trading energy between one another [
10].
While much has been hypothesized about the potential of IoT to stabilize grid operations, in a Smart Grid [
10] or The Enernet, the inevitable convergence of the smart grid with the IoT [
14], little has been done to quantify the possible effects, either using optimization models or through real-world implementation.
Xu et al. [
10] have presented a statistical study to evaluate the differences in cost per kWh and the reliability between three types of national grid structures, namely:
- (1)
The current typical structure in which the electricity is generated in large industrial units, far from end users;
- (2)
A grid with integrated DERs; and
- (3)
A grid with integrated DERs and battery storage units. With respect to the end-user energy costs, the findings were that the DER integrated network is cheaper than the traditional grid.
Furthermore, battery storage introduces greater reliability to the network, albeit at a greater cost [
10]. This reflects the findings of authors who have written solely about residential DER design optimization. Wouters et al. [
7], for example, found that battery storage was not economically viable for residential DERs in South Australia.
The work presented in the following sections focuses on the use of IoT-type integration and communication strategies for the control of residential DERs, as opposed to the control of entire grids, as presented by Xu et al. [
10]. With respect to dynamic pricing, the paper is unique from the work of Finn and Fitzpatrick [
12], in that it investigates both demand- and supply-side control, with focus on residential housing as opposed to industrial buildings. Furthermore, it also expands on the potential use of batteries and investigates the maximum cost at which batteries become economical under various scenarios, thus giving greater insight into their use than previous models which have simply stated that they are not economical [
7].
Finally, the framework presented here builds on the capabilities of those presented by Mechleri et al. [
6] and Wouters et al. [
7], by expanding the models in terms of different types of technologies used and also the different approaches for the case studies, and builds on the preliminary results presented in [
15].
The remainder of the paper is split into the following:
Section 2 describes the methodology followed in this work,
Section 3 reports the results of the various scenarios considered, while
Section 4 focuses on the conclusions. Finally, the Appendix includes a glossary of technical terms and the appendices which contain detailed data used for the MILP model.
2. Methodology
A baseline scenario, with a pre-designed DER network is used to investigate the effects of dynamic pricing. An optimization problem is formulated for the design of the network without dynamic pricing, based on several available renewable technologies. The set of candidate technologies (Tech) are:
Absorption chillers (Abs);
Air conditioning units (AC);
Biomass boilers (BB);
Gas boilers (GB);
Combined heat and power generators (CHP), which can be fuel cells, internal combustion engines or Sterling engines;
Gas heaters (GH);
Heating/cooling pipelines (PLs);
Microgrid controllers (MGCC);
Photovoltaic cells (PV);
Thermal storage (TS);
Wind turbines (WTs).
A second scenario is investigated where the design of the network is done considering dynamic pricing control strategies. The two scenarios are then compared to evaluate the contribution of the dynamic pricing in the design of the network.
A case study [
15] is used to illustrate the approach: a residential house arrangement based on streets in Guildford, Surrey, UK, with five (5) houses (
Figure 1), numbered i1–i5 and spaced 10 m apart. This spacing represents the lengths that would be needed to make connections via pipelines (PLs) or microgrid (MG) cables.
The homes can generate heating, cooling, domestic hot water (DHW) and/or electrical power to meet their own demands but also the needs of the other homes in the network, transmitted via pipelines and MG cables. Excess electricity generated can be sold to the national grid and electricity and gas imported from the grid, at a price, to the network.
Although the application is demonstrated on a UK case study, the framework can be applied easily to scenarios for other countries by replacing sizes and costs of technologies, as well as demands, pricing or policy-related costs, as required by the local situation.
2.1. Seasonal and Hourly Variations in Supply and Demand (Multiperiod Operation Discretization)
Real seasonal and hourly variations in supply and demand of the various energy requirements and renewable generating assets are considered for the models, as discussed in [
15]:
- (1)
UK weather and demand data is split into four seasons: m1—spring; m2—summer; m3—autumn; and m4—winter, which each lasts for three months, March–May, June–August, September–November, and December–February, respectively.
- (2)
Each month within a season has the same number of days, taken as the average of the three months: m1 = 31; m2 = 31, m3 = 30, m4 = 30.
- (3)
Furthermore, every day is split into six time periods, p1 to p6, in which demand, weather operation and thus renewable energy supply are constant.
The length and the start and end times for each of the six time periods are shown in
Table 1.
The information on weather data considered for the model are shown in
Table 2 and
Table 3.
The demand profiles considered for all utilities can be seen, as a function of the time of day and season, in
Figure 2, constructed from data in [
4,
16,
17,
18,
19]. A representative day for each season (Spring, Summer, Autumn and Winter) is chosen in determining the demand profiles.
A scalar is considered for each house (
Table 4) that is used to define five different demand profiles for each utility at any one time.
It is ensured that using the five scalars, the average demand per household for each utility is the one defined in
Figure 2.
2.2. Mathematical Formulation
A simplified version of the model was previously described in [
15]. In the following sections, the mathematical formulation, with all equations and assumptions will be presented in more detail.
2.2.1. Objective Function
The model’s objective function minimizes the annualized cost of the energy network (
). The total cost is calculated as the sum, for all technologies, of the investment costs (
) annualized over the project lifetime (
n) and the annual operating and maintenance costs (
), the annual cost of carbon emission taxation (
), the annual cost of imported energy and fuel (
), minus the annual income (
) made from exporting electricity to the grid, feed in tariffs (FITs) and renewable heat incentives (RHIs), as shown in Equation (1):
2.2.2. Annualized Investment Costs
The annualized investment costs are calculated based on a capital recovery factor (
CRF) using the interest rate,
r (%), and the project lifetime,
n (years). For all technologies, the values of
r and
n are kept constant and based on the values available from similar investigations ([
6,
7,
20,
21,
22,
23]), allowing for ease of comparison between different technologies, scenarios and research conditions:
The investment cost of the installed capacity can be calculated as:
Here,
is the maximum amount of
Utility generated by the technology
Tech in a house
i, in a year. The following utilities are considered: electricity (E), space heating (H) and cooling (C), and domestic hot water (DHW). Hence
represents the required installed capacity of the technology
Tech and must be at least equal to or greater than the highest yearly demand:
The installed capacity of the technologies
Tech is bounded by upper and lower bounds of typically available equipment:
Here, is a binary variable which decides if a technology Tech is placed in a home i.
Equations (4) and (5) apply to all technologies
Tech presented at the beginning of
Section 2. In case of the photovoltaic network, the investments are calculated based on the installed surface area of the panels. Furthermore, the electricity generation from PV cells is a function of the solar irradiance,
, which varies with the time of day and season, the surface area installed,
, which is an optimisation variable, and the panel efficiency,
, a set parameter:
The generation from the PV cells is also a function of the rated panel capacity,
RPC:
In case of the WTs, the investments costs are based on the number of turbines installed and not on the capacity. The maximum capacity is set to 1.5 kW and is equal with the rated capacity, as used in [
24]. Furthermore, it was decided that it was impractical to have WTs in adjacent houses, hence a rule was put in place that there must be at least a one house gap between WT installations:
Finally, for the sake of space within homes, only one CHP is permitted in each dwelling:
In case of the network operation, the capital costs,
are calculated based on the length of the cable or pipe installed. Furthermore, for the batteries, besides the cost expressed by Equation (3) there is a secondary investment cost associated with the storage unit controller:
The capital cost of the microgrid controller is calculated from:
With Z being a binary variable which determines the existence of the microgrid.
2.2.3. Operating and Maintenance Costs
The operating and maintenance (
O&M) costs,
, of the technologies
Tech are calculated using fixed and variable costs, as used in other case studies ([
6,
7,
20,
21,
22,
23]):
Finally, since the microgrid itself uses electricity it has an
O&M cost as well:
2.2.4. Environmental Costs
A tax is put in place by the UK government to penalize electricity generators for CO
2 emissions. The cost of carbon emission taxation,
, is calculated from the government tax of CO
2 emitted,
, the carbon intensity of electricity purchased from the grid,
, and the carbon intensities of emitting technologies (gas boiler, CHP, gas heater). Given that all of these units consume natural gas, this can be considered as the carbon intensity of the grid gas,
, divided by the technology
Tech’s efficiency,
:
where
is the electricity purchased from the grid.
Accordingly, the CO2 cost of the network is calculated by calculating the volume of CO2 produced by each technology Tech throughout the year.
2.2.5. Costs of Fuel
The cost of fuel,
is equal to the volume of gas and electricity purchased multiplied by the respective cost of each:
The cost of gas from the national grid is defined by (£/ kWh).
2.2.6. Income
The income,
, is calculated from the FITs per kWh of low carbon
Utility generated,
, the electricity sales to the grid,
, and the annual payment for utilisation of a biomass boiler under the RHI,
. The electricity generated in the network, which exceeds network demands, can be sold to the national grid:
To ensure that homes do not become power plants and are, in fact, only selling excess and not generating electricity purely for profit, a rule is put in place stating that the total electricity sold from all technologies
Tech cannot exceed that generated by all units, minus the demand of the home:
The technologies contributing to the terms in Equations (16) and (17) are: PV arrays, WTs, ICHP, SCHP and FCHP.
At any time, a house is either buying or selling electricity, not both:
With = upper bound on the volume of electricity sold; = binary stating whether the house i sells electricity in the time period m, p.
A similar equation is defined for buying electricity from the grid:
2.2.7. Demand Equations
The energy demands of the houses relate to the four utilities considered (Utility = E, H, C, and DHW). Balance equations are written to meet these demands with the appropriate technologies.
The demand balance equations can be written in the following form:
The
Utility transferred via the network from house
j to house
i is subtracted from the demand, while the
Utility transferred from
i to house
j is added to the demand. Hence, the left side of Equation (20) is the net
Utility demand and is met by the right-side terms, namely the
Utility imported from the grid, the
Utility from the Distributed Generation (DG) technologies, and the
Utility from Storage facilities available within a house
i. The term
is the loss coefficient [
6], dependent on the cable or pipe length,
li,j, between two homes,
i and
j, as shown in Equation (21), where
is a scalar of low value:
The
Utility transfer term must equal the energy obtained from contributing technologies
Tech within house
i:
In case of the electricity balance equation, the AC electricity demand is added to the total electricity demand, as it is a variable, dependent on the number of AC units installed and their times of operation. The DG technologies contributing to the self-use term, in case of electricity are WTs, CHPs and PV cells.
In case of the heat balance equation, there is no heat imported from the grid. The transferred hot water term does not refer to the hot water for DHW use but represents only the water shared between homes. The technologies that are generating self-heating are CHPs, GBs, BBs and GHs. According to legislation, the BB units are not allowed to provide heat to more than one home [
25]:
There is also a limitation on the annual heat produced from BBs which receive RHI payments [
25]:
Heat is a waste product of the CHPs, generated in volumes proportional with the electricity generation via the heat to electricity ratio (HER). This heat is used directly for heat and DHW, or indirectly in the Abs units.
For the cooling balance equation, the thermal storage of cold loads is not possible since it is assumed that only one thermal storage unit is installed per house. Two units would be required to separate cold and DHW water.
Similarly, to heat, no import of cooling from the grid is considered. The sharing term is done through pipeline connections between two houses i and j. The technologies considered for the self-generation terms are the AC and Abs units. The Abs units cannot be modulated to meet demand because their power is a result of heat produced as a side product during the CHP electricity generation.
Finally, for the DHW balance, there are no Transfer terms. The DHW is not shared, since hot water is needed in the summer and would heat the pipelines transferring cold loads. All thermal DGs except for the GH units can supply DHW, as is thermal storage. No technologies add to DHW and no import from the grid is considered.
2.2.8. Microgrid and Heat Network Operation
The connection between two homes,
i and
j is defined by the binary parameters,
, with
k =
MG in the case of the microgrid, and
k =
PL, in the case of the heat network operation. A
Utility (
E for the microgrid, and
H for the pipelines) cannot move between unconnected nodes of the network. The binary parameter,
defines the microgrid connection between two homes,
i and
j:
With = the upper bound for Utility transfer.
Moreover, a house cannot transfer itself
Utility, hence
and
. The
Utility flows in one direction through the network are in this case:
Looped networks are mitigated by Equation (27), where
is a numerical position of a microgrid or pipeline connected home, respectively. A higher order home cannot transfer
Utility to lower order homes. They can only import it (no looped microgrid or heat networks can be formed). The term
is the value
i.
In case of the microgrid, one controller operates the whole neighborhood and a binary variable,
Z, is defined that decides the microgrid’s existence:
2.2.9. Battery Storage
The balance equation for the electrical storage is written as:
where
= the static loss coefficient (%) and
= the charge and discharge rates (%), respectively.
The energy which enters the battery plus the volume already present in the battery cannot exceed its capacity:
Additionally, what is withdrawn from the battery cannot exceed what it is within it:
The energy entering the battery is equal to the total energy from the contributing technologies,
Tech:
The technologies contributing to the battery storage are ICHP, SCHP, FCHP, PV cells and WTs. The energy entering the battery is also bounded by the maximum charge rate:
The outlet from the battery is similarly bounded by the maximum discharge rate:
The volume of energy in the battery at any time is calculated using the Depth of discharge (
DoC), which is a measure of what percentage of the battery’s capacity must be left uncharged:
Equations (30) and (33) are adapted for the case study in which batteries are able to sell electricity to the grid. Equation (30) is adapted so that power in the battery could also be sold to the grid, using the term
. The volume of energy in the battery units, calculated in Equation (33), is adapted in such way that it could also charge from the grid with the term
:
2.2.10. Thermal Storage
The balance of energy in thermal storage is determined using:
The energy entering the thermal storage unit is the sum of the energy stored by all contributing technologies (GBs, BBs, CHPs):
The installed capacity of the thermal storage unit must be equal to or larger than the input to the unit plus the energy already in it at any time (
m,
p). The capacity must also be within the limits of the available equipment:
A balance is needed to calculate how much energy can be withdrawn and is based on the energy provided in the previous time:
Additionally, the volume of heat stored in the unit cannot exceed its capacity:
2.2.11. Wind Turbines
Generation from the WT units is based on wind data provided to the model as a function of time and season,
.
The power generated by the WTs is not a continuous function of the available wind resources, but is instead a piecewise function dependent on the “rated”, “cut in” and “cut-out” speeds for the units available:
For the model of WT used the following parameter values are considered: VCI = 3 m/s, VCO = 60 m/s and Crate = 1.5 kW.
2.3. Variable Import and Export Pricing
Currently, the energy market has various time-differentiated retail pricing schemes, e.g., time-of-use (TOU), real-time pricing (RTP), critical-peak pricing (CPP) or curtailable/interruptible pricing tariffs, which reflect fluctuating wholesale prices and explore end-user demand flexibility [
26,
27,
28]. Various approaches are used to determine the dynamic import and export prices, ranging from game theory to blockchain applications [
29,
30,
31,
32]. The proposed framework aims at providing support with the design and operation of residential DER networks, and the design of pricing strategies is out of the scope of this approach. For this reason, the concept of dynamic import pricing is investigated through the use of the strategy described further. The scheme assumes that the dynamic import and export prices are obtained using the state-of-the-art approaches and will be provided by the retailers.
Equation (43) calculates the price, in British pence (p) per kWh, of electricity purchased from the grid into the DER network:
where
= the price of electricity purchased during the time (
m,
p) [p/kWh];
= the average electricity price [p/kWh];
= the national demand at time (
m,
p) [kWh];
= the yearly average national demand [kWh].
The price is based on data from the UK national grid for consumption throughout the year in 2017 [
16], which is shown in
Figure 3.
Overall, the price of electricity, taken as a time average, is
= 16.29 p/kWh. The calculated values of the price of purchased electricity are shown in
Table 5.
A similar equation is used to create a variable export price to determine the electricity sold to the grid from the DER network. The export price, given by the UK government [
24], was also scaled against the national demand at time (
m,
p) and the average national demand, as shown in Equation (44):
Overall, the price of exported electricity taken as a time average, is
= 5.03 p/kWh. The calculated values of the price of electricity sold to the grid are shown in
Table 5.
4. Conclusions
The short optimization times of the fixed network models is encouraging, since it demonstrates that optimization using dynamic pricing control is achievable and technologies should be able to rapidly and reliably respond to incoming demand and pricing data, even for time periods in the distant future.
Of course, in a real-world scenario the optimization might be completed for many more houses and other buildings across considerably greater distances, with many more variables to be considered, including real-time weather data and network failure updates.
However, in these circumstances the optimization would be run on specially designed software, on more powerful, purpose-built computers, possibly super-computers, and may also be completed for only a few days or even hours ahead of time, due to the inaccurate nature of weather or other data beyond this period.
The models used in this investigation have extremely low usage of grid imports and, consequently, the results into the use of dynamic import pricing are limited in revealing greater understanding of residential DERs, for all scenarios. However, this could not have been known until all simulations were run, i.e., the use of batteries or dynamic charging may have encouraged trading of energy.
The work is also limited by the small number of time slots used per day, the number of houses and demands and the use of average weather data, instead of, for example, a statistical model in which wind and solar data are derived from a Weibull distribution [
7]. A statistical distribution could also have been used for the demand data.
It would be useful to model a scenario in which the network has less access to self-generation and thus relies more on the grid, since the effect of dynamic import pricing is small due to the network’s ability to almost always meet its own demands. Dynamic FIT values could also be investigated to encourage stability from reliable electricity sources.
Scale-up effects of the scenarios investigated should be explored for increased number of homes, businesses and demand profiles across multiple regions or nations.
Critically, if dynamic pricing is to be used as a control strategy by national grids, more work is needed to develop the theory of how dynamic prices might be generated to meet the needs of the grid and how often they should be updated.
An alternative to using dynamic pricing could be the use of dynamic limitations on imports or exports. For example, during times of high demand, the grid might impose that a DER network cannot withdraw more than a set limit. This would force the DER network to supply its own energy. On the other hand, in times of high supply, for example on a windy, sunny summer’s day, the grid might impose that a DER cannot export more than a certain amount of electricity to the grid. More work is clearly needed to investigate if it is better to force generation and consumption or offer economic incentives, which make generation and consumption more likely.