Method for optimising the energy expenditure and comfort of a building

Abstract

A method for optimizing the energy expenditure and the comfort of a building, including comfort systems provided with an online consumption sensor, local environment data sensors associated with an identifier of a zone of the building, and at least one server for collecting and recording the timestamped data remotely includes the following steps: constructing and saving a simplified digital model of the thermal behavior of the building; a step of calibrating the simplified digital model calculated during the preceding step; a step of validating the calibrated digital model calculated during the preceding step by comparing the digital variables obtained by predictive processing of the calibrated model and the digital variables stored by the server over a period of a few days; a step of calculating digital parameters for resource allocation by applying a Pareto optimum calculation applied to the validated calibrated digital model.

Claims

1. A method for optimizing energy expenditure and comfort of a building that includes: a plurality of comfort systems provided with an online consumption sensor, suitable for periodically remotely transmitting consumption data associated with an identifier of a comfort system of the plurality of comfort systems; and a plurality of local environment data sensors associated with an identifier of a zone of the building; wherein, the method comprises the following steps: constructing and saving on a remote server a simplified digital model of a thermal behavior of the building for which parameters of the simplified digital model are estimated using th-e-timestamped data remotely transmitted to the server by the online consumption sensors and by the plurality of local environment data sensors, by integration processing of characteristics of the building and the plurality of comfort systems depending on a location of the building, construction materials used in the building, the overall architecture of the building, and a chosen energy concept; a step of calibrating the simplified digital model calculated during the preceding step; a step of validating the calibrated simplified digital model by comparing digital variables obtained by predictive processing of the calibrated simplified digital model and digital variables stored by the server over a period of a few days; and wherein the parameters of the simplified digital model of the thermal behavior of said building are estimated during said step of constructing and saving a simplified digital model, using said timestamped data transmitted by said online consumption sensors as well as the timestamped data transmitted by said local environment data sensors, said method further comprises an additional step of calculating numerical parameters aimed at finding a compromise between minimizing total energy consumption and providing user-specified thermal comfort by applying a Pareto optimum calculation to said validated calibrated simplified digital model, for: optimization with historical target temperatures to determine parameters for reducing energy consumption without altering thermal comfort or optimization with new target temperatures to modify temperature setpoints to improve efficiency with a new thermal comfort reference.

2. The method of claim 1, wherein the Pareto optimum calculation is implemented by using a genetic NSGA-II algorithm.

3. The method of claim 1, wherein a criterion for the Pareto optimum calculation is determined by historical target temperatures.

4. The method of claim 3, wherein the Pareto optimum calculation is implemented by using a genetic NSGA-II algorithm.

5. The method of claim 1, wherein a criterion for the Pareto optimum calculation is determined by a set of new target temperature values.

6. The method of claim 5, wherein the Pareto optimum calculation is implemented by using a genetic NSGA-II algorithm.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) While this disclosure concludes with claims particularly pointing out and distinctly claiming specific embodiments, various features and advantages of embodiments within the scope of this disclosure may be more readily ascertained from the following description when read in conjunction with the accompanying drawings, in which:

(2) FIG. 1 is a schematic representation of an environment in which embodiments of the present disclosure may be implemented;

(3) FIG. 2 is a summary of the parameters linked to the building control strategy;

(4) FIG. 3 shows the calibration results in the table of parameters and time series; and

(5) FIG. 4 shows the result of the optimization with historical target temperatures.

DETAILED DESCRIPTION

(6) The present disclosure will be described in more detail with reference to a non-limiting embodiment.

(7) Material Architecture

(8) The description that follows presents an example of effective multi-objective methodology to improve energy efficiency and maintain thermal comfort, without intervention to renovate or modify the building envelope.

(9) The building comprises a plurality of comfort systems such as heating systems, light sources, air-conditioning systems, aeration systems, how or cold water supply points, etc.

(10) These systems are associated with sensors 10 communicating with a server 30 via the wired network or via a radiofrequency network to communicate information on the state of the associated system and on the main energy consumption points. The building also comprises local environment data sensors 20 transmitting data to the server 30 on a comfort parameter in the local zone where the sensor is installed.

(11) The sensors 10, 20 provide information in the form of digital sequences comprising an identifier of the sensor and at least one digital value for the parameter measured. The server 30 controls the timestamping of the data received and recording in permanent storage.

(12) The server 30 also receives and stores timestamped outdoor environment data, notably weather information from data sources.

(13) The data recorded by the server 30 are subject to processing in accordance with the method according to the disclosure, associating an energy program for the building with optimization processing.

(14) For example, the energy program may be a tool such as the Energy Plus (trade name) application developed based on the BLAST (trade name) and DOE-2 (trade name) tools and incorporating specific modules for introducing systems into the thermal zone energy balance and input and output data structures defined from the digital data recorded by the server 30.

(15) The energy program may consist of the specialized TRNSYS (trade name) application for dynamic thermal simulation applied to buildings. This application allows all the characteristics of a building and its systems (heating, air conditioning) to be incorporated in order to carry out a detailed mono-zone or multi-zone study of its thermal behavior. It incorporates variables for location, construction materials, overall architecture, and the chosen energy concept, including more complex systems such as innovative solar systems.

(16) The function of the optimization processing (single-objective or multi-objective) is to analyze the envelopes, orientations, shading or material characteristics and allow a diagnosis to be made.

(17) It may be carried out using the Global Python Parallel-PyGMO (trade name) toolbox applications constituting a multi-objective optimizer that allows a simplified model to be designed.

(18) The simplified model is obtained either based on a schematic view of the building, or following a complex campaign of time and resource measurement where trained professionals define the parameters that characterize the physical properties of the building.

(19) The parameters of the simplified model are then calibrated using measurements (temperatures, consumptions, programming, etc.) obtained from thousands of communicating sensors placed in a real building to store a very large number of real-time data. The physical parameters of the simplified model are estimated using PyGMO add-on software with the measurements and the CMA-ES algorithm.

(20) Next, the estimated model is validated using the TRNSYS (trade name) program to ensure that the resulting base model imitates the thermal behavior of the actual building.

(21) A multi-objective methodology to improve energy efficiency and maintain thermal comfort is then implemented acting only on the building management system without modifying the physical parameters.

(22) The NSGA-II approach is used to obtain the optimal Pareto parameters. The performance of the methodology is assessed based on data collected in a building situated in the Paris region.

Functional Architecture

(23) The first step of the method according to the disclosure involves designing a simplified model of the building for which the parameters are estimated using measurements obtained from communication-capable sensors.

(24) First Step: Definition of a Simplified Model

(25) The base model was implemented using the TRNSYS IT solution (Type 56 component) and taking account of a plurality of parameter types. Building managers usually know some of these parameters precisely whereas others are not known or not well understood.

(26) The single-zone base model is defined by the following components. A plurality of vertical external walls. Each of these walls is specified by the following parameters: surface area, proportion of windows relative to wall, orientation, thickness and constituent layers such as concrete, insulation, etc. A roof and a floor specified by: surface area and thicknesses of the various constituent layers, including the insulation. Maximum heating and air conditioning (AC) power available at the emitters and at central production. Schedules and temperatures for the heating, ventilation and air conditioning (HVAC) systems. A schedule consists of a start time, a time until shut down, a comfort temperature when occupied and lower temperatures when not occupied. Three (or four) program schedules are considered for each week: (i) Monday, (ii) Tuesday, Wednesday, Thursday, (iii) Friday (and (iv) weekend, if different from Friday). Other parameters characterizing thermal contributions inside the building, such as the number of occupants, the number of items of IT equipment, PCs, and the lighting systems characterized by a number of W/m.sup.2.

(27) In most cases, the structure of the walls, of the roof and of the low floors are fairly well known, as is the overall orientation of the building and the glazed surface area, but the thickness and the nature of the insulation is usually not well known and must be estimated to within a realistic range of values.

(28) Furniture that forms a substantial thermal energy store is summarized in a single parameter known as capacitance, expressed in kJ/? C./m.sup.3 and sized in proportion to the total volume.

(29) Second Step: Calibration of the Simplified Model

(30) As the objective is to improve thermal efficiency without renovation work, the calibration procedure must estimate the parameters linked to the building envelope. Some of these parameters are known or do not need to be calibrated, such as the structure of the external walls, of the roofs and of the low floors, or the window types. The other parameters linked to the building envelope required in order to define the TRNSYS model and the parameters linked to the building control strategy are summarized in the table shown in FIG. 2.

(31) Hereinafter, these parameters are designated by ?. The initial value of ? is chosen according to data determined by the building construction or redevelopment date.

(32) The estimation procedure uses the data recorded by the server 30 over a month, based on hourly readings. The data recorded each hour comprise outdoor temperature T.sub.e.sup.obs, average indoor temperature T.sup.obs measured in the building, energy consumption for heating Q.sub.h.sup.obs and for cooling Q.sub.c.sup.obs. The other data are recorded in a table of variables such as that shown in FIG. 2.

(33) The covariance matrix adaptation evolution strategy (CMA-ES) is implemented with the PyGMO (trade name) application toolbox to optimize iteratively the parameters in the table of variables using the selection (?, ?).

(34) At each iteration, the best descendant parameters (?, ?) from the actual estimation of the parameters are combined to form the population of the next iteration and the other candidates are rejected.

(35) For each parameter ?, the TRNSYS model is executed with the stored weather conditions to produce hourly energy consumptions and the associated indoor temperatures.

(36) The objective function minimized by the CMA-ES takes account of the difference between these hourly simulations and the actual observations measured in the building:
f.sub.calib:?.fwdarw.?.sub.t log(1+?T.sub.i.sup.??T.sub.i.sup.obs?.sub.2)+?.sub.q log(1+?Q.sup.??Q.sup.obs?.sub.2),
where: T.sub.i.sup.? and Q.sup.? designate the indoor temperatures and the total energy consumptions (heating, cooling and other expenditures) formed by the time series produced by the TRNSYS model with a given parameter ?, for any time series s, such that:

(37) ${.Math.s^{?}-s^{obs}.Math.}_2={\left(\frac{1}{n}\underset{k=1}{\overset{n}{.Math.}}{.Math.s^{?}\left(k\right)-s^{\mathrm{obs}}\left(k\right).Math.}^2\right)}^{1/2}$ where n equals the number of samples.
Step 3: Validation of the Model

(38) Once the CMA-ES algorithm has reached a state of convergence, the TRNSYS model is trained using observations to ensure calibration relative to the actual building.

(39) To ensure that the base model thermal behavior corresponds to the thermal behavior of the actual building, the model predictions are compared to the observations recorded for the week following the calibration period and for another subsequent period.

(40) All the parameters linked to the building envelope estimated during the calibration procedure are fixed and are considered as the building signature. Next, the parameters linked to the construction control strategy are fixed to the actual construction parameters for each validation period. The calibrated model is executed using these settings and the stored weather conditions and compared to the observations.

(41) Step 4: Pareto Optimization

(42) When the model has been calibrated and validated, Pareto optimization is carried out so that the energy performance of the building can be analyzed, by optimizing energy consumptions while maintaining a thermal comfort chosen by the model. The parameters used to improve energy efficiency are designated by ?. All the other parameters are defined by the parameters calibrated in ?.

(43) For each parameter ?, the TRNSYS model is executed with the stored weather conditions to produce the associated hourly energy consumptions and indoor temperatures for the following week. The objective function minimized by the NSGA-II algorithm aims to find a compromise between minimizing total energy consumptions and providing a thermal comfort specified by the user:

(44) $f_{optim}\text{:}?.fwdarw.\left(f_{optim}^1\left(?\right),f_{optim}^2\left(?\right)\right),\mathrm{with}$$f_{optim}^1\left(?\right)=\log \left(1+{\left(\frac{1}{p}{?}_{k=1}^P{.Math.T_i^{?}\left(k\right)-T_i^{*}\left(k\right).Math.}^2{1}_{k?I}\right)}^{1\text{/}2}\right),f_{optim}^2\left(?\right)=\log \left(1+{.Math.Q^{?}.Math.}_2\right),$
where T.sub.i* designates the sequence of indoor temperatures desired by the energy managers.

(45) Various optimization configurations may be envisaged. a) Optimization with historical target temperatures. In this case, the sequence T.sub.i* determined to allow adjustment to the temperatures observed in the building during the optimization period. Optimization aims to find parameters to reduce energy consumptions without changing the thermal comfort. b) Optimization with new target temperatures. In this case, the thermal comfort recorded with the sensors is assumed to be too conservative and the optimization procedure allows the temperature set points to be changed to improve efficiency with a new reference thermal comfort.

(46) Optimization is carried out using the NSGA-II method for multi-objective problems implemented in PyGMO, based on a non-dominated descendant selection procedure.

(47) Experimental Results

(48) According to one embodiment, the data used were collected in a 7-story office building of 14,000 m.sup.2 floor space situated in the Paris region for a total volume of 51,800 m.sup.3. Based on a commonly used rule, it is assumed that ? of the total area is occupied by people, which gives a total occupation area of 9240 m.sup.2. Assuming that each occupant has 12 m.sup.2, the initial values are set at 770 occupants and 770?1.2=924 PC in the building during the hours of occupation. The wall areas, on the other hand, are 3.7?7?50=1295 m.sup.2 and 3.7?7?40=1036 m.sup.2, respectively.

(49) In the results obtained, the model is calibrated using a month of data stored every hour such that n=720. The calibration results are shown in the table of parameters in FIG. 3. The CMA-ES algorithm is executed with a population size given by ?=4+3 log (card(?)) and the parameter ? by default. The parameters are initialized randomly within the interval given in Table 1 and the estimation procedure is repeated 50 times and stopped after 800 generations when the algorithm has reached convergence. The estimated average value and the standard deviation over 50 independent trials are given for each parameter in the table in FIG. 2. FIG. 3 shows the outdoor temperature evolution curve 5 (at the top) and the estimated indoor temperature curve 6 as well as the curve 7 showing estimated energy consumption.

(50) These estimated chronological series are compared with the observations from the building sensors. The last graph shows the relative error between the estimated chronological series and the observations over time. For any time series, this relative error is given for any 1?k?n, by:

(51) ${\mathrm{.Math.}}_s\left(k\right)=\frac{.Math.s^{\widehat{?}}\left(k\right)-s^{obs}\left(k\right).Math.}{s^{obs}\left(k\right)}$

(52) The estimated model is used to predict the weekly temperatures and consumptions after the calibration period (FIG. 3) and during a period N (FIG. 4).

(53) FIG. 4 shows the result of the optimization with historical target temperatures.

(54) The average temperature in the building during an hour of occupation was 23.4 degrees with a standard deviation of 0.67. Total energy consumption was 342.6 kWh. The results of the NSGA-II algorithm show that for a similar temperature volatility about 23.4 degrees, other construction parameters may lead to a total energy consumption of 300 kWh. The associated parameters are given in the table of parameters and the time series are displayed in FIG. 3. FIG. 4 shows optimization with new target temperatures. This demonstrates that significant gains may be obtained by reducing the target temperatures if the energy managers accept such a moderation of the target temperature.

(55) The method according to the disclosure allows the lack of information and data imprecision inherent to any building in real use to be overcome in a profitable and generalizable way. It also allows a dynamic physical thermal model to be produced that is very close to the real operation of the building (usually to within a few percentage points of reality). Furthermore, it allows explicit results to be obtained on the improvement actions that should be undertaken (system settings, programming, building work, energy supply contract optimization) and allows the impacts of the improvement actions to be quantified in terms of enhanced comfort and energy efficiency.