Control of building technical equipment

Abstract

The invention relates to a computer-implemented method for determining a parameter setting of a building technology installation, comprising: receiving an environmental condition state detected by means of at least one sensor device; determining a number of environmental condition states that have been stored in the database and that satisfy a similarity criterion related to the detected environmental condition states; and, if the number is less than a predetermined number threshold value, determining a parameter setting of the building technology installation according to at least one predefined parameter setting rule; alternatively, if the number is greater than or equal to a predetermined number threshold value, parameter settings of the building technical equipment are determined according to a machine learning model.

Description

Technical Field

[0001] The invention generally relates to the technical field of building technology installations and in particular to the control of building technology installations, which may include, for example, sun protection installations and / or lighting installations. Background Art

[0002] Technical building equipment includes all permanently installed technical equipment inside and outside a building, which serves the functional use of the building. For example, such equipment is necessary for the function of the building or increases the comfort and / or safety of the users. For example, technical building equipment includes sun protection equipment, lighting equipment, fire alarm equipment, intrusion alarm equipment, photovoltaic equipment or access control equipment. Technical building equipment is sometimes also called residential technical equipment, technical building equipment or supply technology.

[0003] The installation of building technology devices on buildings is usually carried out by specialists. In addition to the actual devices, operating elements are provided so that the user can make his own settings, i.e. adapt the parameter settings of the corresponding building technology devices to his own preferences, in particular to the user's preferences determined by the situation or to preferences determined by environmental conditions.

[0004] Furthermore, rule-based control tools are often provided for building technology devices, which, for example, determine the parameter settings of the respective building technology device based on predefined rules and measured values ​​of sensor devices. The sensor devices are provided for measuring environmental conditions and can include, for example, light sensors or temperature sensors. Depending on the configuration, the user can be given the possibility to influence the rules and adapt them to their preferences. The purpose of such rule-based control tools is to increase the user's comfort and reduce the need for operating operating elements.

[0005] However, for many users, determining rules is a challenge. To determine rules or to change preset rules so that the rules fully reflect the user's preferences, it is first necessary to understand the user's preferences, such as a specific room temperature or a specific light intensity. In addition, these user preferences need to be correctly converted into rules. The conversion is usually performed by the user himself and is not supported. Usually, a room is used by different people, who may have different preferences. Therefore, in practice, when using building technology equipment, it is usually necessary to manually operate the operating elements to match the user's preferences. In other words: the possibility of determining rules or changing preset rules only slightly reduces the need for manual operation of operating elements. Summary of the invention

[0006] Against this background, the object of the present invention is to provide a method which increases the user comfort of a building technology installation and thus at least partially overcomes the above-mentioned disadvantages of the prior art.

[0007] The object of the invention is achieved by the subjects of the independent claims.

[0008] According to the invention, a computer-implemented method for determining parameter settings for a building technology device is provided. The method may include receiving an environmental condition state detected by means of at least one sensor device. The method determines the number of environmental condition states stored in a database that meet a similarity criterion associated with the detected environmental condition state. If the number is less than a predefined number threshold, the method may include determining the parameter settings for the building technology device according to at least one predefined parameter setting rule; or, if the number is greater than or equal to a predefined number threshold, the method may include determining the parameter settings for the building technology device according to a machine learning model.

[0009] Technical building equipment can be stationary or mobile technical equipment installed inside or outside a building, which serves the functional use of the building. Examples of technical building equipment are, in particular, sunshade equipment, air conditioning equipment, heating equipment, lighting equipment, fire alarm equipment, intrusion alarm equipment, photovoltaic equipment, access control equipment or equipment combining a plurality of the above functions in one device. The computer-implemented method is preferably provided for determining parameter settings of technical building equipment designed as sunshade equipment and lighting equipment.

[0010] The environmental condition state may relate to one or more measurement time points at which the environmental state of the building technical equipment was measured, such as temperature, light conditions or wind force and / or wind direction. The environmental condition state may also include generally available information about the environmental state of the building technical equipment, such as time or date. Furthermore, the environmental condition state may include currently existing parameter settings or settings of the corresponding building technical equipment. In this context, the environmental condition state is to be understood as one or more environmental conditions that are perceptible or verifiable by a user.

[0011] The at least one sensor device can be configured to detect the state of the environmental condition in the form of measurement data and can in particular have a plurality of sensors, for example a first temperature sensor for the temperature outside the building, a second temperature sensor for the temperature inside the building, a light sensor and / or a wind sensor. In particular, a scaling of the state of the environmental condition and / or the parameter settings can be provided. Particularly preferably, the measurement data detected by means of the at least one sensor device are scaled. The state of the environmental condition and / or the parameter settings can be scaled in a value range between 0 and 1, for example when the state of the environmental condition and / or the parameter settings are present as vectors.

[0012] For example, an environmental condition state x_1 may contain five information x_1_1, x_1_2, x_1_3, x_1_4, x_1_5 about the environment and may be present in the form of a vector of measured values. The values ​​x_1_1, x_1_2, x_1_3, x_1_4, x_1_5 may preferably be fully or partially scaled, or scaled after measurement or recording, in particular scaled over a value range between 0 and 1. For example, the value x_1_1 may be a scaled measured value of a light sensor. For example, the value x_1_2 may be a scaled measured value of a temperature sensor. For example, the value x_1_3 may be a scaled measured value of a wind sensor. For example, the value x_1_4 may be an unscaled measured value of a room occupancy sensor that detects information about the number of people in the corresponding room. For example, the value x_1_5 may be an unscaled time.

[0013] In the sense of the present invention, a distinction can be made between detected environmental condition states and stored environmental condition states. The detected environmental condition states can be environmental condition states that exist at the current point in time, in particular when the method is running. The stored environmental condition states can be environmental condition states that have been stored in a database, for example, the environmental condition states can already exist at an earlier point in time or have been pre-given for the commissioning of the building technical equipment. Therefore, the environmental condition states detected at an earlier point in time or provided in the preparation phase can be stored in the database. In particular, the stored environmental condition states can be stored in pairs in the database with parameter settings related to the building technical equipment. The parameter settings of the environmental condition states include, for example, the settings of the sunshade device and the settings of the lighting in the room.

[0014] In the method step of determining the number of environmental condition states stored in the database that meet the similarity criteria associated with the detected environmental condition states, the difference and / or similarity of the detected environmental condition states and the stored environmental condition states can be checked. When the environmental condition states exist as vectors as described above, the check can be performed by a distance calculation of the vectors, wherein the modulus of the vectors is compared, in particular, by a difference calculation of the moduli of the vectors. Alternatively or additionally, other distance calculation methods common in vector calculations can also be considered, in particular Euclidean distance calculations. Alternatively, the differences in the individual values ​​of the vectors can be calculated and considered by a difference calculation of the vectors. For each pair of environmental condition states that are calculated, the similarity criteria can be checked. Depending on the type of distance calculation method, different similarity criteria can be used. In addition, both the distance calculation method and the similarity criteria can vary depending on the type of building technical equipment to be controlled. In particular, a variety of distance calculation methods and / or similarity criteria can be provided, from which they can be selected.

[0015] In the case of the Euclidean distance of the environmental conditions, an example for a similarity criterion is that the Euclidean distance is below a certain threshold value. The number of stored environmental condition states that meet the similarity criterion is preferably determined by simple counting.

[0016] If the number is less than a predefined number threshold, the parameter setting of the building technical equipment can be determined according to at least one predefined parameter setting rule. In this way, the parameter setting or adjustment of the settings of the building technical equipment is advantageously ensured in the absence or lack of a data basis in the form of stored environmental condition states. For example, if sunshade devices on the windows of a building have only recently been put into operation, there are only a few previously stored environmental condition states in the database. If the sunshade device is parameterized or set in such a way that it is completely or almost completely retracted, sunlight can enter the room unhindered. At times of day when the sunlight exposure is high, it is generally expected that the sunshade device can shield the room from the sunlight to a certain extent. According to the predefined parameter rules, the sunshade device is at least partially deployed when the light sensor detects high sunlight exposure.

[0017] If the number is greater than or equal to a predetermined number threshold, the parameter settings of the building technical equipment are determined according to the machine learning model. The machine learning model is capable of learning the preferences of one or more users, so that as long as a sufficient data basis is given, the parameter settings of the building technical equipment are closer to the actual preferences of one or more users with the help of the machine learning model. As explained at the beginning, if the number of stored environmental condition states that meet the similarity criteria related to the detected environmental condition state is greater than the number threshold, there is a sufficient data basis. It should be mentioned at this point that the stored environmental condition states and the respectively related stored parameter settings can exist in pairs in the database. In other words: as long as there are enough user preference experience values, the machine learning model can be used. These experience values ​​can be gradually detected over a longer period of use after being put into use. For example, each time a user manually changes the parameter settings or settings of the building technical settings with the help of an operating element, these changes are stored in such a way that the current environmental condition state and the manual parameter settings are stored in pairs in the database.

[0018] The machine learning model can therefore replace the predefined parameter setting rules after a period of time. Since the predefined parameter setting rules may in particular include rules for, for example, average preferences of the population, these parameter setting rules cannot fully reflect the preferences of a specific user or specific users of the building technical equipment. However, after there are enough stored environmental condition states in the database for access, the machine learning model can reflect these preferences with relatively high accuracy. In other words, if an environmental condition state is detected and there are too few similar stored environmental condition states (and related parameter settings) relative to the environmental condition state, the functionality of the building technical equipment is ensured by applying the predefined parameter setting rules. As a result, even directly after the newly installed building technical equipment is put into use, a minimum user comfort is provided, so that the user does not have to make each manual setting via the operating element each time. However, once there are enough stored environmental condition states (and related parameter settings) that are sufficiently similar to the detected environmental condition state relative to the detected environmental condition state, the machine learning model determines the parameter setting or settings of the building technical equipment and performs the parameter setting close to the user's preferences in an advantageous manner.

[0019] The proposed operating principle becomes more apparent with regard to the service life of a building technology device, starting from the initial installation. Initially, there are no stored environmental condition states (and associated parameter settings), but only predefined parameter setting rules, which can be specified, for example, by the manufacturer of the building technology device and can be for an assumed average user. During operation, the environmental condition states are detected continuously or at discrete time intervals by a sensor device and reference is made to the predefined parameter setting rules. This can be regarded as the first stage of the service life of the building technology device. After a period of time, the user (for example, the resident of the building for which the building technology device is intended) has made manual settings with the aid of an operating element, or if the parameter settings made according to the predefined parameter setting rules meet their preferences, no manual settings are made. With manual settings and without manual settings (for example, for a predetermined time period), the stored environmental condition states (and associated parameter settings) are stored in a database. In the case of frequently repeated detected environmental condition states, these environmental condition states can be called up with the aid of a machine learning model. In the case of rarely occurring detected environmental condition states, the predefined parameter setting rules continue to be relied on. In this stage, which can be regarded as the second stage of the service life of the building technology device, a machine learning model and predefined parameter setting rules are needed and used. After a further period of time has passed and the machine learning model can call up a sufficient number of stored environmental condition states (and associated parameter settings), the machine learning model is primarily used, ideally only used. This state can be regarded as the third phase of the service life of the building technology equipment. If a change of user is to occur (e.g. a change of tenants in a building), the machine learning model can gradually adapt to possible new preferences in the manner described above. Preferably, a reset function can be provided for the database of stored environmental condition states (and associated parameter settings).

[0020] The determination of the parameter settings can be carried out in particular by means of a control tool of the building technology device. For example, the control tool can be a central control tool having a calculation unit and a measurement data receiving interface, via which the measurement data of the at least one sensor device are received. Alternatively or additionally, the determination of the parameter settings can be supported by a server, wherein in particular a database can be stored on the server. It is particularly advantageous if the calculations are at least partially carried out on the server, for example, to determine the number of stored environmental condition states that meet a similarity criterion and / or to determine the parameter settings according to predefined parameter setting rules and / or to determine the parameter settings according to a machine learning model. Particularly preferably, all calculation steps are carried out by means of a server. In the case of using a server, the building technology device is communicatively connected to the server, in particular via a data network.

[0021] As explained in the above example, the training of the machine learning model can be specified in particular as follows:

[0022] In response to manual parameter settings of building technical equipment or parameter settings that have not been made within a predefined time period, environmental condition states and related parameter settings of the building technical equipment can be stored in a database, and training of a machine learning model can be performed using at least a portion of the environmental condition states and related parameter settings in the database.

[0023] The training of the machine learning model can also include training / test validation and / or cross-validation partitioning of the database. Training / test partitioning and cross-validation partitioning are respectively known methods in the field of machine learning, which can combat the so-called overfitting and underfitting problems of machine learning models. Here, the goal is to find the best choice for the machine learning model, which does not include overfitting or underfitting as much as possible. Overfitting of the machine learning model exists when the model is too targeted at the existing training data and thus has difficulty classifying new observations. Machine learning models with overfitting react especially to stored data and with a higher error rate in the case of new observations. In the field of machine learning, this is called unnecessary noise consideration. If the stored data is not fully taken into account, underfitting exists. This may also lead to an increase in the error rate of the machine learning model.

[0024] Preferably, the method further comprises sending a command in order to change a currently existing parameter setting of the building technology device to a specific parameter setting, wherein preferably the sending is only performed if the specific parameter setting deviates from the currently existing parameter setting by a predefined parameter setting threshold value.

[0025] Sending a command to change a parameter setting to a specific parameter setting can be carried out in particular by means of a control tool for a building technology system. For example, based on a specific parameter setting, a command can be sent to fold up or unfold a sunshade device. By setting a parameter setting threshold, it is ensured that slight deviations between a specific parameter setting and the currently existing parameter setting are ignored. This is particularly advantageous with regard to the energy requirements of the building technology system, since no unnecessary changes occur. The comfort of the user is also increased by the parameter setting threshold, since the user is not disturbed unnecessarily by possible noise (for example when folding up or unfolding a sunshade device). In this respect, by setting a parameter setting threshold, conflicts of interest can be resolved in a simple manner, for example in the case of a sunshade device between an optimal lighting state and a noise background that is as low as possible, allowing for concentrated work.

[0026] It can also be provided that predefined constraints, in particular safety constraints, are (always) observed when determining the parameter settings.

[0027] Within the framework of legal regulations, which may include, for example, fire protection regulations, it is often not possible to establish optimal conditions from the user's point of view. In other words, the optimal situation is limited by legal regulations. In addition, the user may also want a certain theft protection within the framework of his security needs, wherein said needs can only be ensured with non-optimal parameter settings of the building technology equipment. By setting predefined constraints, these legal regulations or other mandatory preferences are always taken into account when determining the parameter settings, regardless of whether the parameter settings are determined with the help of predefined parameter setting rules or with the help of machine learning models.

[0028] Particularly preferably, the method further comprises: receiving presence information and / or identification information, wherein the presence information comprises at least information about the number of persons in a room, a building technology device is set up for the room, wherein the identification information comprises at least information enabling clear identification of persons located in the room or persons entering or leaving the room, and wherein the presence information and / or identification information is taken into account for determining parameter settings.

[0029] Furthermore, it can be provided that, based on the presence information and / or based on the identification information, at least one predefined parameter setting rule and / or machine learning model is selected from a plurality of parameter setting rules or a plurality of machine learning models for determining the parameter setting.

[0030] The presence information contains information about how many people are in a room or in a building and can perceive the parameter settings of the building technology system. For example, this presence information can be detected at an entry point such as a door using an entry sensor. Such an entry sensor can be designed as part of the sensor device. The presence information can be part of the detected environmental condition state and / or a stored environmental condition state.

[0031] The identification information contains information that enables the unambiguous identification of the person present in a room or a building. In this regard, the unambiguous identification does not necessarily necessarily include personal data from which the person's identity (name, appearance, etc.) can be inferred. Unambiguous identification of a person only requires differentiation from other persons and can also be carried out using the unique number (UID) of the access authorization medium (e.g. access card). For example, the identification information can be detected at an access control device at an entry point such as a door. Such an access control device can be constructed as part of a sensor device. The identification information can be part of a detected environmental condition state and / or a stored environmental condition state.

[0032] By precisely adapting to the occupancy status of the room or building, the presence information can be advantageously taken into account. For example, if there are a large number of people in a room, the air conditioning system of the room can be parameterized in such a way that the dehumidification can be increased even before an increase in air humidity is measured by the humidity sensor. This allows for a predictable determination of the parameterization.

[0033] Taking into account the identification information also enables the personal preferences of different users to be taken into account. In particular for so-called "shared offices" (shared workspaces), situations with different users can be envisaged. For this purpose, it can be provided in particular that different predefined parameter setting rules and / or machine learning models are set for different people. If there are multiple people with different preferences in a room or a building, it can preferably be provided that, in order to determine the parameter settings, a rule or a machine learning model is selected which corresponds to as many different preferences as possible. Particularly preferably, it can be provided that, in this case, a temporary rule or a temporary machine learning model is formed, on the basis of which the parameter settings are determined. In particular, the temporary rule or the temporary machine learning model can reflect the best average value of the preferences of the different people in the room or the building.

[0034] Preferably, it is provided that the state of the environmental condition detected by means of the sensor device comprises at least one of the following information:

[0035] The measured value of the light sensor,

[0036] The wind sensor's measurements,

[0037] The measured value of the temperature sensor,

[0038] The difference between temperature and perceived temperature,

[0039] Measurements from room occupancy sensors,

[0040] time,

[0041] The measured values ​​of a touch sensor, which is preferably located on the window handle,

[0042] Date, preferably without year,

[0043] The condition of building elements, preferably windows,

[0044] The condition of the air conditioning equipment,

[0045] The state of the heating equipment,

[0046] The status of the shading equipment,

[0047] the status of lighting equipment;

[0048] and / or parameter settings of a building technology system, said parameter settings comprising at least one or more of the following parameter settings:

[0049] of a building element, preferably a window,

[0050] Air conditioning equipment,

[0051] Heating equipment,

[0052] Sunshade equipment,

[0053] Lighting equipment.

[0054] Furthermore, it can be provided that a similarity criterion of the stored environmental condition states is checked according to a cell-lists algorithm, wherein the similarity criterion for the stored environmental condition states is preferably met only when the difference between the detected environmental condition state and the stored environmental condition state does not exceed a certain value.

[0055] The use of a cell-list algorithm provides an efficient method with which the similarity criterion can be checked. Thus, the number of stored environmental condition states which are sufficiently similar to the detected environmental condition state can be determined in a simple manner.

[0056] According to the present invention, a computer-implemented method for training a machine learning model is also provided. The machine learning model may be configured for use in the above method. In response to manual parameter setting of a building technology device or no parameter setting for a predefined time period, the method may include storing the state of the environmental condition and the associated parameter settings of the building technology device in a database. The method may include using at least a portion of the state of the environmental condition and the associated parameter settings in the database to train the machine learning model.

[0057] The method for training a machine learning model can in particular be provided for (self-learning) sunshade control and lighting control by means of correspondingly constructed building technology equipment. In the following, the method is explained in the example of a building technology equipment constructed as a sunshade device. Taking into account the service life of the sunshade device, starting from the initial installation, there are initially no stored environmental condition states (and the associated parameter settings), but only predefined parameter setting rules, which can be specified, for example, by the manufacturer of the building technology equipment and can be aimed at an assumed average user. During operation, the environmental condition states are detected continuously or at discrete time intervals by a sensor device and the predefined parameter setting rules are called. In some cases, the generated parameter settings conform to the user's preferences, but in some cases not.

[0058] If the parameter setting meets the user's preferences, the user will not change the parameter setting via the operating element of the sunshade device. For example, in order to train a machine learning model, after the parameters have been determined according to predefined rules, the time when the user has not made changes via the operating element can be detected. If the user has not made changes within a certain time, it can be inferred that the parameter setting meets the user's actual preferences under the current environmental condition state. The detected environmental condition state and the related parameter settings are stored as stored environmental condition states in the database and used as training data for the machine learning model.

[0059] However, if the parameter setting does not conform to the user's preferences, the user manually changes the parameter setting through the operating element and, for example, the sunshade device is further retracted or further unfolded. For example, in order to train a machine learning model, after manual parameter setting by the user, it can be inferred that the new parameter setting conforms to the user's preferences under the current environmental condition state. The detected environmental condition state and the related parameter settings are stored in the database as stored environmental condition states and used as training data for the machine learning model. Alternatively or additionally, after manual parameter setting by the user, the time when the user does not make any subsequent changes through the operating element can be detected. If it is detected that no further changes are made through the operating element within a predefined time period after the manual parameter setting, it can be inferred that there is now a parameter setting that conforms to the user's preferences for the current environmental condition state. The detected environmental condition state and the related parameter settings are stored in the database as stored environmental condition states and used as training data for the machine learning model.

[0060] After a period of time, the users (e.g. residents of a building equipped with sunshade devices) have generated enough training data so that the need to make manual changes via the operating elements becomes less and less. If the users change (e.g. the tenants of the building change), the machine learning model can gradually adapt to possible new preferences in the manner described above.

[0061] Furthermore, a storage cleanup can be provided, wherein a check is performed to determine whether the number of stored environmental condition states exceeds a predetermined maximum value. In particular, the maximum value can be the maximum permissible number of stored environmental condition states. With the aid of storage cleanup, balanced data quality and data quantity are ensured, thereby ensuring the proper functioning of the machine learning model.

[0062] In the context of storage cleaning, in particular, provision can be made for the use of a cell-list algorithm. In this case, it can be particularly advantageous to specify a maximum value for each cell in the cell-list algorithm. This is particularly advantageous because by specifying a maximum value for each cell, data gaps in the sense that no stored environmental condition states exist, for example for environmental condition states that occur rarely, are avoided. In other words: it is ensured that a sufficient data basis for stored environmental condition states is maintained over the entire defined range of detectable environmental condition states.

[0063] If the number of stored environmental condition states exceeds a maximum value or a maximum value within a cell of the cell-lists algorithm, the oldest stored environmental condition state can be deleted until the number of stored environmental condition states no longer exceeds the maximum value, thereby supervising the machine learning model in a simple manner.

[0064] Furthermore, it can be provided that, when using the method for training a machine learning model for determining parameter settings of a building technology device in one of the above-described methods, the operation of an operating element of a building technology device by a user and / or the occupancy of a room as well as the non-operation of an operating element within a predefined time period are detected and in particular the current environmental condition state (with the associated parameter settings) is stored in a database. It can be particularly preferably provided that, when another criterion is met, a certain environmental condition state (with the associated parameter settings) stored in the database is given a higher or lower weight than other environmental condition states (with the associated parameter settings) of the environmental condition states stored in the database. For example, a further criterion leading to a higher weighting of a stored environmental condition state (with the associated parameter settings) can be the non-operation of an operating element within a further, longer time period.

[0065] It can also be provided that the training of the machine learning model is performed with the aid of supervised and / or reinforcement machine learning.

[0066] Especially in the context of supervised training machine learning models, it is possible to set the stored environmental condition states (and related parameter settings) to use training / test validation and / or cross validation partitioning (see above). Additionally or alternatively, nearest neighbor classification can be specified, preferably based on a cell-lists algorithm. Nearest neighbor classification is a common parameter-free method in the field of density functions, which enables joint category assignment of adjacent values ​​(current environmental condition states). Particularly preferably, neural networks and / or decision trees can be used. Compared with nearest neighbor classification, they have the advantage of better identification and ignoring unimportant measured values.

[0067] For example, in the context of reinforcement machine learning, a weight distribution can be performed, similar to the "reward and penalty" principle. Here, a reward may result in a higher weight for a stored environmental condition state, while a penalty may result in a lower weight for a stored environmental condition state. In particular, only a higher weight (reward) or a lower weight (penalty) can be set. In particular, as the duration of time during which the user does not operate the operating element increases, a reward and therefore a higher weight for the environmental condition state may be triggered. For example, a reward or a higher weight may continue to increase with the duration during which the operating element is not operated. In particular, when the user operates the operating element, a penalty or a lower weight for the environmental condition state may be triggered. The more frequently the user operates, the greater the penalty or the lower weight may be.

[0068] Reinforcement machine learning preferably includes criteria that combine certain user behaviors (especially operations of operating elements) with rewards or higher weights for environmental condition states or with penalties or lower weights for environmental condition states. In particular, these criteria can be further developed within the framework of their own standard learning model in order to optimize the conditions and amplitude of weight changes. Dead times that occur, for example, during heating and cooling processes can thus also be taken into account in an advantageous manner. Dead times may occur in particular due to differences between perceived temperature and actual temperature and may, for example, be the time that elapses before perceived temperature and actual temperature become consistent.

[0069] Preferably, a reset function for the machine learning model is provided, wherein in particular stored training data are deleted.

[0070] According to the invention, an apparatus, in particular a building technology installation, is provided, which comprises means for carrying out one of the above-described methods.

[0071] In particular, the device can have operating elements. These operating elements are preferably designed to detect subjective inputs of the user via preferably unique input values.

[0072] According to the present invention, there is provided a computer program comprising instructions which, when the program is executed by a computer, cause the computer to execute one of the above-mentioned methods.

[0073] The technical advantages and embodiments described with respect to the method according to the invention also apply to the device according to the invention and the computer program according to the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The preferred embodiments of the present disclosure are described below with reference to the accompanying drawings:

[0075] Figure 1 A first schematic diagram shows an embodiment of the present invention.

[0076] Figure 2A second schematic diagram shows an embodiment of the present invention. DETAILED DESCRIPTION

[0077] Figure 1 The present invention is a highly simplified illustration of an embodiment of a method according to the invention for controlling a technical building system. The technical building system may be a system having the functions of a sun protection system, an air conditioning system, a heating system, a lighting system, a fire alarm system, an intrusion alarm system, a photovoltaic system or an access control system.

[0078] Figure 1 A control device 20 is shown which is configured to carry out the method. For example, the control device 20 can be designed as part of a building technology installation or as a remote control, wherein the building technology installation has a communication connection to the remote control. Figure 2 A higher level of detail regarding control device 20 is included.

[0079] The input values ​​of the control device 20 are: the state of the environmental conditions recorded by means of at least one sensor device 11, the operation or non-operation of the operating element 12 and the currently existing parameter settings 13 of the building technical equipment. It should be understood that embodiments of the present invention can also include only a subset of these input values. Based on these input values, the control device 20 determines new parameter settings 30 of the building technical equipment.

[0080] Please refer to the attached Figure 2 The working principle of the control device 20 is explained in more detail. In this example, the control device 20 includes predefined parameter setting rules 21, at least one machine learning model 22 and a database management 23. In addition, the control device 20 is connected to and accesses a database 40. The database 40 can be constructed as part of a remote server system. The data stored in the database 40 is managed with the help of the database management 23.

[0081] The machine learning model 22 (maschinelle Lernmodell), hereinafter also referred to as a machine learning model (Machinenlern-Modell), may include or be an artificial neural network (ANN; artificial neural network), in particular using a multilayer perceptron (MLP). Alternatively or additionally, a decision tree and / or a random forest method is preferably used. Typically, the machine learning model 22 may be based on a model using machine learning or an algorithm using machine learning. Machine learning may involve algorithms and statistical models that may use a computer system to perform specific tasks without the use of explicit instructions, without relying on models and reasoning. For example, in machine learning, data transformations derived from the analysis of historical data and / or training data may be used instead of rule-based data transformations. For example, the content of an image may be analyzed using a machine learning model or using a machine learning algorithm. In order to enable the machine learning model to analyze the content of an image, the machine learning model may be trained using a training image as input and training content information as output. By training a machine learning model using a large number of training images and / or training sequences (e.g., words or sentences) and associated training content information (e.g., tags or annotations), the machine learning model "learns" to recognize the content of the image, so that the machine learning model can be used to recognize image content not included in the training data. The same principle can also be applied to other types of sensor data: by training the machine learning model using training sensor data and desired outputs, the machine learning model "learns" the conversion between sensor data and output, which can be used to provide output based on non-training sensor data provided to the machine learning model. The provided data (e.g., sensor data, metadata, and / or image data) can be preprocessed to obtain feature vectors, which are used as input to the machine learning model.

[0082] The machine learning model can be trained using training input data. The above example uses a training method called "supervised learning". In supervised learning, a machine learning model is trained using multiple training sample values, wherein each sample value may include multiple input data values ​​and multiple expected output values, that is, each training sample value is associated with an expected output value. By giving the training sample values ​​and the expected output values, the machine learning model "learns" which output value should be provided based on the input sample values ​​similar to the sample values ​​provided during training. In addition to supervised learning, semi-supervised learning can also be used. In semi-supervised learning, some training sample values ​​lack the expected output value. Supervised learning can be based on supervised learning algorithms (such as classification algorithms, regression algorithms, or similarity learning algorithms). When the output is limited to a limited set of values ​​(classification variables), that is, the input is classified as one of a limited set of values, a classification algorithm can be used. If the output shows any numerical value (within a certain range), a regression algorithm can be used. A similarity learning algorithm may be similar to a classification algorithm and a regression algorithm, but is based on learning from examples using a similarity function that measures how two objects are similar or related. In addition to supervised learning or semi-supervised learning, unsupervised learning can also be used to train machine learning models. In unsupervised learning, it is possible to (only) provide input data and the unsupervised learning algorithm can be used to discover structure in the input data (e.g., by grouping or clustering the input data, discovering commonalities in the data). Clustering is the assignment of input data containing multiple input values ​​into subsets (clusters) so that according to one or more (predefined) similarity criteria, input values ​​within the same cluster are similar, but dissimilar to input values ​​contained in other clusters.

[0083] Reinforcement learning is a third class of machine learning algorithms. In other words, reinforcement learning can be used to train machine learning models. In reinforcement learning, one or more software agents (so-called "Software Agents") are trained to perform actions in an environment. Rewards are calculated based on the actions performed. Reinforcement learning selects actions based on the training of one or more software agents so that the accumulated rewards are increased, which causes the software agents to become better at the tasks assigned to them (as evidenced by the increased rewards).

[0084] In addition, some techniques can be applied to certain machine learning algorithms. For example, feature learning can be used. In other words, the machine learning model can be trained at least in part using feature learning, and / or the machine learning algorithm can include a feature learning component. Feature learning algorithms (referred to as representation learning algorithms) can take the information in their input, but transform it in such a way that it becomes useful, usually as a preprocessing stage before performing classification or prediction. For example, feature learning can be based on principal component analysis or cluster analysis.

[0085] In some instances, anomaly detection (i.e., outlier detection) may be used with the goal of providing identification of input values ​​that are suspicious because they differ significantly from the majority of input data and training data. In other words, a machine learning model may be trained at least in part using anomaly detection, and / or a machine learning algorithm may include an anomaly detection component.

[0086] For example, in some examples, a machine learning algorithm may use a decision tree as a prediction model. In other words, a machine learning model may be based on a decision tree. In a decision tree, observations of an object (e.g., a set of input values) may be represented by branches of the decision tree, and output values ​​corresponding to the object may be represented by leaves of the decision tree. Decision trees may support both discrete and continuous values ​​as output values. If discrete values ​​are used, the decision tree may be called a classification tree, and if continuous values ​​are used, it may be called a regression tree.

[0087] Association rules are another technique that can be used in machine learning algorithms. In other words, a machine learning model can be based on one or more association rules. Association rules are created by identifying relationships between variables in a large amount of data. Machine learning algorithms can identify and / or use one or more relationship rules that represent knowledge derived from the data. For example, these rules can be used to store, manipulate, or apply knowledge.

[0088] Machine learning algorithms are typically based on machine learning models. In other words, the concept of a "machine learning algorithm" may refer to a set of instructions that can be used to create, train, or use a machine learning model. The concept of a "machine learning model" may refer to a data structure and / or a rule set that represents learned knowledge (e.g., based on training performed by a machine learning algorithm). In an embodiment, using a machine learning algorithm may mean using an underlying machine learning model (or multiple underlying machine learning models). Using a machine learning model may mean that the machine learning model and / or the data structure / rule set that serves as the machine learning model is trained by a machine learning algorithm.

[0089] For example, a model for machine learning can be an artificial neural network (ANN). ANN is a system inspired by biological neural networks, such as those found in the retina or brain. ANN includes multiple intermediate connecting nodes and multiple connections between nodes called edges. There are usually three types of nodes: input nodes that receive input values, hidden nodes that are (only) connected to other nodes, and output nodes that provide output values. Each node can represent an artificial neuron. Each edge can send information between nodes. The output of a node can be defined as a (non-linear) function of the input (e.g. the sum of its inputs). The input of a node can be used in a function based on the "weight" of the edge or node that provides the input. The weights of nodes and / or edges can be adjusted during the learning process. In other words, the training of an artificial neural network can include adjusting the weights of the nodes and / or edges of the artificial neural network, i.e. achieving the desired output for a specific input.

[0090] Alternatively, the model of machine learning can be a support vector machine, a random forest model or a gradient boosting model. Support vector machine (i.e., support vector network) is a supervised learning model with an associated learning algorithm, which can be used to analyze data (e.g., in classification analysis or regression analysis). Support vector machine can be trained by providing an input with multiple training input values ​​belonging to one of two categories. Support vector machine can be trained to assign new input values ​​to one of these two categories. Alternatively, the model of machine learning can be a Bayesian network, which is a probabilistic directed acyclic graph model. Bayesian networks can use directed acyclic graphs to represent a set of random variables and their conditional dependencies. Alternatively, the model of machine learning can be based on a genetic algorithm, which is a search algorithm and heuristic technique that mimics a natural selection process.

[0091] Regardless of how the machine learning model is implemented, the purpose of the embodiment shown is to determine new parameter settings 30 for a building technology system that take into account the preferences or requirements of users of a building equipped with the building technology system. To this end, the state of an environmental condition detected by means of at least one sensor device 11 is first received.

[0092] The concept of "ambient condition status" is based on one or more measurement time points at which the ambient conditions of the building technical equipment are measured, such as temperature, light conditions and / or wind force and direction. The ambient condition status can also include generally available information about the ambient conditions of the building technical equipment, such as the time or date. In addition, the ambient condition status can include currently existing parameter settings 13 or settings of the corresponding building technical equipment. In this context, the ambient condition status is to be understood as one or more ambient conditions that are perceptible or verifiable by a user.

[0093] The at least one sensor device 11 is configured to detect ambient conditions in the form of measurement data and may in particular have a plurality of sensors, for example a first temperature sensor for the building exterior temperature, a second temperature sensor for the building interior temperature, a light sensor or a wind sensor.

[0094] exist Figure 2 The embodiment shown in the figure includes the control of building technical equipment (particularly sunshade equipment and lighting equipment), wherein the above-mentioned computer-implemented method is set to determine parameter settings 30, and the method also includes the above-mentioned method for training a machine learning model 22.

[0095] In a first step, an environmental condition state detected by means of at least one sensor device 11 is received. The environmental condition state is preferably scaled, in particular in a value range between 0 and 1. Particularly preferably, the detected environmental condition state is detected as a vector, or the measured values ​​detected by means of the at least one sensor device 11 are converted into a vector form, wherein when scaling is performed, each value of the detected environmental condition state vector is scaled.

[0096] In a second step, it can be provided that the number of environmental condition states stored in the database 40 is determined, which meet the similarity criteria associated with the detected environmental condition states. The stored environmental condition states are preferably also scaled, and can exist in particular as vectors. When determining the number of environmental condition states stored in the database 40 that meet the similarity criteria associated with the detected environmental condition states, the detected environmental condition states can be checked for differences and / or similarities with the stored environmental condition states. This can be done in the manner described above, in particular using a cell-lists algorithm.

[0097] If the number is less than a predefined number threshold, for example 3, the parameter setting 30 can be determined according to at least one predefined parameter setting rule 21. This advantageously ensures that the parameter setting or adjustment of the settings of the building technical system is carried out even if a data basis in the form of stored ambient condition states does not exist or is insufficient and that a minimum user comfort is also ensured in this case. For example, the at least one predefined parameter setting rule 21 can be aimed at the usual expected preferences of ordinary users. In particular, in this example, it is the case that during sunny periods of the day, it is generally expected that the sunshade device can shade the room from sunlight to a certain extent. According to the predefined parameter rule 21 reflecting this, the sunshade device is at least partially deployed when the light sensor detects strong sunlight.

[0098] If the number is greater than or equal to a predefined number threshold, the parameter setting can be determined according to the machine learning model 22. The machine learning model 22 is able to learn the preferences of one or more users so that, given a sufficient data basis, a new parameter setting 30 for the building technology system with the aid of the machine learning model 22 comes closer to the actual preferences of the user or users. As explained at the outset, a sufficient data basis is available if the number of stored environmental condition states that meet a similarity criterion with respect to the detected environmental condition state is greater than the number threshold.

[0099] At this point it is mentioned that the stored environmental condition states can exist in pairs with the respectively associated stored parameter settings in the database 40. In other words: the machine learning model 22 can always be used when there are enough experience values ​​for the user's preferences. These experience values ​​can be gradually collected over a longer period of use after startup. For example, every time a user manually changes a parameter setting or a setting of a building technical setting by means of an operating element, these changes are stored in such a way that the current environmental condition state and the manual parameter setting are stored in pairs in the database 40. The machine learning model 22 thus replaces the predefined parameter setting rules 21 after a period of time. Since the predefined parameter setting rules 21 can only be rules for average preferences of the population, for example, the parameter setting rules 21 cannot fully reflect the preferences of a specific user or a specific user of a building technical device. However, after there are enough stored environmental condition states in the accessible database 40, the machine learning model 22 can reflect these preferences with a relatively high degree of accuracy.

[0100] The currently existing parameter setting 13 is specifically adjusted to the (new) parameter setting 30 determined by the method, and it is particularly preferred that the currently existing parameter setting 13 can be carried out by sending a command to the specific (new) parameter setting 30. It is particularly preferred to provide that the adjustment or the command is only made when the specific (new) parameter setting 30 deviates from the currently existing parameter setting 13 by a predefined parameter setting threshold. It is thus ensured that slight deviations between the specific (new) parameter setting 30 and the currently existing parameter setting 13 are ignored. Considering the energy demand of the building technical equipment, this is particularly advantageous because no unnecessary changes occur. The user's comfort is also improved by the parameter setting threshold because the user is not disturbed by possible noise (for example, when the sunshade device is folded or unfolded) when it is not necessary. For example, the parameter setting threshold can be defined as a deviation of 5% from the desired or optimal brightness in Lux. If the actual brightness deviates by a maximum of 5% from the desired or optimal brightness in Lux, this is ignored as a slight deviation and does not lead to a change in the parameter setting.

[0101] The machine learning model used within the framework of the method is in particular trained by a training method, in particular continuously. The training method may in particular include supervised machine learning and / or reinforcement machine learning. Figure 2 The training method can be designed as follows:

[0102] In response to manual parameter setting of the building technical device or the absence of parameter setting within a predefined time period, the method includes storing the environmental condition status and the associated parameter settings of the building technical device in the database 40. In addition, the method includes training the machine learning model 22 using at least a portion of the environmental condition status and the associated parameter settings in the database 40.

[0103] In this embodiment, manual parameter setting of the building technical equipment is equivalent to operating the operating element of the building technical equipment by the user. In this embodiment, no manual parameter setting of the building technical equipment is equivalent to no operation of the operating element of the building technical equipment.

[0104] In all embodiments of the invention, it can optionally be provided that no parameter settings have been made within a predefined time period and that the room and / or building is otherwise occupied by at least one user, which only results in the state of the environmental conditions and the associated parameter settings of the building technology system being stored in the database 40. For example, the occupancy situation can be determined by means of an occupancy sensor designed as a sensor device 11.

[0105] If the parameter setting meets the user's preference, the user will not change the parameter setting through the operating element 12 of the sunshade device. For example, in order to train the machine learning model 22, after determining the parameter setting according to the predefined parameter setting rule 21, the time when the user did not change it through the operating element 12 (no manual parameter setting or no operation of the operating element) can be detected. If the user does not perform manual parameter setting within a certain time, it can be inferred that the parameter setting meets his actual preference under the current environmental condition state. The detected environmental condition state and the related parameter settings are stored as stored environmental condition states in the database 40 and used as training data for the machine learning model 22.

[0106] In other words, it can be provided that when training the machine learning model, the operation of the operating elements of the building technology system by the user and / or the occupancy of the room and the non-operation of the operating elements within a predefined period of time are detected, and in particular the current environmental condition state (with the associated parameter settings) is stored in the database 40. It can be particularly preferably provided that the environmental condition state (with the associated parameter settings) stored in the database 40 is given a higher or lower weight than other environmental condition states (with the associated parameter settings) stored in the database 40 when another criterion is met. For example, another criterion that leads to a higher weighting of the stored environmental condition state (with the associated parameter settings) can be the non-operation of the operating element within a further longer period of time.

[0107] On the contrary, if the parameter setting does not meet the user's preferences, the user manually changes the parameter setting through the operating element 12 and, for example, further retracts or further unfolds the sunshade device. For example, in order to train the machine learning model 22, after manual parameter setting by the user, it can be inferred that the new parameter setting 30 meets the user's preferences under the current environmental condition state. The detected environmental condition state and the related parameter settings are stored in the database 40 as stored environmental condition states and used as training data for the machine learning model 22. Alternatively or additionally, after manual parameter setting by the user, the time when the user subsequently did not make changes through the operating element 12 can be detected. If it is detected that no further changes are made through the operating element 12 within a predefined time period after the manual parameter setting, it can be inferred that there is currently a parameter setting that meets the user's preferences for the current environmental condition state. The detected environmental condition state and the related parameter settings are stored in the database 40 as stored environmental condition states and used as training data for the machine learning model 22.

[0108] For example, in the context of reinforcement machine learning, a weight distribution can be performed, similar to the "reward and penalty" principle. Here, a reward may result in a higher weight of a stored environmental condition state, while a penalty may result in a lower weight of a stored environmental condition state. In particular, only a higher weight (reward) or only a lower weight (penalty) can be set. In particular, as the duration of time during which the user does not operate the operating element increases, a reward and therefore a higher weight of the environmental condition state may be triggered. For example, a reward or a higher weight may continue to increase with the duration during which the operating element is not operated. In particular, when the user operates the operating element, a penalty or a lower weight of the environmental condition state may be triggered. The more frequently the user operates, the greater the penalty or lower weight may be.

[0109] Reinforcement machine learning preferably includes criteria that combine certain user behaviors (especially operations of operating elements) with rewards or higher weights for environmental conditions or with penalties or lower weights for environmental conditions. In particular, these criteria can be further developed within the framework of their own standard learning model in order to optimize the conditions and amplitude of the weight changes. Dead times that occur, for example, during heating and cooling can also be taken into account in an advantageous manner. Dead times may be caused in particular by differences between the perceived temperature and the actual temperature and may, for example, be the time that elapses before the perceived temperature and the actual temperature become consistent.

[0110] In addition, a storage cleanup can be provided, in which it is checked whether the number of stored environmental condition states exceeds a pre-given maximum value. In particular, the maximum value can be the maximum permissible number of stored environmental condition states. In the context of storage cleanup, it can be provided in particular that a cell-lists algorithm is used. In this case, it can be particularly advantageous to provide a maximum value for each cell in the cell-lists algorithm, for example 100 environmental condition states per cell. If the number of stored environmental condition states exceeds a maximum value or a maximum value within a cell of the cell-lists algorithm, the oldest stored environmental condition state can be deleted until the number of stored environmental condition states no longer exceeds the maximum value. The provision of storage management complements the above-mentioned advantages. Storage management can be understood as a part of supervision in the sense of supervised machine learning.

[0111] It can also be provided that the training of the machine learning model 22 is carried out by means of supervised and / or reinforced machine learning. It is particularly preferred that when training the machine learning model 22, a training / test validation and / or cross-validation split of stored environmental condition states (and associated parameter settings) is also used (see the above explanation).

[0112] The proposed operating principle becomes more apparent overall with regard to the service life of a building technology device, starting from the initial installation. Initially, there are no stored environmental condition states (and the associated parameter settings), only predefined parameter setting rules 21, which can be specified, for example, by the manufacturer of the building technology device and can be for an assumed average user. During operation, the environmental condition states are detected continuously or at discrete time intervals by the sensor device 11 and reference is made to the predefined parameter setting rules 21. This can be seen as the first phase of the service life of the building technology device. After a period of time, the user (for example, the resident of the building for which the building technology device is intended) has made manual settings with the aid of an operating element, or has not made manual settings if the parameter settings made according to the predefined parameter setting rules 21 are in line with their preferences. With manual settings and without manual settings (for example, for a predefined time period), the stored environmental condition states (and the associated parameter settings) are stored in a database 40. In the case of frequently repeated detected environmental condition states, these environmental condition states can be called up with the aid of a machine learning model 22. In the case of rarely occurring detected environmental condition states, the predefined parameter setting rules 21 are also relied on. In this phase, which can be regarded as the second phase of the service life of the building technical equipment, the machine learning model 22 and the predefined parameter setting rules 21 are needed and used. After a further period of time has passed and the machine learning model 22 can call up a sufficient number of stored environmental condition states (and the associated parameter settings), the machine learning model 22 is mainly used, and ideally only the machine learning model 22 is used. This state can be regarded as the third phase of the service life of the building technical equipment. If the user changes (for example, the tenant of the building changes), the machine learning model 22 can gradually adapt to possible new preferences in the manner described above. Preferably, a reset function can be provided for the database 40 of stored environmental condition states (and the associated parameter settings).

[0113] The above-described enhanced machine learning method, such as the case of weightings associated with time intervals of stored ambient condition states, is particularly advantageous for use in shading devices and lighting devices or in shading and lighting control devices.

[0114] The special application of the method according to the invention can also be used in the control of air conditioning and heating equipment as part of building technology equipment. Usually, the control of air conditioning and heating equipment is based on rules. Usually, the temperature setting value can only be changed by the corresponding operating element 12. When controlling cooling, the challenge is that the cooling may cause air flow that is noticeable to the user according to the room conditions. Due to this so-called wind chill, the body temperature will be lower than the measured temperature. This effect may lead to an increase in the temperature setting value among the affected users. Due to the lack of a power-off switch for controlling cooling, some users also try to stop cooling by opening the window (and thus operating the window contactor). This leads to increased energy consumption, because after closing the window, the temperature usually rises and the room is cooled more strongly. The optimal control of cooling is usually different for each user, and may even be different for each room, and in addition to reaching the desired temperature setting value, it also includes minimizing the burden of the air flow caused by cooling on the user. Using the method according to the invention, especially for cooling equipment, the user's preferences can be taken into account without the user providing a specific temperature setting value. This is particularly advantageous because the user can usually only express his subjective temperature feeling inaccurately in the form of a temperature setting value. The method may also take into account dead time, during which the perceived temperature deviates from the actual temperature. As part of the sensor device 11, a touch sensor may also be provided on the window, which registers and takes into account the user's contact with the window.

[0115] Although some aspects are described in the framework of devices, it is clear that these aspects also represent a description of the corresponding method, where a block or a device corresponds to a method step or a function of a method step. Similarly, aspects described in the framework of method steps also represent a description of the characteristics of the corresponding blocks or elements or corresponding devices.

[0116] Embodiments of the present invention may be implemented in a computer system. The computer system may be a local computer device (e.g., a personal computer, a laptop, a tablet computer, or a mobile phone) having one or more processors and one or more storage devices; or it may be a distributed computer system (e.g., a cloud computing system having one or more processors or one or more storage devices, which are distributed at different locations, such as at a local client and / or one or more remote server farms and / or data centers). The computer system may include any circuit or combination of circuits. In one embodiment, the computer system may include one or more processors, which may be of any type. According to local usage, the processor may refer to any type of computing circuit, such as, but not limited to, a microprocessor, a microcontroller, a microprocessor with a complex command set (CISC), a microprocessor with a reduced instruction set (RISC), a microprocessor with a very long command word (VLIW), a graphics processor, a digital signal processor (DSP), a multi-core processor, a field programmable gate array (FPGA), or any other type of processor or processing circuit. Other types of circuits that may be included in a computer system may be custom circuits, application specific integrated circuits (ASICs), or similar circuits, such as one or more circuits (e.g., communication circuits) for wireless devices such as mobile phones, tablet computers, laptops, walkie-talkies, and similar electronic systems. A computer system may include one or more storage devices, which may include one or more storage elements suitable for the respective application, such as main memory in the form of random access memory (RAM), one or more hard disks and / or one or more drives for handling removable media such as CDs, flash memory cards, DVDs, etc. A computer system may also include a display device, one or more speakers, and a keyboard and / or a controller, which may include a mouse, a trackball, a touch screen, a voice recognition device, or any other device that allows a system user to input information into a computer system and receive information from it.

[0117] Some or all method steps can be performed by (or using) hardware devices, such as processors, microprocessors, programmable computers or electronic circuits. In some embodiments, one or more main method steps can be performed by such devices.

[0118] The implementation can be carried out using a non-volatile storage medium, such as a digital storage medium, such as a floppy disk, DVD, Blu-ray, CD, ROM, PROM and EPROM, EEPROM or flash memory, on which electronically readable control signals are stored, which cooperate (or can cooperate) with a programmable computer system so that the corresponding method is implemented. Therefore, the digital storage medium can be computer readable.

[0119] Some exemplary embodiments according to the invention comprise a data medium having electronically readable control signals which can interact with a programmable computer system in such a way that one of the methods described herein is carried out.

[0120] Generally speaking, the embodiments of the present invention can be implemented as a computer program product with a program code, wherein when the computer program product runs on a computer, the program code effectively performs one of the methods. The program code can be stored on a machine-readable carrier, for example.

[0121] Other exemplary embodiments comprise the computer program for performing one of the methods described herein, stored on a machine readable carrier.

[0122] In other words, one exemplary embodiment of the present invention is, therefore, a computer program having a program code for performing one of the methods described herein, when the computer program runs on a computer.

[0123] Therefore, another embodiment of the present invention is a storage medium (or data carrier or computer-readable medium), which, when executed by a processor, comprises a computer program stored thereon, the computer program being used to perform one of the methods described herein. The data carrier, the digital storage medium or the recorded medium is typically tangible and / or non-transient. Another embodiment of the present invention is a device comprising a processor and a storage medium as described herein.

[0124] Therefore, another embodiment of the present invention is a data stream or a signal sequence, which presents a computer program for executing one of the methods described herein. For example, a data stream or a signal sequence can be configured so that it is transmitted via a data communication connection (e.g., via the Internet).

[0125] A further embodiment comprises a processing means, for example a computer or a programmable logic device, configured or adapted to perform one of the methods described herein.

[0126] A further exemplary embodiment comprises a computer on which the computer program for performing one of the methods described herein is installed.

[0127] Another embodiment according to the present invention includes a device or system configured to transmit (e.g., electronically or optically) to a receiver a computer program for performing one of the methods described herein. The receiver may be, for example, a computer, a mobile device, a storage device, etc. The device or system may include, for example, a file server for transmitting the computer program to the receiver.

[0128] In some embodiments, a programmable logic device (e.g., a field programmable gate array, FPGA) can be used to perform some or all of the functions of the methods described herein. In some embodiments, a field programmable gate array can work in conjunction with a microprocessor to perform one of the methods described herein. Typically, these methods are preferably performed by any hardware device.

[0129] Reference numerals list

[0130] Sensor device 11

[0131] Operating element 12

[0132] Current parameter settings 13

[0133] Parameter setting 30

[0134] Control device 20

[0135] Predefined parameter setting rules 21

[0136] Machine Learning Model 22

[0137] Database Management 23

[0138] Database 40

Claims

1. A computer-implemented method for determining parameter settings (30) of a building technology device, comprising: Receiving a state of an environmental condition detected by means of at least one sensor device (11); determining the number of environmental condition states stored in a database (40) that meet a similarity criterion associated with the detected environmental condition state; as well as If the number is less than a predefined number threshold, determining a parameter setting (30) for the building technology system according to at least one predefined parameter setting rule (21); or If the quantity is greater than or equal to a predefined quantity threshold, parameter settings for the building technology system are determined based on a machine learning model (22).

2. The method of claim 1, further comprising: Sending a command to change the currently existing parameter setting (13) of a building technology device to a specific parameter setting (30), Preferably, the sending is performed only when the specific parameter setting (30) deviates from the currently existing parameter setting (13) by a predefined parameter setting threshold.

3. The method according to claim 1 or 2, wherein: Predefined constraints, in particular safety constraints, are observed when determining the parameter settings (30).

4. The method according to any one of claims 1 to 3, further comprising: Presence information and / or identification information is received, wherein the presence information contains at least information about the number of persons in a room for which a building technology system is set up, wherein the identification information contains at least information that enables a person located in the room or entering the room to be clearly identified, and wherein the presence information and / or the identification information is taken into account for determining the parameter setting (30).

5. The method according to claim 4, wherein: Based on the presence information and / or based on the identification information, at least one predefined parameter setting rule (21) and / or machine learning model (22) for determining the parameter setting (30) is selected from a plurality of parameter setting rules or a plurality of machine learning models (22).

6. The method according to any one of claims 1 to 5, wherein: The environmental condition state detected by means of the sensor device (11) includes at least one of the following information: The measured value of the light sensor, The wind sensor's measurements, The measured value of the temperature sensor, The difference between temperature and perceived temperature, Measurements from room occupancy sensors, time, The measured values ​​of a touch sensor, which is preferably located on the window handle, Date, preferably without year, The condition of building elements, preferably windows, The condition of the air conditioning equipment, The state of the heating equipment, The status of the shading equipment, the status of lighting equipment; and / or parameter settings (30) of a building technology system, said parameter settings comprising at least one or more of the following parameter settings: of a building element, preferably a window, Air conditioning equipment, Heating equipment, Sunshade equipment, Lighting equipment.

7. The method according to any one of claims 1 to 6, wherein: The similarity criteria of the stored environmental condition states are checked according to the cell-lists algorithm, wherein the similarity criteria for the stored environmental condition states are preferably met only when the difference between the detected environmental condition state and the stored environmental condition state does not exceed a certain value.

8. A computer-implemented method for training a machine learning model (22), for use in a method according to any one of claims 1 to 7, comprising: In response to manual parameter setting of the building technology equipment or no parameter setting for a predefined period of time, storing the state of the environmental conditions and the parameter settings associated with the building technology equipment in a database (40); as well as The machine learning model (22) is trained using at least a portion of the environmental condition states and associated parameter settings in the database (40).

9. A device, in particular a control device (20) for a building technology installation, comprising means for carrying out the method according to one of claims 1 to 8.

10. Computer program comprising instructions which, when the program is executed by a computer, cause the computer to execute the method according to one of claims 1 to 8.

Images