Method for predicting temporal-spatial indoor air pollution levels for vehicles

Abstract

The invention relates to a method for predicting the temporal-spatial indoor air pollution load (I) for vehicles (1, 2) of a vehicle fleet, wherein the indoor air pollution load (I) of at least a subset of the vehicles (1, 2) is continuously determined. According to the invention, it is provided that - the determined indoor pollutant load (I) is transmitted to a central computer unit (5) that is linked to the vehicles (1, 2) and a location-averaged time-dependent distribution map of the indoor pollutant load (I) is created, - the recorded wind information is transmitted to the central computer unit (5) and a location-averaged time-dependent wind distribution map is created in each case, - based at least on the wind distribution maps, a time-delayed model (M) for predicting the temporal-spatial indoor pollutant load (I) is created and trained and - the time-delayed model (M) is transmitted to the respective vehicle (1, 2) of the vehicle fleet.

Description

[0001] The invention relates to a method for predicting a temporal-spatial interior pollutant load for vehicles of a vehicle fleet, wherein an interior pollutant load of the vehicle is continuously determined.

[0002] DE 10 2020 003 282 A1 discloses a method for adjusting the outside air supply to the interior of a vehicle. Using signals recorded by a pollutant sensor located in the interior, the interior pollutant load is continuously determined while the vehicle is being driven. Outside air pollutant load is predicted for a section of road ahead of the vehicle, and the outside air supply is regulated depending on the predicted outside air pollutant load.

[0003] Furthermore, DE 10 2018 210 587 A1 describes a method for ventilating the interior of a vehicle. The method determines a future position of the vehicle, which the vehicle is expected to have reached at a future time. Furthermore, predictive air quality information for the future position at that future time is displayed. A ventilation device of the vehicle is controlled depending on the predictive air quality information.

[0004] The invention is based on the object of specifying a method for predicting a temporal-spatial interior pollutant load for vehicles of a vehicle fleet.

[0005] The object is achieved according to the invention by a method which has the features specified in claim 1.

[0006] Advantageous embodiments of the invention are the subject of the subclaims.

[0007] A method for predicting temporal and spatial interior pollutant levels for vehicles in a fleet provides for continuously determining the interior pollutant levels of at least a subset of the vehicles. According to the invention, the determined interior pollutant levels are transmitted to a central computer unit linked to the vehicles, and a location-averaged, time-dependent distribution map of the interior pollutant levels is created. Wind information acquired by the vehicle and / or otherwise is transmitted to the central computer unit and / or another processing unit, and a location-averaged, time-dependent wind distribution map is then created.Based at least on the wind distribution maps, a time-delayed model for predicting the temporal-spatial indoor pollutant load is created and trained, and the time-delayed model is transmitted to the respective vehicle in the fleet.

[0008] By applying the method, driving comfort for occupants of the respective vehicle in the vehicle fleet and protection of the health of the occupants can be increased, in particular by deactivating an external air supply of an interior ventilation system of the respective vehicle by means of an additional control unit if an interior pollutant load is expected to exceed, for example, a predetermined limit value.

[0009] In one embodiment of the method, the time-delayed model for predicting the expected interior pollutant load at a specific location in the vehicle is transmitted to the vehicle in the fleet. This means that, based on the model, the expected interior pollutant load of the respective vehicle can be determined for any location. Based on the predicted temporal and spatial interior pollutant load, a measure to prevent or reduce the interior pollutant load can be initiated by the vehicle and / or by the occupants of the respective vehicle.

[0010] Because the model is available to the vehicles in the fleet, the predicted, expected interior pollutant load on a route traveled by the vehicle can be continuously determined and, if interior pollutant load is expected, measures can be initiated in order to largely rule out the possibility of the interior pollutant load reaching a level that poses a health risk to the occupants.

[0011] In a further embodiment, depending on the forecast of the temporal and spatial interior pollutant load, a recommendation is issued in the respective vehicle to close vehicle openings and / or to activate the recirculation function of the vehicle's air conditioning system. These measures are intended, in particular, to prevent an unacceptable increase in the interior pollutant load in the vehicle as far as possible.

[0012] In one version, wind direction and wind speed are recorded as wind information and transmitted to the central computer unit. Based on this wind information, the dispersion of a pollutant, for example, when passing a business, can be determined, allowing it to be determined whether the vehicle is exposed to wind-dependent dispersion of the pollutant along its route.

[0013] In one possible implementation, the time-delayed model is trained using machine learning, whereby the model is trained to predict a measured temporal-spatial indoor pollutant load as a target variable. For example, a regression model is applied through reinforcement learning. A time delay and a temporal integration of an outdoor pollutant load are taken into account in the model, for example, by using time series information.

[0014] In a further embodiment, a measured value of the interior pollutant load is determined at a location at least when the vehicle window is open, in particular to ensure that at least one measured value of the interior pollutant load is available for each location, thus allowing the creation of a location-averaged, time-dependent distribution map of the interior pollutant load. This distribution map is then used to create the time-delayed model for predicting the temporal-spatial interior pollutant load.

[0015] Furthermore, one embodiment of the method provides that, to increase the number of measured values, a subset of vehicles in the vehicle fleet performs a test measurement of the indoor pollutant load at this location, and the measured values ​​are transmitted to the central computer unit. The more measured values ​​of indoor pollutant load are available at a location, the more accurately the wind-dependent dispersion of a pollutant and thus the temporal-spatial indoor pollutant load can be predicted using the time-delayed model.

[0016] Embodiments of the invention are explained in more detail below with reference to drawings.

[0017] Showing: Fig. 1 schematically shows a vehicle with a device for predicting interior pollutant levels and other vehicles of a vehicle fleet, Fig. 2 schematically show vehicles of the vehicle fleet at a location at different times and a diagram with aggregated location- and time-resolved wind and pollutant information of the vehicles of the vehicle fleet, Fig. 3 schematically shows a wind-dependent forecast of indoor pollutant levels and Fig. Figure 4 shows a further wind-dependent forecast of indoor pollutant levels.

[0018] Corresponding parts are provided with the same reference numerals in all figures.

[0019] Fig. 1 shows a vehicle 1 with a device for predicting an interior pollutant load I in the vehicle 1 and further vehicles 2, wherein the vehicle 1 and the further vehicles 2 belong to a vehicle fleet, for example of a vehicle manufacturer.

[0020] Fig. Figure 2 shows vehicle 1 and another vehicle 2 of the vehicle fleet at different times t1 to t4 at one location and a diagram D with aggregated location- and time-resolved wind information as well as interior pollutant loads I of vehicles 1, 2.

[0021] The Fig. 3 and Fig. 4 each show a forecast of the indoor pollutant load I to be expected at a location.

[0022] The vehicles 1, 2 each have at least one wind sensor 3 for determining wind conditions WV, wherein the wind sensor 3 detects signals during driving of the respective vehicle 1, 2, on the basis of which wind information, in particular a wind direction W and a wind strength WS, is continuously determined.

[0023] Furthermore, each vehicle 1, 2 has a pollutant sensor 4, which also continuously detects signals while the vehicle 1, 2 is in operation, based on which the interior pollutant load I of the respective vehicle 1, 2 is determined. It is also conceivable that an odor sensor for detecting odors is installed in the respective vehicle 1, 2.

[0024] In the following, a method for predicting an interior pollutant load I of the vehicle 1 is described, using a time-delayed model M, which is Fig. 4, a temporal-spatial interior pollutant load I is predicted for the vehicle 1 at a respective location.

[0025] The vehicles 1, 2 of the vehicle fleet are data-linked to a central computer unit 5, to which wind information and interior pollutant levels I determined by the respective vehicle 1, 2 are transmitted. For communication between the respective vehicle 1, 2 and the central computer unit 5, the respective vehicle 1, 2 has a communication unit 6.

[0026] In the central computer unit 5, the received wind information is aggregated and a location-averaged, time-dependent wind distribution map is created. Essentially, local wind information is available for each location, which was recorded at various previous times t1 to t3 at that location. For example, a time t4 represents a current time t4.

[0027] As described above, an interior pollutant load I in the vehicles 1, 2 is continuously determined based on signals recorded by the pollutant sensor 4 while the respective vehicle 1, 2 is driving. These measured values ​​reflecting the interior pollutant load I are a time-delayed and time-integrated function of a pollutant load S of an external environment of the respective vehicle 1. The interior pollutant loads I continuously determined while the respective vehicle 1 is driving are also transmitted to the central computer unit 5 via the respective communication unit 6.

[0028] In the central computer unit 5, the transmitted indoor pollutant levels I are aggregated and a location-averaged, time-dependent distribution map of the indoor pollutant levels I is created. For the most part, a local indoor pollutant level I exists for each location, which was recorded at different times t1 to t4. This information can be transmitted to vehicles 1 and 2 of the vehicle fleet.

[0029] Based on the spatially averaged time-dependent wind distribution maps, a time-delayed model M is trained using machine learning methods. The model M is trained to predict the measured temporal-spatial indoor pollutant load I as a target variable, for example, using a regression model with reinforcement learning.

[0030] In particular, the time delay and the time integration of the pollutant load S outside the respective vehicle 1, 2 of the vehicle fleet are taken into account, for example by using time series information, as described in the Fig. 3 and Fig. 4 is shown.

[0031] The model M trained in this way is transmitted back to the vehicles 1, 2 of the vehicle fleet, in particular to a control unit 7, and used in the respective vehicle 1, 2 to predict an expected future interior pollutant load I of the respective vehicle 1, 2 for a current position of the respective vehicle 1, 2. If, for example, a future interior pollutant load I is predicted which exceeds a predetermined limit value, a ventilation device for ventilating the interior of the corresponding vehicle 1, 2 can be controlled by means of a further control unit 8 such that an air supply to the vehicle 1, 2 is reduced and, for example, an air recirculation function is set.

[0032] Such a time-delayed model M implicitly allows a determination of a time delay between a comparatively high pollutant load measured by an indoor pollutant load I at a first time t1 and a location and a later second time t2 and another location.

[0033] Thus, it is possible to extract information from model M that goes beyond a passive diffusion model regarding the wind-dependent dispersion of pollutants, such as car exhaust or particulate matter, in a localized area, such as a city. In particular, it is possible to determine the dispersion of pollutants from a local source, such as an urban highway, depending on recorded wind conditions WV. In conjunction with a wind distribution averaged over a relatively long period of time, the influence of a local source of pollutants, particularly on various residential areas and / or pollutant monitoring stations, can be extracted and taken into account in urban planning measures.

[0034] In order to obtain a measured value of the interior pollutant load I of vehicles 1, 2 for each location within a specified period of time, even if an increased pollutant load S is temporarily predicted for this location and therefore the air supply from outside should be reduced, the measured values ​​of the interior pollutant load I are used, which were recorded with the vehicle windows and / or vehicle doors open. This recording occurs passively, in particular, when a vehicle window and / or vehicle door is opened. This means that no active opening of the vehicle window and / or vehicle door is required to generate the measured values.

[0035] To increase the number of measured values ​​from this location, in one possible embodiment, a subset of vehicles 1, 2 of the vehicle fleet can perform a test measurement and transmit it to the central computer unit 5. The subset is determined in such a way that the total load of the individual vehicles 1, 2 of the vehicle fleet is balanced.

[0036] Such a model M can be pre-trained during vehicle development, whereby the models M can also be trained and / or further trained in vehicles 1 and 2 of the vehicle fleet in a user-specific and / or region-specific manner.

[0037] General models M, user-specific and / or region-specific models M can be combined by means of distributed learning from the experiences of vehicles 1, 2 of the vehicle fleet using the central computer unit 5. The respective model M can then be transmitted back from the central computer unit 5 to the vehicles 1, 2 of the vehicle fleet.

[0038] Vehicles 1 and 2 of the vehicle fleet can generally be any vehicle on land, water, or in the air. In particular, they can also be a ship navigating a body of water, in which case wind conditions WV and pollutant loads S on the body of water are determined and sent to the central computer unit 5.

[0039] Fig. 2 shows the vehicles 1, 2 of the vehicle fleet at different times t1 to t4, whereby the respective vehicle 1, 2 records the wind conditions WV at a different time t1 to t4 at one and the same location and determines the interior pollutant loads I based on the pollutant load S. The wind information determined based on the wind conditions WV and the respective interior pollutant loads I are transmitted to the central computer unit 5 and aggregated there. Thus, a wind magnitude represented by a rhombus |v→s|, the wind direction W is represented by triangles and the interior pollutant load I by stars for each time point t1 to t4 based on the information from vehicles 1, 2 transmitted to the central computer unit 5.

[0040] The wind information and indoor pollutant loads I are aggregated in a spatially and temporally resolved manner so that a time-dependent wind and pollutant record is created for each location, as shown in diagram D.

[0041] In the Fig. The forecast of the interior pollutant load I shown in Figure 3 shows, by way of example, a route section A and the interior pollutant load I represented by grayscale. This was determined with time and location resolution based on the wind information transmitted by vehicles 1, 2 of the vehicle fleet to the central computer unit 5. In this exemplary embodiment, there is no wind. For locations that are not traveled to at all or comparatively rarely by vehicles 1, 2, the transmitted wind information is interpolated for a nearby location in order to forecast the interior pollutant load I.

[0042] Fig.4 shows a further forecast of an indoor pollutant load I. The indoor pollutant load I is forecast based on the aggregated wind information, with the wind direction W being shown in a middle graphic by means of arrows, a wind force WS being shown by means of a selected arrow size, and calm wind being shown by means of filled circles. The unfilled circles represent trees as an example, with an oval showing a lake as an example. From a predicted local distribution of the indoor pollutant load I, areas B1, B2 along the route section A can be determined, with the interior ventilation being able to be switched on in area B1 by means of the further control unit 8 and being switched off in another area B2 or being switched off automatically by means of the further control unit 8.

[0043] An application example of the method is described in more detail below: A driver of a high-quality vehicle 1, 2, in particular a so-called luxury vehicle, drives the same route to work every day. Depending on the weather conditions, the wind conditions WV near the ground change and an odor from a nearby sugar beet factory is blown onto a bypass leading to work. Now, many vehicles 1, 2 of a fleet drive in and around the sugar beet factory and detect both the local wind conditions WV and an odor development, in particular an interior pollutant load I. Using the model M created in the central computer unit 5, an odor development can be predicted based on these local wind conditions WV and also based on the locations and times t1 to t4, where an odor nuisance is to be expected on the way to work.In the area of ​​these locations, the air conditioning system of vehicle 1, 2 is set to recirculation mode so that the odors cannot penetrate into vehicle 1, 2. These locations of odor nuisance can change daily depending on local wind conditions and production activity at the sugar beet factory. List of reference symbols 1, 2 vehicle 3 wind sensor 4 pollutant sensor 5 central computer unit 6 Communication unit 7 Control unit 8 additional control units A section of the route B1, B2 area D diagram M model I Indoor pollutant levels S Pollution W Wind direction WS wind force WV wind ratio t1 to t4 time QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2020 003 282 A1

[0002] DE 10 2018 210 587 A1

[0003]

Claims

[1] Method for predicting a temporal-spatial interior pollutant load (I) for vehicles (1, 2) of a vehicle fleet, wherein an interior pollutant load (I) of at least a subset of the vehicles (1, 2) is continuously determined, characterized by , that - the determined interior pollutant load (I) is transmitted to a central computer unit (5) which is data-linked to the vehicles (1, 2) and a location-averaged, time-dependent distribution map of the interior pollutant load (I) is created, - recorded wind information is transmitted to the central computer unit (5) and / or another processing unit and a location-averaged time-dependent wind distribution map is created, - based at least on the wind distribution maps, a time-delayed model (M) for predicting the temporal-spatial indoor pollutant load (I) is created and trained, and - the time-delayed model (M) is transmitted to the respective vehicle (1, 2) of the vehicle fleet. [2] Method according to claim 1, characterized by that the time-delayed model (M) for predicting an expected interior pollutant load (I) at a respective location of the vehicles (1, 2) is transmitted to the vehicle (1, 2). [3] Method according to claim 1 or 2, characterized by that, depending on the forecast of the temporal-spatial interior pollutant load (I), a recommendation is issued in the respective vehicle (1, 2) to close vehicle openings and / or to set a recirculation function of an air conditioning system of the respective vehicle (1, 2). [4] Method according to one of the preceding claims, characterized by that a wind direction (W) and a wind strength are recorded as wind information and transmitted to the central computer unit (5). [5] Method according to one of the preceding claims, characterized bythat the time-delayed model (M) is trained using machine learning. [6] Method according to one of the preceding claims, characterized by that a measured value of the interior pollutant load (I) is determined at a location at least when the vehicle window and / or vehicle door of the respective vehicle (1, 2) is open. [7] Method according to claim 6, characterized by that a subset of vehicles (1, 2) of the vehicle fleet carries out a test measurement of the interior pollutant load (I) at this location and the measured values ​​determined are transmitted to the central computer unit (5).

Images