IBM Research - Zurich
GmbH
Saeumerstrasse 4
8803 Rueschlikon
Switzerland
Authors:
email:
Date:
Version:
Fabian Mueller
Olle Sundstroem
Dieter Gantenbein
{fmu,osu,dga}@zurich.ibm.com
26.08.2014
3.0
Subject:
IBM EcoGrid Direct-Price Agent Implementation Status
1
1
Scope
This document specifices the IBM EcoGrid Direct-Price Agent v3.0 behavior, as an enhancement over
the versions 1.0 and 2.1 used in the 2013/14 heating season [1][2].
The new version 3.0 of the price agent aims at increasing the demand-response potential without
sacrificing any customer comfort. As introduced in this document, it distinguishes between heat
pump installations and direct electric heating installations. Moreover, it redefines the construction of
the upper and lower temperature bounds based on the user-set reference temperature and the chosen
flexibility level, as well as many timing parameters.
2
Definitions of temperature bands
The user specifies a reference temperature Tr in the GWR UI, see Figure 1 left screenshot (native
Danish language version used on Bornholm). Optionally, the user may also specify individual comfort
priority and flexibility level, see Figure 1 right screenshot:
Figure 1 GWR UI user settings: The user can decide whether or not to allow automation “Tillad styring”, and define individual
comfort priorities “Individuelle indstillinger”, and modify the reference temperature “Afbryd styring under” (stop control
below this temp). Comfort priority settings comprise the day time priority “Prioriter” {Morning, Daytime, Evening} as well as
the desired flexibility level “Profil” {Low, Medium, High}. They become active on Save “Gem”.
Flexibility Tflex(t) [°C]
The user-set flexibility levels {Low, Medium, High} translate into a temperature range Tflex(t) that
determines the allowed temperature above the reference temperature Tr. The mapping from flexibility
level to temperature range also depends on the comfort priority setting and, thus, can vary over time.
In case the user does not select individual settings, cf. left screenshot in Figure 1, a default
temperature range is set that is valid for the entire day; it corresponds to the “otherwise” in Table 1
which also provides detailed information about how flexibility and comfort priority settings are
translated to allowed temperature ranges, for all comfort priority settings.
Comfort priority
Morning
06:00-10:00 otherwise
Daytime
08:00-18:00 otherwise
Evening
17:00-23:00 otherwise
Default 3
3
3
3
3
3
Low
2
Medium 3
2
4
2
3
2
4
2
3
2
4
High
6
4
6
4
6
4
Table 1 Flexible temperature range Tflex(t) [°C] for different levels of flexibility and comfort priorities. These values apply to both
heat pump and electric heating installations. In case the user does not specify individual settings, the default values “otherwise”
are applied.
In addition to the user settings, the temperature set point for the heating devices’ internal thermostat
control Tth must be set such that Tth>=Tr+max{Tflex(t)}. This lower bound on Tth is required to guarantee
2
that the heating devices will turn on whenever they are enabled by the IBM price agent. However, Tth
should be chosen such that this temperature is acceptable in the worst case, see Section 5 for more
details.
3
Heat Pump Model and Constraints
The version of the direct-price agent described in this document uses a very simple heat pump model. It is
assumed that the heat pump consumes a constant amount of power PON when in operation, and a constant
amount of power POFF when idle.
The numeric power level values can be identified for every house individually. Table 2 provides a set of
typical values.
Heat Pump Operation State
ON
OFF
Constant Power
Consumption
4000W
130W
Table 2 Example of heat pump power levels.
The heat pump runtime constraints used in the new price agent version 3.0 are provided in Table 2 for
different flexibility levels. In contrast to previous versions of the direct-price agent, the maximum total offtime is now specified for 24 hours.
User-set Flexibility
Level
Low
Medium
High
Min. Off-time
Max. Off-time
Min. On-time
30min
30min
30min
30min
60min
90min
120min
90min
60min
Max. total Offtime per 24h
180min
360min
720min
Table 3 Heat pump (HP) operational constraints for different user settings.
4
Direct Electric Heating Model and Constraints
The model used for direct electric heating is similar to the heat pump model introduced above. It also
assumes constant power consumption if the device is ON. If it is OFF, no power is consumed at all.
Electric Heating Operation
State
ON
OFF
Constant Power
Consumption
4000W
0W
Table 4 Example of electric heating power levels.
The electric heating runtime constraints used in the new price agent version 3.0 are provided in Table 5 for
different flexibility levels. The dynamics of an electric heater are assumed to be very fast. Thus, only short
minimum ON/OFF times are considered.
User-set Flexibility
Level
Low
Medium
High
Min. Off-time
Max. Off-time
Min. On-time
5min
5min
5min
10min
15min
20min
5min
5min
5min
Max. total Offtime per 24h
180min
360min
720min
Table 5 Direct electrical heating (EH) operational constraints for different user settings.
3
5
Behavior of Price Agent
This section describes the behavior of the new price agent v3.0. There are 3 different behavioral modes
depending on the current temperature T(t), the thermostat set point Tth, and the user settings.
1. The temperature is below the reference temperature (T(t)<Tr):
In this case, the heating devices are constantly enabled.
2. The temperature is within the flexible temperature range
(Tr<=T(t)<=Tr+Tflex(t)): In this temperature range, the heating devices are
enabled or disabled with the goal of minimizing energy costs based on the
EcoGrid real-time prices. All the operational constraints as defined in Tables
3 and 5 are considered.
3. The temperature is above the allowed temperature range (T(t)> Tr+Tflex(t)):
In this case, the automation switches to the flexibility level `High’ that allows
for maximal OFF-times to correct over temperatures, but still under the
operational constraints of High as defined in Tables 3 and 5.
Disclaimer: Note that it is possible for the temperature to go below Tr or exceed Tr+Tflex(t). Even
when the agent will switch to behavior (3) in above list and run the automation with a flexibility level
`High’, decreasing of the temperature may not always succeed. The actual house indoor temperature
depends on many external factors. Examples include: open doors and windows, the current weather
situation, solar irradiation through windows, the thermal characteristics of the house, and also on
potentially other active but uncontrolled heating devices.
Figure 2 Definition of different temperature bands. The minimum required temperature T r is set by the user in the GUI. If the
actual temperature drops below this value, the heating devices are enabled constantly. The upper temperature limit is given by
adding the flexible temperature range Tflex(t) to Tr. If the temperature exceeds this value, the heating devices are controlled with
high flexibility with the aim of bringing the temperature below Tth again. A large thermostat set point temperature Tth, where
Tth>=Tr+max{Tflex(t)}, is required for the devices to start heating whenever they are enabled by the IBM price-agent.
4
6
Energy-Procurement Cost Minimization
Given a price forecast for a fixed planning horizon, the IBM direct-price agent version 3.0 computes a
heating device enablement schedule that (i) minimizes the total energy-procurement cost accumulated
over the planning horizon based on the EcoGrid real-time price and that (ii) simultaneously satisfies all
operational constraints as specified in Tables 3 and 5 depending on the device type.
7
Security Features
In addition to the operational constraints discussed in Sections 3 and 4, the following security features
are implemented to guarantee pilot participant freedom and best-possible efforts to establish an indoor
air temperature higher or equal to the user-set minimum comfort temperature at all times.
As soon as at least one of the following conditions is true, any planned minimum procurement-cost
enablement schedule will be pre-empted and the heat pump is unconditionally enabled to operate
according to its native parameters.
i.
ii.
iii.
iv.
v.
The indoor air temperature as measured by the GreenWave temperature sensor is below
the user-set minimum temperature Tr.
The latest indoor air temperature values received from the GreenWave temperature
sensor are older than 4 hours ago.
The communication between server and GreenWave gateway in house is lost for more
than 30 minutes (in this case the automatic scene will be executed).
Oestkraft or the participant operationally bail out the household from automation by
unchecking the optimize flag in the GreenWave user interface.
The IBM-agent administrative automation state for the household is set to disabled
(example as done over Christmas Holiday Season 2013/14).
References
[1]
Mueller, F., Sundstroem, O., Gantenbein, D. (2013). IBM EcoGrid Direct-Price Agent
Implementation Status v.1.0. IBM Research Zurich.
[2]
Mueller, F., Sundstroem, O., Gantenbein, D. (2014). IBM EcoGrid Direct-Price Agent
Implementation Status v2.1. IBM Research Zurich.
5