A method for controlling an air handling unit, an air handling unit, an air handling system, a computer program and a computer program product

Abstract

The disclosure relates to a method (100) for controlling an air handling unit, AHU, the AHU comprising: a rotor comprising a desiccant material; a heater configured to heat the regeneration air upstream of the rotor; and a control unit configured to control the AHU, the method (100) comprising continuously: generating (S400), using a machine learning, ML, model, an on / off-command configured to switch the heater on or off based on: one or more measured humidity levels relating to a humidity level of process air downstream of the rotor and / or a humidity level of a room to which dehumidified process air is fed downstream of the rotor; and switching (S600) the heater on or off using the generated on / off-command. The disclosure further relates to an air handling unit, an air handling system, a computer program and a computer program product.

Description

[0001] A method for controlling an air handling unit, an air handling unit, an air handling system, a computer program and a computer program product

[0002] Technical field

[0003] The present disclosure relates to a method for controlling an air handling unit, an air handling unit, an air handling system, a computer program and a computer program product. More specifically, the disclosure relates to a method for controlling an air handling unit, an air handling unit, an air handling system, a computer program and a computer program product as defined in the introductory parts of the independent claims.

[0004] Background art

[0005] Air handling units for dehumidification, such as air handling units comprising a desiccant rotor, have many passive and powered components that operate in tandem to process the air. The complex interplay of the passive and powered components of the air handling unit may lead to inefficient power usage. There is thus a need in the art for air handling units with improved energy efficiency.

[0006] Summary

[0007] It is an object of the present disclosure to mitigate, alleviate or eliminate one or more of the above-identified deficiencies and disadvantages in the prior art and solve at least the above-mentioned problem.

[0008] The present disclosure is based on an experimental observation where it has been noticed that one can save considerable amount of energy when the control strategy is in a such way that the reactivation, i.e. the desorption of bound water from a desiccant rotor by letting hot so-called reactivation air pass through the rotor, turns off when a controlled humidity level is below a desired setpoint, given that some oscillatory behaviour around the setpoint can be afforded.

[0009] According to a first aspect there is provided a method for controlling an air handling unit, AHU. The AHU comprises a rotor comprising a desiccant material. The AHU further comprises a process air circuit configured to conduct process air through a first sector of the rotor. The AHU also comprises a regeneration air circuit configured to conduct regeneration air through a second sector of the rotor. The AHU additionally comprises a heater configured to heat the regeneration air upstream of the rotor. The AHU further comprises a control unit configured to control the AHU. The method comprises continuously: generating, using a machine learning, ML, model, an on / off-command configured to switch the heater on or off based on: one or more measured humidity levels relating to a humidity level of process air downstream of the rotor and / or a humidity level of a room to which dehumidified process air is fed downstream of the rotor; and switching the heater on or off using the generated on / off- command.

[0010] In contrast to traditional methods, where the humidity level is set to track a humidity level setpoint as close and stable as possible, the present disclosure uses a non-equilibrium method for keeping the humidity level of process air within a predetermined interval.

[0011] The amplitude of the humidity level oscillation is dependent on when the reactivation turns on. By using machine learning it is possible to predict when to switch the heater on or off without the humidity level falling outside the predetermined interval of allowable humidity levels. The machine learning model thereby acts synergistically with the choice to turn the heater on or off.

[0012] According to some examples, generating the on / off command further comprises: generating, using the ML-model, a prediction of one or more future humidity levels based on the one or more measured humidity levels and temperature measurements of the regeneration air downstream of the heater; and generating the on / off-command based on a comparison between the predicted one or more future humidity levels and a predetermined range of allowable humidity levels.

[0013] According to some examples, generating the on / off-command based on said comparison further comprises generating an on-command based on the comparison indicating a future humidity level exceeding the highest allowable humidity level in the predetermined range of allowable humidity levels; and generating an off-command on the comparison indicating a future and / or present humidity level falling below the lowest allowable humidity level in the predetermined range of allowable humidity levels.

[0014] The predicted future humidity levels enable direct and / or indirect prediction if a future humidity level will potentially fall outside of the predetermined range of allowable humidity levels. By "direct" we herein mean that a predicted future humidity level can be directly compared to the predetermined range of allowable humidity levels. By "indirect prediction" we herein mean that a trend or pattern of the predicted future humidity levels can indicate, e.g. via extrapolation, that some humidity level even further in the future will potentially fall inside or outside of the predetermined range of allowable humidity levels. The indication does not be explicitly calculated in a separate step, but can be determined by the machine learning model as part of the generation of the on / off-command.

[0015] According to some examples, the control unit comprises a cascade controller, the cascade controller comprises a first control unit configured to receive a humidity setpoint relating to the humidity level of process air downstream of the rotor and / or the humidity level of a room to which dehumidified process air is fed downstream of the rotor, and generate the temperature setpoint based on the difference between at least one of the one or more measured humidity levels and the humidity setpoint. The cascade controller further comprises a second control unit configured to receive the temperature setpoint and control the heater based on a difference between the temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint. The control unit further comprises a third control unit configured to receive the generated on / off-command and to switch the first control unit, the heater, and optionally the second control unit, on or off based on the received on / off-command. Generating the on / off-command further comprises receiving, at the first control unit, the humidity setpoint. Generating the on / off- command also comprises generating, at the first control unit, the temperature setpoint based on based on the difference between the one or more measured humidity levels and the humidity setpoint. Generating the on / off-command further comprises determining, at the second control unit, the difference between temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint. Generating the on / off-command additionally comprises transmitting, to the ML-model, data relating to the one or more measured humidity levels, the data optionally further the method comprises the difference between temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint for the temperature of the regeneration air downstream of the heater. Generating the on / off- command further comprises generating, using the ML-model, the on / off-command based on the transmitted data. Generating the on / off-command also comprises transmitting the on / off- command to the third control unit, and switching the first control unit and the heater, and optionally the second control unit, on or off based on the on / off-command.

[0016] The use of a cascade controller supports different ways of generating a temperature setpoint for the heater when it is switched on, which can synergistically be combined with the generation of the on / off-command to further improve energy efficiency. According to some examples, generating the on / off-command is further based on a difference between temperature measurements relating to a temperature of the regeneration air downstream of the heater and a temperature setpoint for the temperature of the regeneration air downstream of the heater.

[0017] By also taking said temperature difference into account, the machine learning model can improve the accuracy of its predictions, thereby allowing the generation of the on / off- command to have the humidity levels come closer to the limits of the allowed interval, which in turn enables further energy savings.

[0018] According to some examples, the setpoint for the temperature is a predetermined fixed temperature.

[0019] A predetermined fixed temperature allows targeting a temperature of the regeneration air downstream of the heater that correlates to a temperature of the regeneration air downstream of the heater where the energy consumption required to remove a predetermined amount of moisture from the process air is lowest.

[0020] According to some examples, the setpoint for the temperature is generated by a PID- controller and / or a model predictive controller, MPC.

[0021] PID-controllers offer very low cost and reliable controllers, while MPC controllers can optimize both setpoints for humidity levels and temperature of the regeneration air downstream of the heater.

[0022] According to some examples, the method comprises obtaining a time series of the one or more measured humidity levels comprising a current humidity level. Generating the on / off- command comprises generating, using the ML-model, a prediction of a time series of future humidity levels based on the obtained time series of measurements of humidity levels, the time series of future humidity levels comprising at least one future humidity level. Generating the on / off-command further comprises generating the on / off-command based on the generated prediction of a time series of future humidity levels.

[0023] By obtaining a time series of humidity levels enables an improvement in the prediction of future humidity levels, which in turn enables generating on / off-commands that enable greater energy savings.

[0024] According to some examples, the method comprises obtaining a time series of temperature measurements relating to the temperature of the regeneration air downstream of the heater, the time series of temperatures the method comprises a measurement of a current temperature of the regeneration air downstream of the heater, and wherein generating the prediction of a time series of future humidity levels is further based on the time series of temperature measurements.

[0025] The time series of temperatures relates to adsorbed humidity being removed from the desiccant rotor and is therefore related to the corresponding time series of future humidity levels. By also generating the prediction of the time series of future humidity levels based on the time series of temperature measurements, more accurate predictions of future humidity levels can be made.

[0026] According to some examples, the method comprises transmitting the one or more humidity levels, and optionally temperature measurements, to a server; and transmitting the generated on / off-command to the AHU; wherein generating the on / off-command comprises generating the on / off-command at the server.

[0027] The computational burden can thereby be transferred to the server. The server can act as a cloud service and integrate real world operational data from multiple AHUs in multiple different environments, and thereby train machine learning models that are capable of handling a more diverse set of scenarios than a machine learning model trained at a local AHU in a specific environment.

[0028] According to some examples, generating the on / off-command is further based on sensor data relating to at least one of: a humidity level of process air upstream of the rotor, an energy consumption of at least one powered component of the AHU and / or at least one operational parameter of a powered component of the AHU.

[0029] In addition to enabling an improvement in accuracy of generating on / off-commands that allows the AHU to better use the full range of allowable humidity levels, the sensor data further enables tracking the performance of the AHU over time. The machine learning model can therefore be updated to take problems, such as a malfunctioning powered component or a clogged filter, into consideration and adjust the switching behaviour to account for the new reality.

[0030] According to a second aspect there is provided an air handling unit, AHU. The AHU comprises a rotor comprising a desiccant material. The AHU further comprises a process air circuit configured to conduct process air through a first sector of the rotor. The AHU also comprises a regeneration air circuit configured to conduct regeneration air through a second sector of the rotor. The AHU additionally comprises a heater configured to heat the regeneration air upstream of the rotor. The AHU further comprises a control system configured to carry of the method according to any of the first aspect.

[0031] The AHU implements the method according to the first aspect, and consequently has all the corresponding technical effects and advantages.

[0032] According to some examples, the control unit comprises a cascade controller. The cascade controller comprises a first control unit configured to receive a humidity setpoint relating to the humidity level of process air downstream of the rotor and / or the humidity level of a room to which dehumidified process air is fed downstream of the rotor, and generate the temperature setpoint based on the difference between the one or more measured humidity levels and the humidity setpoint, and a second control unit configured to receive the temperature setpoint and control the heater based on the difference between the temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint.

[0033] The AHU thereby implements method steps according to the first aspect comprising corresponding cascade controllers, exhibiting the same technical effects and advantages.

[0034] According to a third aspect there is provided an air handling system comprising an air handling unit, AHU, according to the second aspect, and a server. The AHU and the server are communicatively connected to each other. The system is configured to carry out the method according to the first aspect.

[0035] The system implements the method according to the first aspect and has all the associated technical effects and advantages.

[0036] According to a fourth aspect there is provided a computer program comprising computer program code which, when executed by a processor of the control unit of an AHU according to the second aspect, causes the AHU to perform the method according to the first aspect.

[0037] The computer program implements the method according to the first aspect and has all the associated technical effects and advantages.

[0038] According to a fifth aspect there is provided a computer program product comprising a non-transitory computer-readable storage medium having thereon a computer program comprising program instructions, the computer program being loadable into a processor of the control unit of an AHU according to the second aspect and configured to cause the AHU to perform the method according to the first aspect.

[0039] The computer program product implements the method according to the first aspect and has all the associated technical effects and advantages

[0040] Effects and features of the second through fifth aspects are to a large extent analogous to those described above in connection with the first aspect. Examples mentioned in relation to the first aspect are largely compatible with the the second through fifth aspects.

[0041] The present disclosure will become apparent from the detailed description given below. The detailed description and specific examples disclose preferred examples of the disclosure by way of illustration only. Those skilled in the art understand from guidance in the detailed description that changes and modifications may be made within the scope of the disclosure.

[0042] Hence, it is to be understood that the herein disclosed disclosure is not limited to the particular component parts of the device described or steps of the methods described since such device and method may vary. It is also to be understood that the terminology used herein is for purpose of describing particular examples only, and is not intended to be limiting. It should be noted that, as used in the specification and the appended claim, the articles "a", "an", "the", and "said" are intended to mean that there are one or more of the elements unless the context explicitly dictates otherwise. Thus, for example, reference to "a unit" or "the unit" may include several devices, and the like. Furthermore, the words "comprising", "including", "containing" and similar wordings does not exclude other elements or steps.

[0043] Brief of the

[0044] The above objects, as well as additional objects, features and advantages of the present disclosure, will be more fully appreciated by reference to the following illustrative and non-limiting detailed description of example examples of the present disclosure, when taken in conjunction with the accompanying drawings.

[0045] Figures la and lb illustrate the disclosed method according to the first aspect;

[0046] Figure 2 illustrates schematically an air handling unit according to the second aspect; and Figure 3 illustrates schematically an air handling system according to the third aspect.

[0047] Detailed description

[0048] The present disclosure will now be described with reference to the accompanying drawings, in which preferred example examples of the disclosure are shown. The disclosure may, however, be embodied in other forms and should not be construed as limited to the herein disclosed examples. The disclosed examples are provided to fully convey the scope of the disclosure to the skilled person.

[0049] Figures la and lb illustrate the disclosed method 100 for controlling an air handling unit, AHU, according to a first aspect of the present disclosure. The AHU comprises a rotor comprising a desiccant material. The AHU further comprises a process air circuit SI configured to conduct process air through a first sector of the rotor. The AHU also comprises a regeneration air circuit configured to conduct regeneration air through a second sector of the rotor. The AHU additionally comprises a heater configured to heat the regeneration air upstream of the rotor. The AHU further comprises a control unit configured to control the AHU.

[0050] The method 100 comprises continuously: generating S400, using a machine learning, ML, model, an on / off-command configured to switch the heater on or off based on: one or more measured humidity levels relating to a humidity level of process air downstream of the rotor and / or a humidity level of a room to which dehumidified process air is fed downstream of the rotor; and switching S600 the heater on or off using the generated on / off-command.

[0051] Traditionally the humidity level of the process air is regulated by obtaining an equilibrium at a steady state with respect to a humidity set point, typically by means of a PID controller. The disclosed method 100 instead continuously switches the heater on or off in a non-equilibrium process. For example, the humidity level of the process air may be allowed to rise with the heater off and the heater is switched on to prevent the humidity level to deviate above a predetermined threshold. Similarly, the humidity level may be allowed to drop below the humidity set point before the heater is switched off. The process of switching the heater on or off results in the humidity level of process air oscillating, and the amplitude is determined at least in part on when the heater is switched on or off. The heater temperature, and consequently also the temperature of the regeneration air downstream of the heater, can have an impact on the oscillating humidity level of the process air. Thus, if there exists an interval of allowable humidity levels, the disclosed method can provide significant energy savings compared to equilibrium methods of the prior art. The greater the span of the interval, the greater the potential energy savings.

[0052] Because of the complex behaviour of the humidity level when switching the heater on or off, the method 100 employs a machine learning model that acts synergistically with the on / off-command generation, in that the machine learning model can typically increase the energy savings with respect to PID-controllers.

[0053] Thus, the idea is that a machine learning model is trained in a way that it learns the timing when to turn on / off the reactivation in order not to exceed a certain oscillation amplitude around the humidity setpoint. The machine learning model can be trained as follows: the dehumidifier is being turned on / off in the environment where the strategy is to be employed. Data is then used to train the ML-model which will be the core of the proposed control strategy. An advantage is that a user can specify a tolerance for the oscillation and the control unit is supposed to deliver it while saving energy in the times that the reactivation is off. Thus, according to some examples, the method further comprises obtaining information configured to specify a predetermined range of allowable humidity levels. The information can be obtained via user input, for instance via a graphical user interface, GUI.

[0054] According to some examples, generating S400 the on / off command further comprises: generating S402, using the ML-model, a prediction of one or more future humidity levels based on the one or more measured humidity levels and temperature measurements of the regeneration air downstream of the heater; and generating S404 the on / off-command based on a comparison between the predicted one or more future humidity levels and a predetermined range of allowable humidity levels. In some examples, the one or more measured humidity levels comprise a current measured humidity level. According to some examples, the one or more measure humidity levels comprise at least on past humidity level. By past humidity level we herein mean a humidity level that was measured at a time preceding a current measured humidity level.

[0055] A direct comparison between the predicted future humidity levels and the minimum and maximum humidity levels of the predetermined range of allowable humidity levels allows detection of the humidity level of the process air being in danger of falling outside of the range of allowable humidity levels and the ML-model can generate an on / off-command to mitigate the risk. Thus, according to some examples, generating S404 the on / off-command based on said comparison further comprises: generating S406 an on-command based on the comparison indicating a future humidity level exceeding the highest allowable humidity level in the predetermined range of allowable humidity levels; and generating S408 an off- command on the comparison indicating a future and / or present humidity level falling below the lowest allowable humidity level in the predetermined range of allowable humidity levels.

[0056] A comparison between the predicted one or more future humidity levels and the predetermined range of allowable humidity levels may optionally comprise an analysis of trends and / or patterns of the predicted future humidity levels, for instance by extrapolation and / or a determination performed by the ML-model. Thus, even if a direct comparison does not indicate any of the predicted future humidity levels falling outside of the allowable range, the comparison may still be able to predict future behaviour by examining said trends and / or patterns and generate the on / off-command based on the comparison.

[0057] According to some examples, the ML-model comprises a reinforcement learning, RL, model. An RL-model is ideal for control and can make effective use, via its representation of states and actions, of both past measurements of humidity levels as well as predicted future humidity levels.

[0058] According to some examples, the control unit comprises a cascade controller. The cascade controller comprises a first control unit configured to receive a humidity setpoint relating to the humidity level of process air downstream of the rotor and / or the humidity level of a room to which dehumidified process air is fed downstream of the rotor, and generate the temperature setpoint based on the difference between at least one of the one or more measured humidity levels and the humidity setpoint. The cascade controller further comprises a second control unit configured to receive the temperature setpoint and control the heater based on a difference between the temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint. The control unit further comprises a third control unit configured to receive the generated on / off-command and to switch the first control unit, the heater, and optionally the second control unit, on or off based on the received on / off-command.

[0059] In some examples with the control unit comprising a cascade controller as described above, generating S400 the on / off-command further comprises receiving S410, at the first control unit 242, the humidity setpoint. Generating S400 the on / off-command also comprises generating S412, at the first control unit 242, the temperature setpoint based on based on the difference between the one or more measured humidity levels and the humidity setpoint. Generating S400 the on / off-command additionally comprises determining S414, at the second control unit 244, the difference between temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint. Generating S400 the on / off-command further comprises transmitting S416, to the ML-model, data relating to the one or more measured humidity levels, the data optionally further the method comprises the difference between temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint for the temperature of the regeneration air downstream of the heater. Generating S400 the on / off-command also comprises generating S418, using the ML-model, the on / off-command based on the transmitted data. Generating S400 the on / off-command additionally comprises transmitting S420 the on / off-command to the third control unit 246, and switching S422 the first control unit 242 and the heater 220, and optionally the second control unit 244 on or off based on the on / off-command.

[0060] In some further examples, the third control unit comprises the ML-model. The generation of the on / off-command thereby takes place at the AHU. In some examples, the third control unit and the ML-model are separate units. This enables the ML-model to be located at a server, as will be described further below.

[0061] Since the temperature of the regeneration air downstream of the heater influences how much adsorbed moisture is released from the desiccant rotor, and hence the ability of the desiccant rotor to dehumidify the process air, the ML-model may be further based on the temperature of the regeneration air downstream of the heater.

[0062] Thus, according to some examples, generating S400 the on / off-command is further based on a difference between temperature measurements relating to a temperature of the regeneration air downstream of the heater and a temperature setpoint for the temperature of the regeneration air downstream of the heater.

[0063] In some examples the setpoint for the temperature is a predetermined fixed temperature. In some further examples, the fixed temperature is determined based on the humidity level of process air upstream of the desiccant rotor. According to some examples, the predetermined fixed temperature is based on a temperature of the regeneration air downstream of the heater where the energy consumption required to remove a predetermined amount of moisture from the process air is lowest. According to some examples, the setpoint for the temperature is generated by a PID- controller and / or a model predictive controller, MPC.

[0064] In some further examples, the predetermined fixed temperature setpoint and / or the PID-controller and / or a model predictive controller, MPC generated setpoint are provided by a first control unit of a cascade controller as described above.

[0065] According to some examples, the ML-model comprises a reinforcement learning, RL, model. The use of RL models, which can predict future states and generating optimally timed on / off-commands parameters based on both historical and real-time data, enables a timeseries prediction capability that improves long-term system performance and energy efficiency. RL enables the system using the disclosed method to continuously improve based on real-time feedback, making the disclosed method 100 more adaptable and capable of handling dynamic environments.

[0066] In some examples, the on / off commands are further based on previous control parameters. In other words, previous control parameters as inputs for the RL algorithm, which allows the system to learn from past actions and continuously refine its control strategy, that is when to switch the heater on or off, for better performance.

[0067] Thus, in some examples the method 100 comprises: obtaining S100 a time series of the one or more measured humidity levels the method comprises a current humidity level. In these examples generating S400 the on / off-command also comprises generating S428, using the ML-model, a prediction of a time series of future humidity levels based on the obtained time series of measurements of humidity levels, the time series of future humidity levels comprising at least one future humidity level. Generating S400 the on / off-command comprises further comprises generating S430 the on / off-command based on the generated S428 prediction of a time series of future humidity levels.

[0068] According to some examples, the method 100 comprises obtaining S200 a time series of temperature measurements relating to the temperature of the regeneration air downstream of the heater. The time series of temperatures comprises a measurement of a current temperature of the regeneration air downstream of the heater. Generating S428 the prediction of a time series of future humidity levels is further based on the time series of temperature measurements.

[0069] According to some examples, the method 100 comprises transmitting S300 the one or more humidity levels, and optionally temperature measurements, to a server; and transmitting S500 the generated on / off-command to the AHU; wherein generating S400 the on / off-command comprises generating S450 the on / off-command at the server. By generating the on / off-command at a server, the computational burden of the ML-model can be taken of individual AHUs. The use of a cloud-based server that aggregates data from multiple AHU units to optimize the on / off-command generation centrally allows for global optimization of the multiple AHUs. A cloud-based server that gathers data from multiple AHUs, applies RL algorithms, and generates on / off-commands accordingly provides scalability and coordination across large systems.

[0070] According to some examples, generating S400 the on / off-command is further based on sensor data relating to at least one of: a humidity level of process air upstream of the rotor, an energy consumption of at least one powered component of the AHU and / or at least one operational parameter of a powered component of the AHU.

[0071] Figure 2 illustrates schematically an air handling unit, AHU, 200 according to the second aspect. The AHU 200 comprises a rotor 210 comprising a desiccant material. The AHU 200 further comprises a process air circuit SI configured to conduct process air through a first sector zl of the rotor 210. The AHU 200 also comprises a regeneration air circuit S2 configured to conduct regeneration air through a second sector z2 of the rotor 210. The AHU additionally comprises a heater 220 configured to heat the regeneration air upstream of the rotor 210. The AHU also comprises a control system 230 configured to carry of the method 100 according to the first aspect. The disclosed AHU 200 thereby implements the method 100 according to any of the examples of the first aspect, as described above, and consequently have all the associated technical effects and advantages.

[0072] According to some examples, the control unit 230 comprises a cascade controller 240. The cascade controller comprises a first control unit 242 configured to receive a humidity setpoint relating to the humidity level of process air downstream of the rotor and / or the humidity level of a room to which dehumidified process air is fed downstream of the rotor, and generate the temperature setpoint based on the difference between the one or more measured humidity levels and the humidity setpoint. The cascade controller also comprises a second control unit 244 configured to receive the temperature setpoint and control the heater based on the difference between the temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint. The cascade controller 240 enables the AHU to perform all the functions involving a cascade controller as discussed in relation to Figures la and lb, above, with the AHU exhibiting the corresponding technical effects and advantages.

[0073] Figure 3 illustrates schematically an air handling system 3000 according to a third aspect of the present disclosure. The air handling system 3000 comprises an air handling unit, AHU, 300 according to the second aspect. The air handling system 3000 further comprises a server 350. The AHU 300 and the server 350 are communicatively connected to each other. The system 3000 is configured to carry out the method 100 according to the first aspect.

[0074] The air handling system 3000 implements the disclosed method 100 and consequently have all the associated technical effects and advantages. In particular, the use of an external server 350 allows the system 3000 to place the ML-model at the server, the specifics of which have been discussed in relation to figures la and lb above.

[0075] The present disclosure further relates to a fourth aspect in the form of a computer program comprising computer program code which, when executed by a processor 232, 332 of the control unit 230, 330 of an AHU 200, 300 according to the second aspect, causes the AHU to perform the method 100 according to the first aspect.

[0076] The present disclosure further relates to a fifth aspect in the form of a computer program product comprising a non-transitory computer-readable storage medium having thereon a computer program comprising program instructions, the computer program being loadable into a processor 232, 332 of the control unit 230, 330 of an AHU 200, 300 according to claim the second aspect, and configured to cause the AHU 200, 300 to perform the method 100 according to the first aspect.

[0077] The person skilled in the art realizes that the present disclosure is not limited to the preferred examples described above. The person skilled in the art further realizes that modifications and variations are possible within the scope of the appended claims. Additionally, variations to the disclosed examples can be understood and effected by the skilled person in practicing the claimed disclosure, from a study of the drawings, the disclosure, and the appended claims.

Claims

CLAIMS1. A method (100) for controlling an air handling unit, AHU, the AHU comprising: a rotor (210, 310) comprising a desiccant material; a process air circuit (SI) configured to conduct process air through a first sector (zl) of the rotor (210, 310); a regeneration air circuit (S2) configured to conduct regeneration air through a second sector (z2) of the rotor (210, 310); a heater (220, 320) configured to heat the regeneration air upstream of the rotor (210, 310); and a control unit (230, 330) configured to control the AHU, the method (100) comprising continuously:- generating (S400), using a machine learning, ML, model, an on / off-command configured to switch the heater on or off based on: one or more measured humidity levels relating to a humidity level of process air downstream of the rotor and / or a humidity level of a room to which dehumidified process air is fed downstream of the rotor; and switching (S600) the heater on or off using the generated on / off-command.

2. The method according to claim 1, wherein generating (S400) the on / off command further comprises:- generating (S402), using the ML-model, a prediction of one or more future humidity levels based on the one or more measured humidity levels and temperature measurements of the regeneration air downstream of the heater; and- generating (S404) the on / off-command based on a comparison between the predicted one or more future humidity levels and a predetermined range of allowable humidity levels.

3. The method according to claim 2, wherein generating (S404) the on / off-command based on said comparison further comprises- generating (S406) an on-command based on the comparison indicating a future humidity level exceeding the highest allowable humidity level in the predetermined range of allowable humidity levels; andgenerating (S408) an off-command on the comparison indicating a future and / or present humidity level falling below the lowest allowable humidity level in the predetermined range of allowable humidity levels.

4. The method according to any of the preceding claims, wherein the control unit (230) comprises a cascade controller (240), the cascade controller comprising:• a first control unit (242) configured to receive a humidity setpoint relating to the humidity level of process air downstream of the rotor and / or the humidity level of a room to which dehumidified process air is fed downstream of the rotor, and generate a temperature setpoint based on the difference between at least one of the one or more measured humidity levels and the humidity setpoint, and• a second control unit (244) configured to receive the temperature setpoint and control the heater based on a difference between the temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint, wherein the control unit (230) further comprises:• a third control unit (246) configured to receive the generated on / off-command and to switch the first control unit (242), the heater (220), and optionally the second control unit (244), on or off based on the received on / off-command, wherein generating (S400) the on / off-command further comprises: receiving (S410), at the first control unit (242), the humidity setpoint,- generating (S412), at the first control unit (242), the temperature setpoint based on based on the difference between the one or more measured humidity levels and the humidity setpoint, determining (S414), at the second control unit (244), the difference between temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint,- transmitting (S416), to the ML-model, data relating to the one or more measured humidity levels, the data optionally further comprising the difference between17 temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint for the temperature of the regeneration air downstream of the heater,- generating (S418), using the ML-model, the on / off-command based on the transmitted data,- transmitting (S420) the on / off-command to the third control unit (246), and switching (S422) the first control unit (242) and the heater (220), and optionally the second control unit (244), on or off based on the on / off-command.

5. The method according to any of the preceding claims, wherein generating (S400) the on / off-command is further based on a difference between temperature measurements relating to a temperature of the regeneration air downstream of the heater and a temperature setpoint for the temperature of the regeneration air downstream of the heater.

6. The method according to claim 5, wherein the setpoint for the temperature is a predetermined fixed temperature.

7. The method according to claim 5, wherein the setpoint for the temperature is generated by a PID-controller and / or a model predictive controller, MPC.

8. The method (100) according to any of the preceding claims, further comprising obtaining (S100) a time series of the one or more measured humidity levels comprising a current humidity level; and wherein generating (S400) the on / off-command comprises:- generating (S428), using the ML-model, a prediction of a time series of future humidity levels based on the obtained time series of measurements of humidity levels, the time series of future humidity levels comprising at least one future humidity level, and- generating (S430) the on / off-command based on the generated (S428) prediction of a time series of future humidity levels.

189. The method (100) according to claim 8, further comprising obtaining (S200) a time series of temperature measurements relating to the temperature of the regeneration air downstream of the heater, the time series of temperatures comprising a measurement of a current temperature of the regeneration air downstream of the heater, and wherein generating (S428) the prediction of a time series of future humidity levels is further based on the time series of temperature measurements.

10. The method according to any of the preceding claims, further comprising- transmitting (S300) the one or more humidity levels, and optionally temperature measurements, to a server; and- transmitting (S500) the generated on / off-command to the AHU; wherein generating (S400) the on / off-command comprises- generating (S450) the on / off-command at the server.

11. The method according to any of the preceding claims, wherein generating (S400) the on / off-command is further based on sensor data relating to at least one of: a humidity level of process air upstream of the rotor (210, 310), an energy consumption of at least one powered component of the AHU (200, 300) and / or at least one operational parameter of a powered component of the AHU (200, 300).

12. An air handling unit, AHU, (200, 300) comprising: a rotor (210, 310) comprising a desiccant material; a process air circuit (SI) configured to conduct process air through a first sector (zl) of the rotor (210, 310); a regeneration air circuit (S2) configured to conduct regeneration air through a second sector (z2) of the rotor (210, 310);19 a heater (220, 320) configured to heat the regeneration air upstream of the rotor (210, 310); and a control system (230, 330) configured to carry of the method (100) according to any of claims 1-11.

13. The AHU (200, 300) according to claim 12, wherein the control unit (230, 330) comprises a cascade controller (240, 340), the cascade controller comprising:• a first control unit (242, 342) configured to receive a humidity setpoint relating to the humidity level of process air downstream of the rotor and / or the humidity level of a room to which dehumidified process air is fed downstream of the rotor, and generate a temperature setpoint based on the difference between the one or more measured humidity levels and the humidity setpoint, and• a second control unit (244, 344) configured to receive the temperature setpoint and control the heater based on the difference between the temperature measurements relating to the temperature of the regeneration air downstream of the heater and the temperature setpoint.

14. An air handling system (3000) comprising: air handling unit, AHU, (300) according to claim 12 or 13; and a server (350); wherein the AHU (300) and the server (350) are communicatively connected to each other; and wherein the system (3000) is configured to carry out the method according to any of claims 1-11.

15. A computer program comprising computer program code which, when executed by a processor (232, 332) of the control unit (230, 330) of an AHU (200, 300) according to claim 12 or 13, causes the AHU (200, 300) to perform the method according to any of claims 1-11.2016. A computer program product comprising a non-transitory computer-readable storage medium having thereon a computer program comprising program instructions, the computer program being loadable into a processor (232, 332) of the control unit (230, 330) of an AHU (200, 300) according to claim 12 or 13 and configured to cause the AHU (200, 300) to perform the method according to any of claims 1-11.

Images