Machine learning-based control systems for heating, ventilation, and cooling systems

Abstract

This disclosure relates to machine learning and artificial intelligence-based control systems for heating, ventilation, and cooling systems. In some embodiments, a computer device receives flow rate data characterizing the flow rate of a first fluid over a certain time range. The computer device also receives humidity data characterizing the humidity level over the same time range. Furthermore, the computer device receives temperature data characterizing the temperature over the same time range. The computer device generates a training set of features based on the flow rate data, humidity data, and temperature data. Furthermore, the computer device trains a machine learning process based on the training set of features. The computer device stores machine learning model data characterizing the trained machine learning process in a data storage location.

Description

Technical Field

[0001] Cross - Reference to Related Applications This application claims priority based on U.S. Provisional Patent Application No. 63 / 520,391, filed on August 18, 2023, the entire content of which is hereby expressly incorporated by reference into this specification.

[0002] [[ID=IO]] [[ID=II]]The present disclosure generally relates to control systems, and more specifically to control systems for heating, ventilation, and air - conditioning (HVAC) systems.

Background Art

[0003] Heating, ventilation, and air - conditioning (HVAC) systems generally use a vapor - compression refrigeration cycle to cool ambient or room - temperature air. These HVAC systems can include heat exchangers that operate to remove heat from a refrigerant. For example, these heat exchangers can include plates or coils through which the refrigerant flows. By blowing air over these plates or coils with a fan, the refrigerant flowing through the plates or coils can be cooled. Although not as common, these heat exchangers may include a liquid desiccant for dehumidifying air during the cooling process. These conditioning systems may include a thermostat for setting a desired temperature and, in some embodiments, also a humidity level.

Summary of the Invention

[0004] In some embodiments, the device includes a non-transitory, machine-readable storage medium for storing instructions. The device also includes at least one processor connected to the non-transitory, machine-readable storage medium. This at least one processor is configured to execute instructions for receiving flow rate data characterizing the flow rate of a first fluid over a certain time range. This at least one processor is also configured to execute instructions for receiving humidity data characterizing the humidity level over the same time range. Furthermore, this at least one processor is also configured to execute instructions for receiving temperature data characterizing the temperature over the same time range. This at least one processor is also configured to execute instructions for generating a training set of features based on this flow rate data, humidity data, and temperature data. Furthermore, this at least one processor is configured to execute instructions for training a machine learning process based on this training set of features. This at least one processor is also configured to execute instructions for storing machine learning model data characterizing the trained machine learning process in a data storage location.

[0005] In another embodiment, a method using at least one processor includes the step of receiving flow rate data characterizing the flow rate of a first fluid over a time range. The method also includes the step of receiving humidity data characterizing the humidity level over the same time range. Furthermore, the method includes the step of receiving temperature data characterizing the temperature over the same time range. The method also includes the step of generating a training set of features based on this flow rate data, humidity data, and temperature data. Furthermore, the method includes the step of training a machine learning process based on the training set of features. The method also includes the step of storing machine learning model data characterizing the trained machine learning process in a data storage location.

[0006] In other embodiments, the system includes a harmonizing subsystem configured to adjust the airflow based on a regulating fluid. The system also includes a heat exchanger subsystem configured to remove heat from the regulating fluid. Furthermore, the system includes a controller configured to adjust at least one control setting of one or more of the aforementioned harmonizing subsystems and heat exchanger subsystems. The system also includes a computer unit communicatively connected to the controller. The computer unit is configured to receive sensor data from at least one sensor, characterizing at least one of temperature, humidity levels, and flow rates. The computer unit is also configured to generate a set of estimated features based on the sensor data. Furthermore, the computer unit is configured to apply a trained machine learning process to the set of estimated features to generate output data characterizing predicted operating values ​​of one or more of the aforementioned harmonizing subsystems and heat exchanger subsystems over future time intervals. The computer unit is also configured to transmit a signal to the aforementioned controller based on the output data to adjust the aforementioned at least one control setting. [Brief explanation of the drawing]

[0007] The following drawings illustrate specific embodiments of the present disclosure and are not intended to limit the scope of the present disclosure. The drawings are not to scale and are intended to be used in conjunction with the description in the following detailed description.

[0008] [Figure 1] This shows a liquid desiccant air-conditioning (LDAC) system according to one embodiment.

[0009] [Figure 2] Figure 1 shows an example of a computer device for LDAC control of the LDAC system according to one embodiment.

[0010] [Figure 3] Figure 1 shows a portion of the LDAC system according to one embodiment.

[0011] [Figure 4] This shows a timeline for training and applying a machine learning process according to one embodiment.

[0012] [Figure 5] A flowchart illustrating an example of a method for adjusting the control settings of the LDAC system shown in Figure 1, according to one embodiment, is shown.

[0013] [Figure 6] A flowchart illustrating an example of a method for training a machine learning process according to one embodiment is shown. [Modes for carrying out the invention]

[0014] In the following description, prior art features of heat and mass exchangers that are obvious to those skilled in the art will be omitted or briefly described. Various embodiments will be described in detail with reference to the drawings, but it should be noted that the same reference numerals throughout the drawings represent the same parts or assemblies. References to the various embodiments described herein are not intended to limit the scope of the appended claims. Furthermore, any examples described herein are not limiting and merely illustrate a selection of the many implementable embodiments within the appended claims. In addition, certain functions described herein may be used in combination with other described functions in each possible combination and permutation.

[0015] Unless otherwise specified herein, all terms shall be given the broadest and most reasonable interpretation, including the meaning implied herein, the meaning understood by those skilled in the art, and / or the meaning defined in dictionaries, specialized books, etc. It should also be noted that, unless otherwise specified, “a,” “an,” and “the” as used herein and in the appended claims shall be interpreted as including the plural form, and that the terms “includes” and / or “including” herein clearly indicate the presence of the described features, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof. In the detailed description of the invention, relative terms such as “horizontal,” “vertical,” “up,” “down,” “vertex,” and “bottom,” and their derivatives (e.g., “horizontally,” “downward,” “upward,” etc.) shall be interpreted as referring to the direction described or indicated in the drawings described herein. These relative terms are used for explanatory convenience and are not generally intended to require a specific orientation. Terms including "up" versus "down," "inside" versus "outside," and "longitudinal" versus "lateral" should be interpreted relative to each other, or, where appropriate, relative to an extension axis, rotation axis, or center of rotation. Terms related to connection or coupling, such as "connected" or "interconnected," refer to relationships in which structures are fixed or connected to each other directly or indirectly through intermediate structures, and movable or rigid connections or relationships, unless explicitly stated otherwise. Terms such as "operationally connected" or "operably connected" refer to connections, couplings, or relationships that enable the related structures to function as intended. Furthermore, terms related to communication, such as "communicatively coupled," may include wired or wireless connections that enable related structures to communicate with each other.

[0016] Embodiments of this disclosure generally relate to the control of heating, ventilation, and cooling (HVAC) systems, and more specifically, to the control of HVAC systems using machine learning processes. For example, a liquid desiccant air-conditioning (LDAC) system may include a conditioning system that uses a conditioning fluid (e.g., a liquid desiccant, water) to condition the air. For example, this conditioning system may receive a concentrated conditioning fluid from a storage tank and distribute the concentrated conditioning fluid throughout a mass transfer element to dehumidify an incoming airflow (e.g., outside air received through an operating damper and supplied by a blower fan). As this airflow is dehumidified, the conditioning fluid is diluted. The conditioning system recovers the diluted conditioning fluid and supplies it to be returned to the aforementioned storage tank or a separate storage tank. In some cases, the conditioning system distributes an operating fluid (e.g., water, a liquid desiccant) to cool the incoming airflow. For example, the harmonizing system may include a dehumidification stage for dehumidifying the airflow through which it passes, and a cooling stage for cooling the airflow through which it passes. Furthermore, the harmonizing system may discharge exhaust (e.g., air used for cooling equipment) to the outside (e.g., through an actuated damper) and supply air to the space to be cooled. In some embodiments, the harmonizing system returns a portion of the supplied air as exhaust. The LDAC system may also include a regenerator system that receives the diluted conditioning fluid from the aforementioned storage tank and concentrates the conditioning fluid by removing water from the diluted conditioning fluid. The regenerator system supplies the concentrated conditioning fluid back to the aforementioned storage tank.

[0017] Furthermore, the LDAC system may include an LDAC-controlled computer device, such as a server (e.g., a cloud-based server), which is communicably connected to one or more sensors. The sensors may include, among many others, flow sensors, humidity sensors, temperature sensors, carbon dioxide (CO2) sensors, carbon monoxide (CO) sensors, and smoke detectors. A flow sensor may measure the flow rate of a fluid, such as air, and generate sensor data characterizing this measured flow rate. A humidity sensor may detect the humidity level of the environment (e.g., 0% to 100%) and generate sensor data characterizing this detected humidity level. A temperature sensor may detect the temperature of the environment and generate sensor data characterizing this detected temperature. A CO2 sensor may detect the CO2 concentration and generate sensor data characterizing this detected amount of CO2, while a smoke detector may generate sensor data when smoke is detected. The LDAC-controlled computer device can receive sensor data from one or more of these sensors.

[0018] Furthermore, the LDAC control computer device may be communicatively connected to the LDAC controller. In some cases, the LDAC controller is or includes a thermostat. The LDAC controller may be communicatively connected to the aforementioned harmonizing system and / or regenerator system and may supply control signals to the harmonizing system and / or regenerator system to adjust one or more control settings of the harmonizing system and / or regenerator system. Control settings may include, among many others, a desired temperature setting, a desired humidity level setting, a desired flow rate setting for (e.g., supply air, exhaust air, return air, working fluid, conditioning fluid), a desired concentration of the conditioning fluid, an on or off setting for a fan (e.g., supply air fan), or an on or off setting for one or more of the harmonizing system and regenerator system. The LDAC control computer device can send one or more control signals to the aforementioned LDAC controller to adjust these control settings and other control settings. In some embodiments, the LDAC controller includes a thermostat and may optionally include one or more additional or alternative sensors, such as the aforementioned sensors described herein.

[0019] As described herein, the LDAC-controlled computer device may determine the predicted operating values ​​of the aforementioned LDAC system using trained machine learning or artificial intelligence processes. For example, to train these machine learning or artificial intelligence processes, the LDAC-controlled computer device may generate a training dataset and, optionally, a validation dataset by adaptively selecting features from historical sensor data and historical control setting values ​​for corresponding time intervals. This historical sensor data may include, for example, historical temperature, humidity, and flow rate sensor values. This historical control setting value may characterize temperature, humidity levels, and flow rate settings (for example, values ​​set by the user in the thermostat of the aforementioned LDAC controller). In some embodiments, as described herein, these selected features may include, for example, desired temperature, desired humidity levels, and desired flow rates obtained from users by surveying them through a mobile phone application, a computer application, or an application on other suitable electronic device. The LDAC-controlled computer device may perform operations to train the aforementioned machine learning or artificial intelligence processes based on the training dataset. For example, as described herein, an LDAC-controlled computer device may train the machine learning or artificial intelligence process until at least one metric is met (for example, until the loss function falls below a threshold).

[0020] When the indicator is met, the LDAC-controlled computer may perform operations to validate the machine learning or artificial intelligence process based on a validation dataset. The training described above is completed when this validation satisfies one or more indicators (for example, when the loss function of a calculated precision value, a calculated recall value, or a calculated Area Under Curve (AUC) for a Receiver Operating Characteristic (ROC) curve or Precision-Recall (PR) curve meets the corresponding threshold). Furthermore, based on the results of these training processes, the LDAC-controlled computer may generate model coefficients, parameters, thresholds, and / or other modeling data that comprehensively define the trained machine learning or artificial intelligence process, and may store these generated model coefficients, parameters, thresholds, and / or modeling data in one or more actual non-temporary memory locations.

[0021] Once trained, the LDAC-controlled computer may apply the trained machine learning or artificial intelligence process to sensor data to generate output data characterizing predicted operating values ​​of the LDAC system at future time intervals (e.g., hourly intervals one day, one week, or one month later). For example, the LDAC-controlled computer may receive sensor data from one or more sensors and apply the trained machine learning or artificial intelligence process to this sensor data to generate output data characterizing one or more predicted operating values. In some embodiments, the LDAC-controlled computer applies the trained machine learning or artificial intelligence process to the aforementioned sensor data and also to load data characterizing building loads (e.g., indoor loads), such as the thermal load of a corresponding building. These predicted operating values ​​may be, for example, predicted temperature (e.g., room temperature), humidity levels, or gas concentrations (e.g., CO2 concentration) for future time intervals. In some embodiments, the LDAC control computer device 102 applies the trained machine learning or artificial intelligence process to input data characterizing future operational control settings of the LDAC system that create specific supply air flow rates and / or conditions, and generates output data characterizing room temperature, humidity, and / or carbon dioxide concentration. In this way, the LDAC system may predict the room's response to the LDAC settings based on the predicted thermal load (sensible and / or latent heat) of the building.The trained machine learning or artificial intelligence process may include ensemble learning or decision tree processes, such as gradient boosting decision tree processes (e.g., XGBoost model), clustering processes, unsupervised learning processes (e.g., k-means method, mixture model, hierarchical clustering algorithm, etc.), semi-supervised learning processes, supervised learning processes, statistical processes (e.g., multinomial logistic regression model, etc.), random forests, neural networks such as artificial neural networks or deep neural networks, or association rule processes (e.g., Apriori algorithm, Eclat algorithm, FP-growth algorithm, etc.).

[0022] Based on the generated output data, the computer device for LDAC control may adjust one or more control settings of the LDAC system. For example, based on the output data, the computer device for LDAC control may send a signal to the LDAC controller to adjust the flow rate of the conditioning fluid of the aforementioned conditioning system to change the humidity level of the supplied air (e.g., increasing the flow rate of this conditioning fluid tends to lower the humidity level of the supplied air, and decreasing the flow rate of this conditioning fluid tends to increase the humidity level of the supplied air). As another example, based on the output data, the computer device for LDAC control may send a signal to the LDAC controller to adjust the concentration of the conditioning fluid generated by the aforementioned regenerator system to change the humidity level of the supplied air (e.g., increasing the concentration of this conditioning fluid tends to lower the humidity level of the supplied air, and decreasing the concentration of this conditioning fluid tends to increase the humidity level of the supplied air).

[0023] In another embodiment, based on the output data, the LDAC control computer device may send a signal to the LDAC controller to change the exhaust ratio (e.g., the ratio of exhaust to supply air) to change the supply air temperature. In some embodiments, based on the output data, the LDAC control computer device may send a signal to the LDAC controller to adjust the exhaust flow rate of the harmonizing system to change the supply air temperature (for example, increasing the exhaust flow rate tends to decrease the supply air temperature, and decreasing the exhaust flow rate tends to increase the supply air temperature). For example, the LDAC controller may send a signal to adjust the fan speed of the harmonizing system. In some embodiments, the LDAC controller may receive sensor data indicating the exhaust flow rate and determine whether the exhaust flow rate has been reached based on this sensor data. In some embodiments, the LDAC controller determines the flow rate of two or more fluids based on receiving measured values ​​for the flow rates of those fluids. For example, the LDAC controller may receive sensor data indicating the flow rate of outside air, the flow rate of return air, and the flow rate of exhaust air. The LDAC controller may determine the flow rate of supply air based on the difference obtained by subtracting the exhaust flow rate from the sum of the outside air flow rate and the return air flow rate. In some embodiments, based on the output data, the LDAC control computer may send a signal to the LDAC controller to adjust the flow rate of the supply air based on the determined flow rate of the supply air. In any of the embodiments described above, the relative amount of air flow (supply air, exhaust air, etc.) may be controlled and / or modified using one or more dampers.

[0024] The LDAC control computer may adjust the control settings of one or more LDAC systems prior to the future time interval to reach a desired operating value during that future time interval. For example, the generated output data may indicate that the predicted room temperature during the future time interval starting in 4 hours is 90 degrees (e.g., based on predicted building load data, or to minimize energy consumption, or to minimize energy costs, or to create a desired load shape to match energy supply from the utility company, or to minimize greenhouse gases). Thus, based on a desired temperature of 75 degrees during the future time interval, the LDAC control computer may, after 2 hours, instruct the LDAC controller to increase the exhaust ratio so that the temperature during the future time interval (e.g., at the start) is at least 75 degrees. Similarly, the generated output data may indicate that the predicted humidity level of the room during the future time interval is 75%. By pre-dehumidifying the air prior to the expected humidity load, the thermal storage desiccant can last longer during the occurrence of high humidity loads, increasing the amount of heat stored and the events of peak load reduction. Thus, based on 50% of the desired humidity level of the room during the future time interval, the LDAC control computer device increases the flow rate of the aforementioned conditioning fluid to the LDAC controller after 2 hours, thereby lowering the humidity level so that it can reach 0% by the future time interval. In fact, by predicting the room's response to the LDAC, the speed at which the building load is handled can be determined. For example, a common problem with conventional air conditioning systems is that they operate with little change in the supply air conditions, without considering the actual cooling load of the building. This can lead to the building being overcooled when the system is started, resulting in an uncomfortable or undesirable indoor environment.In contrast, the computer device with LDAC control can adjust the supply air conditions of the LDAC according to the cooling load of this building, and as a result, make the transition between two different conditions smoother and / or create a more stable indoor environment (for example, when keeping the indoor setting constant).

[0025] In some embodiments, this desired temperature may be stored in the memory device as the recommended temperature for this future time interval. For example, this memory device may store an air conditioning schedule that specifies the recommended temperature for corresponding time intervals for each day of the week, and in some cases may also store an air conditioning schedule that specifies the recommended humidity level. For example, this air conditioning schedule may specify the recommended temperature and the recommended humidity level for every 4-hour interval for each day of the week. In some embodiments, this time interval is variable. For example, this air conditioning schedule may include a first time interval every 2 hours, a second time interval every 8 hours, and a third time interval every 14 hours for a specific day.

[0026] In some embodiments, the LDAC-controlled computer system allows the user to determine time intervals (e.g., days of the week, start time, end time) and select recommended temperature and humidity levels for each time interval. The LDAC-controlled computer system may then generate and / or update the air conditioning schedule and store it in the aforementioned memory device. The LDAC-controlled computer system may then apply a trained machine learning or artificial intelligence process to sensor data and / or building load data to ensure that the recommended temperature and / or humidity levels are achieved during the time interval. In some cases, the LDAC-controlled computer system receives polling information from one or more users, which identifies the user's recommended temperature and / or humidity levels. Based on this polling information, the LDAC-controlled computer system may generate and / or update the air conditioning schedule in the aforementioned memory device. For example, the LDAC control computer may calculate the average recommended temperature (for example, over a corresponding time interval) and update the air conditioning schedule with that average temperature. Similarly, the LDAC control computer may average the recommended humidity levels (for example, over a corresponding time interval) and update the air conditioning schedule with that average humidity level. Furthermore, these settings (e.g., air conditioning schedules) may actually be subjective, for example, by surveying the user with questions such as "I am comfortable," "I am cold," or "I am humid." Based on these questions, the LDAC control computer can determine target temperature and humidity levels that are most likely to satisfy the majority of occupants. The system can then evaluate the effectiveness of these adjustments by detecting changes in the frequency of complaints or by requesting new information from the user through prompts such as "We have adjusted the air conditioning to make it more comfortable. Are you more comfortable now?" and processing the answers to these additional survey questions.

[0027] In some embodiments, the LDAC control computer determines the air conditioning schedule based on the past settings of the LDAC system. For example, the LDAC computer system may calculate statistical indicators, such as the average value over a certain period (e.g., 3 days, 1 week, 1 month, 3 months, 1 year, seasonally, etc.) for the same time interval (e.g., time zone) for LDAC controller settings provided by the user (e.g., thermostat temperature settings and / or humidity settings). For example, the LDAC control computer may calculate the average temperature setpoint and the average humidity level setpoint for the corresponding time zone during that certain period. Based on the calculated average temperature setpoint and average humidity level setpoint, the LDAC control computer may generate an air conditioning schedule.

[0028] In some embodiments, the LDAC control computer determines the time and how to adjust the aforementioned control settings based on temperature tables and / or humidity tables stored in the aforementioned memory device. For example, the temperature table may indicate the time required to raise or lower the temperature of the spatial volume by a corresponding amount for various water flow rates and / or exhaust flow rates. Similarly, the humidity table may indicate the time required to raise or lower the humidity level of the spatial volume by a corresponding amount for various regulating fluid flow rates or concentrations. The LDAC control computer may determine the time and amount to adjust the aforementioned control settings based on output data characterizing predicted operating values ​​of the LDAC system in the future time interval, such as temperature and humidity levels, and the temperature tables and / or humidity tables.

[0029] In some cases, the LDAC control computer determines the aforementioned temperature and / or humidity tables based on past sensor readings and past setting values ​​of the LDAC system. For example, the LDAC control computer may determine a statistical measure, such as an average value, based on the past sensor readings and past setting values. For example, given a starting temperature or starting humidity level, an ending temperature or ending humidity level, a starting value for the control setting (e.g., initial water flow rate), and a final value for the control setting (e.g., adjusted water flow rate), the LDAC control computer may determine the time it took for the LDAC system to reach a specific temperature or humidity level. In addition, such a determination may further consider past load profiles at specific start times and days of the week within a specific period. Therefore, during periods of peak building occupancy, peak sunlight, or other building conditions (e.g., open windows and doors), it may take longer to achieve a desired temperature or humidity change than during periods of low building occupancy or low sunlight. Furthermore, by analyzing the behavior of this system, it can be used to identify the instantaneous heat and humidity loads of the building, and this data can be stored and analyzed to predict future loads. In addition, by comparing the predicted load with the instantaneous load, events that affect the operation of this system may be inferred. For example, based on the comparison of the predicted load and the instantaneous load, the LDAC control computer may determine that all windows are still open and instruct it to turn off the aforementioned air conditioner.

[0030] In some cases, the LDAC-controlled computer device may apply a second trained machine learning or artificial intelligence process to the current sensor readings and current control settings to determine the aforementioned temperature and / or humidity tables (e.g., table values). The second trained machine learning or artificial intelligence process may include ensemble learning or decision tree processes, such as gradient boosting decision tree processes (e.g., XGBoost models), clustering processes, unsupervised learning processes (e.g., k-means algorithms, mixed models, hierarchical clustering algorithms, etc.), semi-supervised learning processes, supervised learning processes, statistical processes (e.g., multinomial logistic regression models, etc.), random decision forests, neural networks such as artificial neural networks or deep neural networks, or association rule processes (e.g., Apriori algorithms, Eclat algorithms, FP-growth algorithms, etc.). The LDAC-controlled computer device may train this second machine learning or artificial intelligence process based on training datasets and, in some embodiments, validation data sets, as described herein. These training and validation datasets include historical sensor readings and characteristics of past control settings. In some embodiments, since the LDAC system can independently change the humidity and temperature of the supply air, the LDAC system performs experimental processes in which the temperature and humidity are changed independently (for example, significantly lowering the temperature without reducing the moisture content of the supply air, or vice versa), and the room conditions based on these changes are recorded to generate training data. In this way, this training data may isolate and characterize the effects of each change.

[0031] Referring to the drawings, Figure 1 shows an example of an LDAC system 100 including a harmonizing system 110, a conditioning fluid (CF) tank 114, and a regenerator system 112, which operates to supply supply air 135 to a building 101. For example, the CF tank 114 may store a conditioning fluid, such as a liquid desiccant. The harmonizing system 110 may receive concentrated conditioning fluid 141 from the CF tank 114 and use this concentrated conditioning fluid 141 to dehumidify an outside air flow 131 (for example, in a dehumidification location). The harmonizing system 110 may recover diluted conditioning fluid (for example, conditioning fluid used to dehumidify the outside air flow 131) and may also supply diluted conditioning fluid 143 back to the CF tank 114.

[0032] Furthermore, in some embodiments, the Harmonization System 110 may cool the outside airflow 131 at a cooling location (for example, using a working fluid such as water). After passing through the dehumidification location, and optionally the cooling location, the outside airflow 131 is supplied to the building 101 as a supply airflow 135. Optionally, the Harmonization System receives return air 137 from the building 101, mixes the return air 137 with the outside airflow 131 to produce mixed air, and then dehumidifies and / or cools the mixed air to supply the supply airflow 135. Optionally, a portion of the mixed air (i.e., return air 137 mixed with outside air 131) is used to cool the components of the Harmonization System 110. Optionally, a portion of the outside air 131 is used to cool the components of the Harmonization System 110 and discharged from the Harmonization System 110 to the external environment as exhaust 113.

[0033] The regenerator system 112 receives diluted conditioning fluid 145 from the CF tank 114, concentrates the diluted conditioning fluid 145 to produce concentrated conditioning fluid 147, and supplies it back to the CF tank 114. The regenerator system 112 may also receive an outside air flow 131 to cool its components, and may also supply an exhaust flow 133 to return the exhaust environment to the external environment.

[0034] The LDAC system 100 also includes an LDAC-controlled computer device 102 which is communicably connected to a database 116, an LDAC controller 105, and several sensors 120A, 120b, 120C, and 120D. Each of the several sensors 120A, 120b, 120C, and 120D may be, in some embodiments, a flow sensor, a humidity sensor, a temperature sensor, a CO2 sensor, a CO sensor, or a smoke detector. As shown in the figure, the several sensors 120A, 120b, 120C, and 120D may be located throughout the building 101. In some embodiments, the sensors 120A, 120b, 120C, and 120D may be located near or inside an intake duct or exhaust duct through which either or both of the supply air 135 or the return air 137 pass. In some embodiments, one or more sensors 120A, 120b, 120C, 120D may be located near or inside the intake for circulating air for the building, and / or near or inside the exhaust for exhaust air for the building. In some embodiments, one or more sensors 120A, 120b, 120C, 120D, such as one or more of a temperature sensor and a humidity sensor, may be located in each room of a multi-room building. For example, sensors 120A, 120b, 120C, 120D may be located near the doorway leading out of each room. In some cases, sensors 120A, 120b, 120C, 120D may be located within the aforementioned harmonizing system 110 or within the aforementioned regenerator system 112 to measure, for example, exhaust flow rate, return air flow rate, supply air flow rate, regulating fluid flow rate, etc. The LDAC-controlled computer device 102 can receive sensor data from each of the multiple sensors 120A, 120b, 120C, and 120D.

[0035] The LDAC controller 105 is communicatively connected to the harmonic system 110 and, in some embodiments, also communicatively connected to the regenerator system 112. The LDAC controller 105 includes, for example, one or more processors and transceivers, and transmits signals to the harmonic system 110 and / or the regenerator system 112 to adjust the control settings of the harmonic system 110 and / or the regenerator system 112. For example, the LDAC controller 105 transmits a signal to the harmonic system 110 to allow the harmonic system 110 to adjust the flow rate of the regulating fluid in the harmonic system. In another embodiment, the LDAC controller 105 transmits a signal to the harmonic system 110 to allow the flow rate of the working fluid water flowing through the harmonic system 110 to adjust the exhaust flow rate of the harmonic system 110. The LDAC controller 105 transmits a signal to the regenerator system 112 to allow the regenerator system 112 to adjust the concentration of the regulating fluid. For example, the LDAC controller 105 is one or more thermostats, or includes one or more thermostats.

[0036] Furthermore, as described herein, the LDAC-controlled computer device 102 can transmit signals to the LDAC controller 105 to cause the LDAC controller 105 to adjust any of the control settings of the harmonic system 110 and / or the regenerator system 112. For example, the LDAC-controlled computer device 102 may receive sensor data from any of the sensors 120A, 120b, 120C, and 120D, and transmit signals to the LDAC controller 105 based on the sensor data to adjust one or more control settings. In an embodiment, the LDAC-controlled computer device 102 may establish any of the trained machine learning or artificial intelligence processes described herein. Furthermore, the LDAC-controlled computer device 102 may receive sensor data from one or more of the sensors 120A, 120b, 120C, and 120D, and may generate features based on the received sensor data. The LDAC control computer 102 may apply the aforementioned established and trained machine learning or artificial intelligence processes to the generated features to generate output data characterizing the predicted operating values ​​of the LDAC system 100 in future time intervals. As described herein, each predicted operating value may be, for example, a predicted temperature or humidity level for a future time interval. Based on this generated output data, the LDAC control computer 102 may send a signal to the LDAC controller 105 to adjust one or more control settings.

[0037] As described herein, the LDAC control computer unit 102 may send a signal to the LDAC controller 105 to adjust these control settings at a point in time prior to the future time interval. For example, the LDAC control computer unit 102 may send this signal to the LDAC controller 105 to reach a desired operating value (e.g., a desired temperature and / or humidity level) in the future time interval.

[0038] In some embodiments, these desired operating values ​​are stored in a database 116. For example, the database 116 stores an air conditioning schedule that identifies preferred temperature and, optionally, preferred humidity levels for the corresponding time periods for each day of the week. In some embodiments, a user can input into the LDAC-controlled computer device 102 to add, delete, and / or adjust desired operating values ​​for the air conditioning schedule stored in the database 116. The LDAC-controlled computer device 102 can also store individual user data and profiles. In some embodiments, the database 116 stores survey data, for example, that includes questionnaire information characterizing preferred operating values. The LDAC-controlled computer device 102 may adjust the air conditioning schedule in the database 116 based on this survey data.

[0039] In some embodiments, the LDAC control computer 102 determines the desired operating values ​​for the air conditioning schedule based on past settings of the LDAC system 100. For example, the LDAC control computer 102 may calculate a statistical index of user-provided LDAC controller settings (e.g., thermostat temperature settings and / or humidity settings) for time intervals repeated over a period of time. The LDAC control computer 102 may update the air conditioning schedule based on this statistical index.

[0040] In some embodiments, the LDAC control computer 102 determines when and by how much to adjust the control settings based on temperature and / or humidity tables stored in the database 116. As described herein, the temperature table may indicate the time required to raise or lower the temperature of the space volume by a corresponding amount for various water and / or exhaust flow rates, while the humidity table may indicate the time required to raise or lower the humidity level of the space volume by a corresponding amount for various regulating fluid flow rates and concentrations. Similar to these temperature and / or humidity tables, the LDAC control computer 102 may determine when and by how much to adjust the control settings based on output data generated by the aforementioned established machine learning or artificial intelligence processes, in order to achieve preferred operating values ​​in the aforementioned future time intervals.

[0041] In some cases, the LDAC control computer device 102 may apply a second trained machine learning process or artificial intelligence process to the sensor data received from one or more of the sensors 120A, 120B, 120C, 120D and the control settings (for example, identifying the current settings) to determine the values ​​in the temperature table and / or humidity table, and may store these values ​​in the temperature table and / or humidity table of the database 116.

[0042] In some embodiments, the LDAC control computer 102 provides a warning display based on output data characterizing the predicted operating values ​​of the LDAC system 100. For example, if this output data indicates that a predicted operating value, such as the level of a CO2 sensor, exceeds a threshold (e.g., is above the threshold), the LDAC control computer 102 may generate and display a warning message. In some embodiments, the LDAC control computer 102 may generate and send a communication (e.g., an email, an SMS message) notifying the warning. This warning display may indicate the sensor and the predicted operating value corresponding to that sensor. In some embodiments, the LDAC control computer 102 may send a signal to the LDAC controller 105 to adjust one or more control settings of the LDAC system 100, such as adjusting (e.g., reducing) the amount of return air 137 mixed with the outside air flow 131. In this way, the return air 137 may be supplied to the aforementioned external environment along with the exhaust flow 113, rather than being recirculated to the building 101. In some cases, this warning may indicate a predictive maintenance event (e.g., maintenance required for the Harmonization System 110 and / or the Regenerator System 112) or an event related to the Building 101 (e.g., a window is open, overcrowding, etc.). In some embodiments, the LDAC control computer 102 determines whether a significant change has occurred in the Building 101 or the LDAC system 100 and generates this warning based on that determination. For example, the LDAC control computer 102 may compare the predicted operating values ​​of the LDAC system 100 with corresponding measured values ​​(e.g., current temperature, humidity, CO2 concentration, etc.) to determine whether an error condition exists (e.g., whether the difference between the aforementioned predicted operating values ​​and the aforementioned measured operating values ​​exceeds a threshold). Furthermore, based on the determined error condition, the LDAC system 100 can determine the most likely cause of the error condition, such as a door being open, a window being open, the LDAC system 100 needing maintenance, or other circumstances that could cause the error condition.

[0043] Figure 2 shows an embodiment of the LDAC-controlled computer unit 102 of the LDAC system 100 of Figure 1. The LDAC-controlled computer unit 102 comprises one or more processors 201, working memory 202, one or more input / output devices 203, instruction memory 207, transceivers 204, one or more communication ports 209, and a display 206, all of which are operably connected to one or more data buses 208. The data buses 208 enable communication between these various devices. The data buses 208 may include wired or wireless communication channels.

[0044] Each of the processors 201 may include one or more independent processors having one or more cores. Each of these independent processors may have the same structure or different structures. The processor 201 may include one or more central processing units (CPUs), one or more graphics processing units (GPUs), application-specific integrated circuits (ASICs), digital signal processors (DSPs), and the like. The processor 201 may be configured to perform specific functions or operations by executing them using executable code stored in the instruction memory 207. For example, the processor 201 may be configured to perform one or more of any functions, methods, or operations disclosed herein.

[0045] The instruction memory 207 can store instructions that are accessible (e.g., readable) and executable by the processor 201. For example, the instruction memory 207 may be a computer-readable non-temporary storage medium, such as read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, removable disk, CD-ROM, non-volatile memory, or other suitable memory. In this embodiment, the instruction memory 207 includes LDAC machine learning model data 207A, which contains instructions characterizing a model of a trained machine learning or artificial intelligence process as described herein. For example, one or more processors 201 may retrieve the LDAC machine learning model data 207A from the instruction memory 207, and may execute the LDAC machine learning model data 207A to establish one of the trained machine learning or artificial intelligence processes described herein.

[0046] Furthermore, the processor 201 can store data in the working memory 202 and read data from the working memory 202. For example, the processor 201 can store a set of working instructions, such as instructions loaded from the instruction memory 207, in the working memory 202. The processor 201 can also use the working memory 202 to store dynamic data generated during the operation of the LDAC-controlled computer device 102. The working memory 202 may be random access memory (RAM), such as static random access memory (SRAM) or dynamic random access memory (DRAM), or other suitable memory.

[0047] The input-output device 203 may include any suitable device capable of inputting or outputting data. For example, the input-output device 203 may include one or more of the following: a keyboard, touchpad, mouse, stylus pen, touchscreen, physical buttons, speaker, microphone, or other suitable input or output devices. The input-output device 203 may allow the user to provide an input to select or characterize a preferred operating value, such as those described herein.

[0048] The communication port 209 may include, for example, a serial port such as a Universal Asynchronous Receiver / Transmitter (UART) connection or a Universal Serial Bus (USB) connection, or other suitable communication port or connection. In some embodiments, the communication port 209 enables the programming of executable instructions into the instruction memory 207. In some embodiments, the communication port 209 enables the transfer (e.g., uploading or downloading) of data such as survey data (e.g., questionnaire information).

[0049] The display 206 can display the user interface 205. The user interface 205 enables user interaction with the LDAC-controlled computer device 102. For example, the user interface 205 may be a user interface for an application that allows the user to input preferred operating values, for example, to update the air conditioning schedule stored in the database 116. In some embodiments, the user can exchange information with the user interface 205 by operating the input-output device 203. In some embodiments, the display 206 may be a touchscreen, and the user interface 205 is displayed on the touchscreen.

[0050] The transceiver 204 enables communication with a network, such as a wireless network, established between the LDAC-controlled computer device 102 and the LDAC controller 105. For example, the transceiver 204 may connect to a WiFi, Bluetooth, cellular, or other suitable wireless network, and may transmit signals (e.g., data) to the LDAC controller 105 and receive signals from the LDAC controller 105. In some cases, the transceiver 204 may additionally or alternatively communicate with one or more sensors, such as sensors 120A, 120B, 120C, and 120D, via the wireless network. The processor 201 is operable to receive data from or transmit data to the wireless network via the transceiver 204.

[0051] Figure 3 shows an exemplary portion of the LDAC system 100 of Figure 1. In this embodiment, the LDAC-controlled computer device 102 can receive sensor data from various sensors. For example, the LDAC-controlled computer device 102 can receive flow rate data 313 from one or more flow rate sensors 302, temperature data 315 from one or more temperature sensors 304, and humidity data 317 from one or more humidity sensors 306. The LDAC-controlled computer device 102 may analyze the flow rate data 313, temperature data 315, and humidity data 317 to generate sensor data elements 330, and store these sensor data elements 330 in the database 116. These sensor data 330 may include a temperature value 330A, a humidity value 330B, and a flow rate value 330C.

[0052] The database 116 may also store LDAC predictive model data 360, which includes model coefficients, parameters, thresholds, and / or other modeling data, and which may collectively specify one or more of the trained machine learning or artificial intelligence processes described herein. For example, an LDAC-controlled computer device 102 may retrieve the LDAC predictive model data 360 from the database 116 and establish one of the trained machine learning or artificial intelligence processes described herein based on this LDAC predictive model data 360.

[0053] The database 116 may also store an air conditioning schedule 362, a temperature table 364, and a humidity table 366. As described herein, the air conditioning schedule 362 stores desired operating values ​​(e.g., preferred temperature and humidity levels) for corresponding time intervals. Furthermore, the temperature table 364 stores values ​​indicating the time required to raise or lower the temperature in a space by a corresponding amount (e.g., for various water and / or exhaust flow rates), while the humidity table 366 stores values ​​indicating the time required to raise or lower the humidity in a space by a corresponding amount (e.g., for various regulating fluid flow rates or concentrations).

[0054] Furthermore, the database 116 may store survey data 332, which includes preferred operating values, such as preferred temperature and / or humidity levels in a corresponding time interval. For example, the survey data 332 may include statistical measurements of questionnaire information received from various users (e.g., residents of a building). In some cases, the survey data 332 may include preferred operating values ​​from one or more users for each of several rooms in a building, such as rooms in building 101.

[0055] Furthermore, the database 116 may store control setting data 334 containing values ​​(e.g., current values ​​and / or historical values) for one or more control settings of the harmonic system 110 and the regenerator system 112. For example, the control setting data 334 may include temperature setting values ​​334A (e.g., current temperature setting values ​​and past temperature setting values), humidity setting values ​​334B (e.g., current humidity level setting values ​​and past humidity level setting values), and flow rate setting values ​​334C (e.g., current flow rate setting values ​​and past flow rate setting values) of the LDAC system 100. The LDAC control computer may receive (e.g., on request) control setting signals 311 characterizing one or more current control settings of the harmonic system 110 and / or the regenerator system 112.

[0056] The database 116 may also store building load data 370, which characterizes the load data of a building, such as building 101. For example, the building load data 370 may include one or more of the following: heat load 370A, sensor load 370B, and occupancy schedule 370C. Heat load 370A may characterize one or more heat loads of a building, such as the heat load of a room in building 101. Sensor load 370B may characterize the load of a sensor, such as the load of a CO2 sensor installed in a room in building 101. Occupancy schedule 370C may characterize the expected occupancy status of a building or a room in a building during a specific time period. In some cases, the LDAC control computer device 102 may apply a trained machine learning or artificial intelligence process described herein, such as a trained machine learning or artificial intelligence process characterized by LDAC prediction model data 360, to a portion of the store building load data 370 to generate output data characterizing one or more predicted operating values.

[0057] The database 116 may also include parameters 380 for the harmonic system and parameters 382 for the regenerator system. The parameters 380 for the harmonic system may include the operating parameters and / or performance parameters of the corresponding harmonic system (e.g., harmonic system 110). The parameters 382 for the regenerator system may include the operating parameters and / or performance parameters of the corresponding regenerator system (e.g., regenerator system 112).

[0058] The LDAC control computer 102 may generate features based on sensor data 330, and may apply established and trained machine learning or artificial intelligence processes to the generated features to generate output data characterizing predicted operating values ​​of the LDAC system in future time intervals. As described herein, each predicted operating value may be, for example, a predicted temperature or humidity level for the aforementioned future time interval. Based on this generated output data, the LDAC control computer 102 may send adjustment signals 321 to the LDAC controller 105 to adjust one or more control settings. For example, the LDAC control computer 102 may send adjustment signals 321 to the LDAC controller 105 to reach a desired operating value (for example, a value indicated in adjustment schedule 362) in the aforementioned future time interval.

[0059] In some embodiments, the LDAC control computer 102 determines when and by how much to adjust the control settings based on the temperature table 364 and / or humidity table 366 stored in the database 116. For example, the LDAC control computer 102 searches for time entries and adjustment entries in the temperature table 364 and / or humidity table 366 that correspond to predicted operating values ​​(e.g., predicted temperature) generated by a trained machine learning or artificial intelligence process for spatial volumes such as the volume of a room in building 101 or the spatial volume of building 101, and the current operating value (e.g., current temperature) corresponding to the predicted operating value. The time entry may indicate the amount of time required to adjust from the current operating value to the predicted operating value, and the adjustment entry may indicate the amount of adjustment required for that adjustment. The LDAC control computer 102 may send an adjustment signal 321 to the LDAC controller 105 at the determined time before the future time interval and adjust the corresponding control settings by the amount of adjustment.

[0060] Figure 4 shows an exemplary timing chart 400 for training one of the machine learning or artificial intelligence processes described herein, and for applying this trained machine learning or artificial intelligence process to the generated features to determine the predicted operating values ​​and for adjusting control settings to reach the desired operating values ​​in future time intervals.

[0061] For example, as illustrated, the LDAC-controlled computer device 102 may train a machine learning or artificial intelligence process during the model learning window 402. For example, as described herein, the LDAC-controlled computer device 102 may generate a set of training data by extracting adaptively selected feature values ​​from historical sensor data, such as the values ​​of sensor data 330, and historical control setting data, such as the values ​​of control setting data 334, for corresponding time intervals. In some embodiments, the LDAC-controlled computer device 102 generates the set of training data based on historical sensor data 330, historical control setting data 334, and one or more operating or performance parameters of the harmonic system 110 and the regenerator system 112. Furthermore, in some embodiments, the LDAC-controlled computer device 102 similarly generates a set of validation data. This set of validation data may be based on historical sensor data and control setting data corresponding to different, non-overlapping time intervals, rather than on historical sensor data and control setting data based on the aforementioned set of training data. Furthermore, during the model training window 402, the LDAC-controlled computer device may perform operations to train the aforementioned machine learning or artificial intelligence process based on the set of training data. For example, as described herein, the LDAC-controlled computer device 102 may train the machine learning or artificial intelligence process until at least one metric is met (e.g., until the loss function falls below a threshold).

[0062] When at least one of the indicators is met, the LDAC-controlled computer device 102 may perform processing to verify the aforementioned initially trained machine learning or artificial intelligence process during the verification window 404. For example, the LDAC-controlled computer device 102 may determine one or more indicators based on the output data generated by applying the aforementioned initially trained machine learning or artificial intelligence process to the set of verification data (for example, based on comparing the output data with expected data, such as in supervised learning), and may determine that the aforementioned initially trained machine learning or artificial intelligence process has been verified when the one or more indicators meet the corresponding thresholds.

[0063] Once verified, the LDAC control computer 102 may, in the inference window 406, apply a trained machine learning or artificial intelligence process to the received sensor data (e.g., sensor data received in real time) to generate output data that characterizes one or more predicted operating values ​​for a future period, such as the prediction window 410. Furthermore, as described herein, the LDAC control computer 102 may determine the timing and amount of adjustments to the control settings of the LDAC system 100 based on the predicted operating values, such as adjustment schedules, such as adjustment schedule 362, which characterize the desired operating values, temperature tables, such as temperature table 364, and / or humidity tables, such as humidity table 366.

[0064] Subsequently, the LDAC control computer device 102 may send an adjustment signal, such as an adjustment signal 321, to the LDAC controller 105 to adjust the corresponding control settings during the adjustment period, such as an adjustment window 408. In this embodiment, the adjustment time required to reach the desired operating state from the predicted operating state is shown, from the time the adjustment signal is sent until the time t6 when the prediction window 410 starts.

[0065] Figure 5 shows an exemplary process 500 for adjusting the control settings of an HVAC system, such as the LDAC system 100 in Figure 1. This exemplary process 500 may be performed by one or more computer devices, such as the LDAC control computer device 102.

[0066] Starting from block 502, sensor data for the LDAC system is received. For example, an LDAC control computer device 102 may receive sensor data from one or more sensors 120A, 120B, 120C, and 120D.

[0067] In block 504, a trained machine learning process is applied to the sensor data to generate output data that characterizes the predicted operating values ​​of the LDAC system in future time intervals. For example, as described herein, the LDAC control computer 102 may generate features based on the received sensor data. The LDAC control computer 102 may apply an established and trained machine learning or artificial intelligence process, characterized, for example, by LDAC prediction model data 360, to the generated features to generate output data that characterizes the predicted operating values ​​of the LDAC system 100 in future time intervals. As described herein, each predicted operating value may be, for example, a predicted temperature or humidity level in a future time interval such as a prediction window 410.

[0068] Furthermore, block 506 transmits a signal to adjust at least one control setting of the LDAC system based on the output data. For example, as described herein, the LDAC control computer 102 may determine the adjustment time and amount of the at least one control setting based on the generated output data, the adjustment schedule 362, and one or more entries in the temperature table 364 and humidity table 366. Furthermore, based on the adjustment time, the LDAC control computer 102 may transmit an adjustment signal 321 to the LDAC controller 105 within a time window, such as an adjustment control window 408, to adjust the at least one control setting of the LDAC system 100.

[0069] Figure 6 shows an exemplary process 600 for training a machine learning process. The exemplary process 600 may be executed by one or more computer devices, such as LDAC-controlled computer device 102.

[0070] Starting from block 602, sensor data is received at predetermined time intervals. For example, as described herein, the LDAC-controlled computer device 102 may obtain historical sensor data 330 and historical control setting data 334 from the database 116. Furthermore, in block 604, a feature training set is generated based on the sensor data. For example, the LDAC-controlled computer device 102 may generate a set of training data based on historical sensor data 330 and historical control setting data 334 corresponding to a first time interval (e.g., a 3-month time interval, a 6-month time interval, a 1-year time interval, etc.). In some embodiments, the feature training set is generated based on historical sensor data 330, historical control setting data 334, and one or more operating or performance parameters of the harmonic system 110 and regenerator system 112, such as harmonic system parameters 380 and regenerator system parameters 382.

[0071] Proceeding to block 606, a machine learning process is trained based on the training set of features. For example, an LDAC-controlled computer device 102 may input elements of the training data set into this machine learning process to generate output data. This training may be performed within a corresponding time window, such as the model training window 402.

[0072] In block 608, it is determined whether the training is complete. For example, based on the output data generated during the training performed in block 606, the LDAC-controlled computer device 102 may calculate one or more indicators and determine whether those indicators meet the corresponding thresholds (e.g., are below the threshold). If those indicators meet the corresponding thresholds, the training is complete. Otherwise, if one or more of those indicators do not meet the thresholds, the training is not complete. If the training is not complete, the process returns to block 602 to receive additional sensor data to continue the training. Otherwise, if the training is complete, the process proceeds to block 610.

[0073] In block 610, data characterizing the trained machine learning process is stored in a data storage location. For example, the LDAC-controlled computer device 102 may generate model coefficients, parameters, thresholds, and / or other modeling data that comprehensively define the trained machine learning process, and may store such generated model coefficients, parameters, thresholds, and / or modeling data in the database 116 as LDAC predictive model data 360.

[0074] In some embodiments, the device includes a non-temporary machine-readable storage medium for storing instructions. The device also includes at least one processor connected to the non-temporary machine-readable storage medium. The at least one processor is configured to execute the following instructions: receive flow rate data characterizing the flow rate of a first fluid over a time range; receive humidity data characterizing the humidity level over the time range; receive temperature data characterizing the temperature over the time range; generate a training set of features based on the flow rate data, humidity data, and temperature data; train a machine learning process based on the training set of features; and store machine learning model data characterizing the trained machine learning process in a data storage location.

[0075] In some embodiments of these above-described embodiments of the device, the at least one processor is configured to execute the following instructions: receive thermostat data characterizing the temperature settings of the thermostat over a given time range; and generate a feature training set based on the thermostat data. In some embodiments of any of these above-described embodiments of the device, the thermostat data characterizes one or more changes to the settings of the thermostat.

[0076] In some embodiment of any of the above-described embodiments of the device, the at least one processor is configured to perform the following instructions: receive survey data characterizing the satisfaction levels of one or more people with comfort within a given time range; and generate a training set of features based on the survey data.

[0077] In some embodiments of any of the above-described embodiments of the apparatus, the at least one processor is configured to execute the following instructions: generate output data in accordance with the training of the machine learning process; generate at least one metric value based on the output data; and complete the training of the machine learning process based on the at least one metric value.

[0078] In some embodiments of any of the above-described embodiments of the apparatus, the at least one processor is configured to perform the following instructions: establish the trained machine learning process based on the machine learning model data; receive second flow rate data from at least a first sensor characterizing the flow rate of a fluid to a location; receive second humidity data from at least a second sensor characterizing the humidity at the location; receive second temperature data from at least a third sensor characterizing the temperature at the location; generate a set of inferred features based on the second flow rate data, the second humidity data and the second humidity data; and apply the trained machine learning process to this set of inferred features to generate second output data. In some embodiments, the output data characterizes at least one of a predicted flow rate, a predicted humidity level and a predicted temperature in a future time interval. In some embodiments, the flow rate setting is a flow rate setting for a liquid desiccant in a liquid desiccant-type harmonization system. In some embodiments, the flow rate setting is a flow rate setting for water in a liquid desiccant-type harmonization system. In some embodiments, adjusting the flow rate setting changes the ratio of supply air to exhaust air in a liquid desiccant-type conditioning system. In some embodiments, adjusting the flow rate setting changes the flow rate of supply air. In some embodiments, the at least first sensor, the at least second sensor, and the at least third sensor are communicably connected to the at least one processor via a wireless network. In some embodiments, the at least one processor is configured to execute instructions that provide a warning display based on the output data. In some embodiments, the warning display indicates at least one of an open door or an open window.

[0079] In some embodiments, a method comprising at least one processor includes the following steps: receiving flow rate data characterizing the flow rate of a first fluid over a time range; receiving humidity data characterizing the humidity level over the time range; receiving temperature data characterizing the temperature over the time range; generating a training set of features based on the flow rate data, humidity data and temperature data; training a machine learning process based on the training set of features; and storing machine learning model data characterizing the trained machine learning process in a data storage location.

[0080] In some embodiments of these above-described methods, the method includes the steps of: receiving thermostat data that characterizes the temperature settings of the thermostat over a given time range; and generating a training set of features based on the thermostat data. In some embodiments of any of these above-described methods, the thermostat data characterizes one or more changes to the thermostat settings.

[0081] In some embodiments of these above-described methods, the method includes the steps of: receiving survey data characterizing the satisfaction level of one or more people with comfort within a given time range; and generating a training set of the features based on the survey data.

[0082] In some embodiments of these above-described methods, the method includes the following steps: generating output data in response to the training of the machine learning process; generating at least one metric value based on the output data; and completing the training of the machine learning process based on the at least one metric value.

[0083] In some embodiments of these above-described methods, the method includes the steps of: establishing a trained machine learning process based on machine learning model data; receiving second flow rate data from at least a first sensor characterizing the flow rate of a fluid to a location; receiving second humidity data from at least a second sensor characterizing the humidity at the location; receiving second temperature data from at least a third sensor characterizing the temperature at the location; generating a set of inferred features based on the second flow rate data, the second humidity data, and the second humidity data; and applying the trained machine learning process to this set of inferred features to generate second output data. In some embodiments, the output data characterizes at least one of a predicted flow rate, a predicted humidity level, and a predicted temperature in a future time interval. In some embodiments, the flow rate setting is a flow rate setting for a liquid desiccant in a liquid desiccant-type harmonization system. In some embodiments, the flow rate setting is a flow rate setting for water in a liquid desiccant-type harmonization system. In some embodiments, adjusting the flow rate setting changes the ratio of supply air to exhaust air in a liquid desiccant type conditioning system. In some embodiments, adjusting the flow rate setting changes the flow rate of supply air. In some embodiments, the at least first sensor, the at least second sensor, and the at least third sensor are communicably connected to the at least one processor via a wireless network. In some embodiments, the method includes the step of providing a warning indicator based on the output data. In some embodiments, the warning indicator indicates at least one of an open door or an open window.

[0084] In some embodiments, the system includes a harmonization subsystem configured to adjust airflow based on a regulating fluid. The system also includes a regenerator subsystem configured to remove heat from the regulating fluid. Furthermore, the system includes a controller configured to adjust one or more control settings of the harmonization subsystem and the regenerator subsystem. The system also includes a computer device communicatively connected to the controller. The computer device is configured to: receive sensor data from at least one sensor, characterizing at least one of temperature, humidity levels, and flow rate; generate a feature inference set based on the sensor data; apply a trained machine learning process to the feature inference set to generate output data characterizing one or more predicted operating values ​​of the harmonization subsystem and the regenerator subsystem over future time intervals; and transmit a signal to the controller based on the output data to adjust the at least one control setting.

[0085] Thus, in at least some embodiments, a machine learning-based process is used to predict environmental conditions at a future point in time, such as temperature and humidity levels, and to pre-adjust the control settings of the HVAC system to meet the desired environmental conditions at that future point in time. For example, as described herein, a computer device may receive sensor data from various sensors, including temperature sensors, humidity sensors, and airflow sensors. This computer device may generate features based on the sensor data and apply a trained machine learning process to these features to generate output data that characterizes predicted environmental conditions, such as predicted temperature and humidity levels. Furthermore, based on this output data, the computer device performs a process to adjust the HVAC system to meet the desired environmental conditions at a future point in time. As described herein, this machine learning process may be trained on features generated from historical sensor data and historical control settings of the HVAC system.

[0086] The various embodiments described herein are provided for illustrative purposes only and should not be construed as limiting the scope of the appended claims. Those skilled in the art will readily understand that various modifications and changes can be made without following the exemplary embodiments and applications shown and described herein, and without departing from the spirit and scope of the appended claims.

Claims

1. A device comprising a non-temporary machine-readable storage medium for storing instructions, and at least one processor connected to the non-temporary machine-readable storage medium, The aforementioned at least one processor is A command to receive flow rate data characterizing the flow rate of a first fluid over a certain time range; A command to receive humidity data characterizing the humidity level within the aforementioned time range; A command to receive temperature data characterizing the temperature in the aforementioned time range; Instructions for generating a training set of features based on the flow rate data and the humidity and temperature data; Instructions for training a machine learning process based on the aforementioned training set of features; and An instruction to store the machine learning model data characterizing the aforementioned trained machine learning process in a data storage location, A device characterized by being configured to perform the following.

2. The aforementioned at least one processor is A command to receive thermostat data characterizing the temperature setting of the thermostat within the aforementioned time range; and, An instruction to generate a training set of the features based on the thermostat data, The apparatus according to claim 1, characterized in that it is configured to perform the following.

3. The apparatus according to claim 2, characterized in that the thermostat data characterizes one or more changes to the settings of the thermostat.

4. The aforementioned at least one processor is A command to receive survey data characterizing the satisfaction level of one or more people with comfort within the aforementioned time range; and, Instructions for generating a training set of the features based on the aforementioned survey data, The apparatus according to claim 1, characterized in that it is configured to perform the following.

5. The aforementioned at least one processor is Instructions for generating output data in accordance with the training of the aforementioned machine learning process; An instruction to generate at least one index value based on the output data; and, An instruction to complete the training of the machine learning process based on at least one of the aforementioned indicator values, The apparatus according to claim 1, characterized in that it is configured to perform the following.

6. The aforementioned at least one processor is Instructions for establishing the trained machine learning process based on the machine learning model data; A command to receive at least a first sensor, second flow data characterizing the flow rate of a certain fluid to a certain location; A command to receive second humidity data characterizing the humidity of the location from at least a second sensor; A command to receive a second temperature data characterizing the temperature of the location from at least a third sensor; Instructions for generating a set of features to be inferred based on the second flow rate data, the second humidity data, and the second humidity data; and, Instructions for applying the trained machine learning process to the set of features to be inferred to generate second output data, The apparatus according to claim 1, characterized in that it is configured to perform the following.

7. The apparatus according to claim 6, characterized in that the output data characterizes at least one of a predicted flow rate, a predicted humidity level, and a predicted temperature in a future time interval.

8. The apparatus according to claim 6, wherein the at least one processor is configured to execute an instruction to adjust the flow rate setting based on the output data.

9. The apparatus according to claim 8, characterized in that the flow rate setting is the flow rate setting of the liquid desiccant in a liquid desiccant type harmony system.

10. The apparatus according to claim 8, characterized in that the flow rate setting is the flow rate setting for water in a liquid desiccant type harmony system.

11. The apparatus according to claim 8, characterized in that the ratio of supply air to exhaust air in a liquid desiccant type conditioned system changes by adjusting the flow rate setting.

12. The apparatus according to claim 8, characterized in that the flow rate of the supplied air changes by adjusting the flow rate setting.

13. The apparatus according to claim 6, characterized in that the at least first sensor, the at least second sensor, and the at least third sensor are communicated with the at least one processor via a wireless network.

14. The apparatus according to claim 6, wherein the at least one processor is configured to execute an instruction to provide a warning display based on the output data.

15. The apparatus according to claim 14, characterized in that the warning indicator shows at least one of an open door or an open window.

16. A method using at least one processor, The steps include: receiving flow rate data characterizing the flow rate of a first fluid over a certain time range; The steps include: receiving humidity data characterizing the humidity level within the aforementioned time range; The steps include: receiving temperature data that characterizes the temperature in the aforementioned time range; The steps include: generating a training set of features based on the flow rate data and the humidity and temperature data; The steps include: training a machine learning process based on the aforementioned training set of features; and The steps include storing the machine learning model data characterizing the trained machine learning process in a data storage location, A method characterized by comprising the following features.

17. The steps include: receiving thermostat data characterizing the temperature setting of the thermostat within the aforementioned time range; and The steps include generating a training set of the features based on the thermostat data, The method according to 16, characterized by comprising:

18. The steps include: receiving survey data characterizing the satisfaction level of one or more people with comfort within the aforementioned time range; and A step of generating a training set of the features based on the aforementioned survey data, The method according to 16, characterized by comprising:

19. A step of generating output data in accordance with the training of the machine learning process; a step of generating at least one metric value based on the output data; and A step of completing the training of the machine learning process based on at least one of the aforementioned indicator values, The method according to 16, characterized by comprising:

20. A harmonizing subsystem configured to adjust airflow based on a regulating fluid, A regenerator subsystem configured to remove heat from the aforementioned adjustment fluid, A controller configured to adjust one or more control settings of the harmonic subsystem and the regenerator subsystem, A system comprising a computer device that is communicatively connected to the controller, The aforementioned computer device Receive sensor data from at least one sensor that characterizes at least one of temperature, humidity level, and flow rate; Based on the aforementioned sensor data, an inference set of features is generated; Applying a trained machine learning process to the aforementioned feature inference set to generate output data characterizing one or more predicted operating values ​​of the harmonic subsystem and the regenerator subsystem in future time intervals; and A signal is sent to the controller based on the output data to adjust the at least one control setting. A system characterized by being configured in such a way.

Images