Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
D.Saker
1*
, P.Coker
2
, M.Vahdati
2
, S.Millward
3
, C.Carey
3
1
Technologies for Sustainable Built Environments, University of Reading, UK
2
School of Construction Management and Engineering, University of Reading, UK
3
SSE Plc, Reading, UK
* Corresponding author: d.saker@pgr.reading.ac.uk
ABSTRACT
In order to maintain a reliable electricity supply the System Operator must continuously
balance supply and demand on a second by second basis. Presently, this is achieved almost
entirely by utilising either part loaded power plant or else by using carbon intensive standby
generators. As the UK increases the amount of renewable energy connected to the grid, it is
likely that more balancing services will be required to maintain a stable electricity supply
system. This article demonstrates how the application of domestic demand response can help
mitigate the reliance on supply driven balancing services and therefore reduce carbon
emissions associated with new intermittent renewable generation.
The described method has the aim of time shifting domestic energy demand from electric
water heating without being disruptive to occupants usual lifestyles or comfort. A number of
dwellings using electric water heating had meters installed to characterise the diversity of
usage. Analysis shows that introducing dynamic immersion switching can provide more
distributed availability to the System Operator, in addition to reducing overall standing heat
losses from the hot water tank compared to how off peak switches are currently operated.
Using a Markov chain occupancy based approach a model is presented that simulates the
aggregated household load from electric water heaters, as well as the stored energy
availability derived from the hot water tanks. Results suggest this form of aggregated demand
response can help integrate renewable energy on the grid without major intervention.
Keywords:
Demand response, energy efficiency, domestic hot water, stochastic modelling
1. INTRODUCTION
Ageing power stations, concerns over air pollution, the risks associated with climate change,
and the increasing volatility of conventional fuel prices has resulted in the UK drafting a new
energy bill that firmly prioritises energy security and the implementation of low carbon
technologies. Due to the large UK wind resource availability and its cost competitiveness,
wind power has become the most attractive source of low carbon electricity supply. It is
predicted that by the year 2025 there will be 30GW of wind generation capacity on the UK
system, which is equivalent to half of the UKs current peak demand [1]. The continuous
growth of variable embedded generation and wind farms connected to the electricity grid,
presents a new set of challenges for the System Operator to overcome in keeping supply and
demand synchronised. In Great Britain this responsibility lies with the National Grid Plc, who
has already reported constraints on the network as a result of excess wind generation
occurring during low demand periods [2].
Aside from pumped storage, which has a round trip efficiency of 78% there is no other large
scale energy storage on the GB power system, for this reason National Grid must either
procure standby generation or else have access to flexible electrical load [3]. It is estimated
that at the moment between 95% and 98% of the ancillary services National Grid tender to
balance the network, are derived from electricity generation supply [4][5]. In contrast,
Demand Side Response (DSR) describes the time shifting or reduction of demand to balance
the power system via the ancillary services market. DSR requires energy to be converted into
its final form, rather than being stored in an intermediary form as is the case with grid
connected energy storage; hence it has a relatively high efficiency. Crucially, DSR can offset
the use of expensive and carbon intensive balancing services and reduces the requirement for
network reinforcement. The greater complexity associated with the domestic sector has so far
limited the use of DSR in households. Nevertheless the importance of utilising domestic DSR
in a low carbon future is echoed by many organisations, including energy regulator Ofgem
and National Grid [6][7]. There is a consensus that smart meters alongside time of use tariffs
or real time pricing will facilitate domestic DSR; however this is a cause for concern, because
there is no evidence that this approach is effective within the UK. Firstly, only automated
DSR can provide the necessary guarantees at the ancillary service scale, whereas the
deployment of time of use tariffs relying on customer interaction may in fact lead to greater
demand uncertainty. Secondly, the rollout of smart meters will not be completed until 2020 at
the earliest, which does not help domestic DSR provide a potentially vital role in forthcoming
years. This is particularly pertinent as the UK faces a drop in capacity margin to as low as 4%
in 2016, thereby increasing the risk of large scale interruptions to supply [8].
The potential for automated domestic DSR without behaviour change using readily available
technology provides the motivation for this research. In particular this paper investigates the
benefits and suitability of aggregated domestic electric water heating (EWH) in hot water
tanks for providing an ancillary service. To further assist the integration of future grid
connected renewable energy supplies and lessen the risk of curtailment, the feasibility of
providing a load on capability from multiple hot water tanks is also investigated. These
objectives are fulfilled by the development of a novel modelling tool that simulates the
aggregation of immersion heater load and energy storage potential at high resolution.
2. LITERATURE REVIEW
Presently the only way of utilising domestic DSR resources is through Radio Tele-switching
(RTS) technology that is installed in approximately 2 million households. RTS permits
remote load switching of electric storage heaters and immersion heaters via a long wave radio
data system. Originally developed in the 1980s as a way of providing overnight demand for
inflexible nuclear generation, RTS only offers one way communications, but it can transmit
signals virtually instantaneously. For energy suppliers and the System Operator RTS offered
important advantages over predecessor technologies that were usually in the form of time
clocks attached to the meter. Firstly, there was the ability to lessen the problem of cold load
pick-up and alleviate artificial demand peaks caused by meters simultaneously being switched
on. Secondly, there was no longer the operational issues associated with time drift and clock
changes. Thirdly, there was much greater flexibility and the twin element RTS system
provided a way of separating electric heating load from the rest of the household. There is
also evidence from an Ipsos Mori survey [6] that finds EWH is the most favoured method for
domestic customers to time shift their load. Despite this apparent willingness to engage and
the added capabilities RTS has offered, UK off peak tariffs have remained largely static in
their operation.
The ubiquitous nature of EWH in power systems and their intensive operation and energy
consumption makes them ideal components for direct load control schemes. Indeed the notion
of curtailing and time shifting domestic electric water heating is synonymous with the original
concept of demand side response that surfaced in the 1970s. The majority of previous studies
within this subject have focused attention on either: (i) energy efficiency analysis, (ii) demand
side response strategies and (iii) the associated issues with cold load pick up. Early models,
for example by Galiani et al, sought to physically describe single EWH systems as detailed
interactions between individual components [9]. Whilst this approach added greater
complexity compared to previous empirical based regression models, it was a necessary step
to analyse the effects from individual EWH perturbations. Malhame and Chong made
significant progress in this field by utilising computing power and building a physically based
stochastic model of EWH, thereby improving the load prediction capabilities [1014]. Their
combined extensive body of research characterises individual EWH load as a continuous state
and discrete state, which are set as tank water temperature and a thermostat controlled switch
respectively. Their output that is self described as a Markovian hybrid state model crucially
simulates hot water demand from the tank as a stochastic process with random amplitudes and
random switching times. When considering the aggregation problem, Malhame proposes a
system of coupled partial differential equations to describe the distribution of tank water
temperatures and therefore the fraction of EWH load in the on state. In most cases other
newer EWH models have used Malhame and Chongs stochastic aggregation framework, with
any variations usually centred on how the simulation of hot water usage is determined [15
17]. A key attraction to this aggregation modelling approach is that individual EWH
uncertainties can be characterised by a probability density function, replacing the need to have
detailed models for each household. However, as Paull points out, the ability to fully
characterise and load shift individual EWHs has several advantages, including the ability to
classify users on their hot water usage rather than household load usage as is typically done
[18]. Paull presents a EWH model that creates individual load simulations based on initially
disaggregating the EWH load from a large dataset (sourced from a Canadian smart meter
trial). In this way domestic hot water usage is estimated across households, allowing usage
profiles to be defined and used for creating future EWH load simulations and DSR control
algorithms. The model presented in the next section of this paper uses similar elementary
energy flow analysis as previously discussed models. Akin to Paulls model, empirical data
has been used to generate time series for individual households, however, a new methodology
is presented that creates hot water demand profiles based on an a Markov chain.
Table 1: Summary of main household sources of hot water in the UK [1923]
Heat Source Current
Stock
kWh/year
Advantages Disadvantages
Gas
Combination
(Condensing)
11,550,000
(6,312,000)
5,000 Easy to maintain,
Less space required,
Always hot water
available,
Low running costs
Low flow rates,
Does not offer any storage
opportunity, higher water
waste,
Poorer DHW output when
space heating required
Regular boiler
(Condensing)
10,910,000
(2,109,000)
4,840 Option for high flow rates,
Relatively inexpensive
Finite storage,
Maintenance issues,
Space required
Immersion primary
(primary and secondary)
2,500,000
(12,000,000)
3,500
1
Low capitol cost,
Easily maintained,
Off-peak option
High running costs,
Carbon intensive,
Finite storage
1
Based on homes using only immersion as the primary heating source, other homes use as secondary heating back up
3. METHODOLOGY
In order to simplify the modelling process the following assumptions are held:
The water in the hot water tank is perfectly mixed and the energy state changes are set
at discrete 1 minute intervals over a 24 hour period.
Each hot water tank has only one active EWH rated at 3 kW that can heat up all water
in the tank. Further work will incorporate an additional EWH element at the top of the
tank that is usually used to provide smaller hot water volumes quickly.
One single thermostat switches the EWH and the deadband temperature is fixed.
3.1. Thermal Representation
The thermal model of a single EWH is shown in figure 1, where the heat into the system is
; heat is lost in the system through either hot water demand H(t), or by conduction
(kWh/C); where
and
(1)
Whilst the thermostat temperature
(t)
S(t)
(t)
H(t)
Temperature
ON OFF
The information shown in figure 1 can be summarised into the following differential equation
describing each EWH:
(2)
Where;
(3)
(4)
In the above equations the tanks surface area is A (m
2
); vc(t) is the volume of hot water
consumed in litres over the time period t and
18 litres.
Remarkably the model output shown in Table 2 came out with the same result and a 95%
confidence of interval of