Temperature control system and method

Abstract

The present invention relates to a system for temperature control in one or more rooms, and a method for temperature control in one or more rooms.

Description

Technical Field

[0001] The present invention relates to a system for temperature control in one or more rooms, and a method for temperature control in one or more rooms. Background Technology

[0002] Most temperature control systems, whether cooling or heating, typically operate based on preset values, with a user-defined "target" temperature or preset value for each room (one temperature for all rooms, or different temperatures for all or some rooms). The principle of operation for a temperature control system is that further cooling or heating is required whenever the preset value is not reached, depending on whether the actual temperature is higher or lower than the target temperature. While such systems are simple and easy to control, unfortunately, they can lead to significant energy waste, where the temperature control system attempts to reach the preset value but is unable to do so due to energy-wasting conditions, or its operation is severely hampered. These energy-wasting conditions can vary considerably; relevant examples include open windows or doors, and unexpected / unwanted heat and / or cold sources in the room that counteract desired temperature (or heat) changes by introducing or removing heat from the room (opposite to the action of the temperature control system).

[0003] JP2008 / 089284 describes a simple prior art system comprising a controller that stores the change in room temperature per unit time during air conditioning operation in a room where indoor units of an air conditioner are installed. It then calculates a standard value for room temperature change based on the stored changes per unit time and stops the air conditioner when the changes per unit time do not reach the standard value. However, the system in JP2008 / 089284 has failed to efficiently and accurately detect wasteful operating conditions.

[0004] The present invention aims to solve the above-mentioned problems and disadvantages. Summary of the Invention

[0005] The present invention aims to at least address some of the problems and drawbacks mentioned above.

[0006] Therefore, the present invention relates to a system for temperature control in one or more rooms according to claim 1, and a method for temperature control in one or more rooms according to claim 9. Different embodiments are derived from the dependent claims.

[0007] According to a first aspect, the present invention relates to a system for temperature control in one or more rooms, each of the one or more rooms including at least one temperature control unit, the temperature control unit including one or more heating and / or cooling units, the system including a supervisory controller configured to control the operation of the temperature control unit. The supervisory controller is also configured to estimate the heat variation in each of the one or more rooms.

[0008] The estimated heat change is based on historical outdoor temperature data indicating outdoor temperature over a longer period of time and historical indoor temperature data indicating indoor temperature in each of the one or more rooms over the same longer period of time.

[0009] The supervisory controller is configured to detect the presence of energy waste in at least one of the one or more rooms based on the estimated heat change in each of the one or more rooms and the current indoor temperature of each of the one or more rooms.

[0010] The monitoring controller is configured to take intervention action when an energy waste state is detected in at least one room, preferably wherein the intervention action is carried out in or for the at least one room, more preferably wherein the intervention action includes at least shutting down the temperature control unit in the at least one room where the energy waste state is detected.

[0011] This system provides a functional method for estimating room heat changes based on a combination of historical indoor and outdoor temperature data. By using the historical data, the estimated heat changes can be predicted, aiming to identify anticipated temperature changes in a first room (due to heat loss / gain). Taking into account the current indoor temperature of the room, the supervisory controller can then detect the presence of energy waste in the room by comparing the estimated heat changes with the current indoor temperature.

[0012] The above estimation is performed from a previous time point relative to the current indoor temperature being compared. Thus, the estimated indoor temperature at the current time (the current time, i.e., the time of the latest indoor temperature registered in the system) can be predicted from the previous time point using the estimated heat change from the previous time point to the current time. In this sense, the current indoor temperature should be considered the latest indoor temperature registered in the system. Conversely, the indoor temperatures (both current and previous time points) can be used to determine the actual heat change, which can then be compared with the estimated heat change.

[0013] If the supervisory controller detects a mismatch between the estimated value and the current indoor temperature, and the supervisory controller cannot provide further explanation, then an energy waste state is detected.

[0014] Based on whether a perceived state of energy waste is detected, the supervisory controller can take intervention actions in the room, such as controlling the room's temperature control unit (shutting it off, reducing its output, etc.).

[0015] Existing technologies, such as JP2008 / 089284, fail to disclose a system and associated methods for estimating heat changes using both indoor and outdoor temperature historical data, which is precisely what makes it possible to establish a reliable representation of the heat exchange characteristics of a room (or multiple rooms), thereby allowing for accurate estimation of heat changes.

[0016] In an implementation, the estimated heat change is estimated via a formula or model based on the historical outdoor temperature data and the historical indoor temperature data, wherein the formula or model receives the indoor temperature at a previous time point as input, preferably at least 1 hour ago, more preferably at least 30 minutes ago.

[0017] Building formulas or models based on historical data (indoor and outdoor temperatures) requires a certain number of (relevant and relatively recent) data points in each dataset (indoor and outdoor). Preferably, the indoor and outdoor temperature datasets are coupled such that each data point in the first dataset is linked to a data point in the other dataset, indicating that they were registered at (substantially) the same time. However, even if the data points are not linked, they can still be used to build formulas or models, for example, by extrapolating curves connecting the data points and selecting appropriate time points for using the extrapolated (indoor and / or outdoor) temperatures.

[0018] Such a formula or model can then be used to estimate heat changes over a period of time following the input temperature. This estimation can take the form of an estimated heat change curve over a time period (e.g., 10 minutes, 20 minutes, 30 minutes, 1 hour, 2 hours, etc.) following the input temperature, extending at least to the current time to allow for comparison with the current indoor temperature. Therefore, if the input temperature is from 1 hour ago, the estimated heat change curve should estimate at least to 1 hour from that time, i.e., the current time. If the input temperature is from 30 minutes ago, the estimated heat change curve should estimate at least to 30 minutes from that time, i.e., the current time.

[0019] Alternatively, it can take the form of one or more discrete estimates of heat change at fixed time intervals (e.g., at 10 minutes, 20 minutes, 30 minutes, 40 minutes, etc.) after the time of the input temperature. This is preferably performed at least up to the current time or later. Thus, if the input temperature is from 1 hour ago, the estimated heat change should be estimated at least up to 1 hour from that time, i.e., the current time. If the input temperature used is from 30 minutes ago, the estimated heat change should be estimated at least up to 30 minutes from that time, i.e., the current time.

[0020] Preferably, the indoor temperature used as input is even more recent, and is at most 20 minutes past, more preferably at most 15 minutes past, or even at most 10 minutes or 5 minutes past.

[0021] Using the indoor temperature at the most recent previous time point as input ensures that the formula or model is reliable, as no substantial changes that the formula or model has not yet taken into account should occur (e.g., activation of temperature control units, the effects of weather changes, etc.).

[0022] In a preferred embodiment, the formula or model used to estimate heat changes is based on the following equation: , This indicates the heat capacity of the room. Indicates the indoor temperature of the room. This indicates the current outdoor temperature. This indicates the thermal resistance between the room and the external environment. This refers to the energy transferred to or removed from the room by the one or more temperature control units. Regarding energy, this refers to the energy actually converted into heat, excluding energy used for the general operation of the unit or energy losses due to inefficient conversion of energy used by the unit into heat. Both types of transferred / removed energy can usually be readily derived from the (known) specifications of the temperature control units. and It was derived from historical indoor and outdoor temperature data.

[0023] Using the aforementioned equation allows for a simple yet highly accurate representation of the thermal properties of a room, which can even be easily determined based on a finite sample set of historical indoor and outdoor temperature data. Only two factors need to be determined in the basic equation ( and In some cases, the energy transferred or removed by one or more of the temperature control units is unknown, but this can also be easily inferred based on historical temperature data, especially if further information is available, such as the operating status of the unit at the time of the data point in the historical temperature data.

[0024] In addition, it should be noted that the energy transferred / removed from the room can be in the form of a function or model defined by the characteristics of the temperature control unit (rise time, fall time, maximum output, etc.).

[0025] In a particularly preferred embodiment, the formula or model is developed or created using machine learning by training the model on an available dataset. The use of machine learning is particularly valuable here because of the large amount of data points that can be obtained and stored in historical indoor and outdoor temperature datasets, and used to train the model, making it more accurate and robust, and suitable for working in any relevant new situations.

[0026] As will be discussed further in the specification, the machine learning model can be further extended by training it on other data that the machine learning model considers.

[0027] In another preferred embodiment, the formula or model receives the outdoor temperature at a previous time point as further input, preferably at least one hour ago, more preferably at least 30 minutes ago, and preferably at the same or close time point as the input indoor temperature. As mentioned regarding the current indoor temperature, the outdoor temperature preferably used as input is more recent, and at least 20 minutes ago, more preferably at least 15 minutes ago, or even at most 10 minutes or 5 minutes ago, for the same reasons as above. In that case, the substantial change is more related to sudden weather changes, but could also be due to human influence.

[0028] To ensure the reliability of the formula or model, the system needs sufficient data points to model the room’s thermal characteristics (e.g., natural heat exchange with the environment, the effect of temperature difference with the environment, the effect of temperature control units, the effect of the presence of users, and the effect of the presence / operation of objects and / or other features).

[0029] Thus, the supervisory controller may include or be linked to a dedicated acquisition section to acquire indoor temperature data via one or more indoor temperature sensors. These indoor temperature sensors may be integrated into a local component of the supervisory controller, and / or a temperature control unit, and / or via separate sensors from which indoor temperature data can be collected. In the case of multiple indoor temperature sensors, an averaging step can be performed to determine individual values ​​(arithmetic mean, geometric mean, harmonic mean, median, mode, middle value, and other types of averages, optionally with a filtering step to remove discrepancies).

[0030] In the same manner, the acquisition section acquires outdoor temperature data from an outdoor temperature sensor and / or at least one external data source. As previously described, the outdoor temperature sensor may be integrated into the supervisory controller (an external component), and / or the temperature control unit (an external component), and / or acquired via the outdoor sensor, from which outdoor temperature data can be collected. External data sources may be, for example, temperature applications, Internet resources, etc.

[0031] As mentioned earlier, in the case of multiple outdoor temperature sensors and / or external data sources, an averaging step can be performed to determine a single value.

[0032] (Indoor and / or outdoor) temperature data are typically collected periodically. The time intervals between collections can vary between 1 second, 10 seconds, 30 seconds, 1 minute, 5 minutes, 30 minutes, 1 hour, 2 hours, 3 hours, or even 6 hours, 12 hours, etc. Preferably, the time intervals are kept relatively short to accommodate sudden temperature changes and day-night cycles without unnecessarily storing too much data.

[0033] Temperature data is typically collected at the same time, with indoor and outdoor temperature data being coupled in this way.

[0034] These collected indoor and outdoor temperature data points are stored in the storage segment of the supervisory controller to create indoor and outdoor temperature historical data sets. The indoor and outdoor temperature historical data preferably only considers data points from previous time periods of a predetermined length, in order to ignore data points from much earlier periods that may no longer be relevant, such as due to changes in infrastructure (insulation), environmental factors, and / or other factors.

[0035] To ensure the accuracy of the formula or model, it is periodically updated and / or regenerated, preferably at fixed intervals, to accommodate changes in the environment, the room, room-related characteristics, etc. The regeneration of the formula or model can be performed, for example, every minute, every 2 minutes, every 5 minutes, every 10 minutes, every 20 minutes, every 30 minutes, every hour, every 2 hours, every 3 hours, every 6 hours, or longer. The formula or model is then typically regenerated based on a more recent subset of historical indoor and outdoor temperature data, which may partially overlap with or may not overlap with the subset used in previously generated formulas or models.

[0036] In another preferred embodiment, the formula or model is also based on historical operating data of the temperature control units, which defines the operating state of each temperature control unit over the longer time period. One of the factors that strongly influences heat variations in a room is the activity of temperature control units (heating and cooling units, or a combination of both). This system and the associated methodology are well-suited for developing models or formulas for estimating natural heat variations (i.e., without any artificial or sudden heat input or output, such as from temperature control units). This can be easily performed if sufficient historical temperature data is available.

[0037] However, while the impact of temperature control units on room heat is technically easy to quantify, it is difficult to predict because their activation, deactivation, and intermediate modes can vary depending on the presence of the occupant, their preferences, and even their mood or physical condition, and differ based on the time of day, day of week, or time of year. Thus, accurate knowledge of their operational status over longer periods is crucial for supplementing natural heat variation models / formulas and deriving more accurate estimates of heat variations.

[0038] In some implementations, the operating history data of the temperature control unit does not necessarily include the operating status of each temperature control unit over the entire longer period of time, but rather the most recent portion extending to the present time, because once the dynamics of the operating status affecting heat changes can be determined (which is usually quite simple), it can be used as a supplement to the model / formula.

[0039] As mentioned earlier, using machine learning can significantly improve reliability in identifying states of energy waste. By training a model with historical data from temperature control unit operations, heat gain or loss as a result of temperature control unit operations can be detected and taken into account in subsequent steps.

[0040] In another preferred embodiment, the supervisory controller is connected to each temperature control unit and includes an identification segment configured to detect the operation of each temperature control unit. Through this direct connection, the supervisory controller can accurately and reliably acquire data at periodic intervals.

[0041] In a preferred embodiment, at least one temperature control unit is a heat medium circulation system comprising at least one radiator and a heat source unit, and further includes a valve for shutting off the circulation of the heat medium in the system, wherein the supervisory controller is configured to control the valve to control the operation of the heat medium circulation system. One of the main applications of the system and methodology of the present invention is to take action on energy waste states, where the temperature control unit is used to heat via the radiator through heat medium circulation. In this case, an energy waste state is detected, and based on the current indoor temperature (and potentially other characteristics, such as the current outdoor temperature and historical operating data of the temperature control unit, ...), the estimated heat change represents a smaller heat loss or a larger heat gain (relative to a certain tolerance) than actually measured. If there is no cause for such a difference, the present invention will consider the existence of unexpected heat loss, such as due to an open window, etc., and will take action accordingly via an intervention action. In particular, one such action would involve shutting off heating in the room where the energy waste state is detected, which can be performed by simply operating the valve to (completely or partially) stop the circulation of the heat medium to the radiator in the room where the energy waste state is detected. In this way, the heat medium circulation can still reach other rooms where the energy waste state is not determined.

[0042] In a preferred embodiment, at least one temperature control unit is an air conditioning system comprising at least one compressor and at least one air conditioning unit, wherein the supervisory controller is configured to control the compressor to control the operation of the air conditioning system.

[0043] One of the main applications of the system and methodology of this invention is to take action on energy waste states, where a temperature control unit is used for cooling via a compressor and an air conditioning unit. In this case, an energy waste state is detected, and based on the current indoor temperature (and potentially other characteristics, such as the current outdoor temperature and historical operating data of the temperature control unit, ...), the estimated heat change represents a greater heat loss or a smaller heat gain (relative to a certain tolerance) than actually measured. If there is no cause for such a difference, the invention will consider the presence of unexpected heat input, such as due to an open window, and will take action accordingly via an intervention. In particular, one such action would involve closing (or reducing) the cooling in the room where the energy waste state is detected.

[0044] As previously stated, many interventions are possible, and each intervention mentioned in the application can be interchanged with other interventions specifically mentioned in any implementation, unless specifically impossible, where such implementations are implicitly disclosed and are part of this specification.

[0045] In a preferred embodiment, the supervisory controller is configured to estimate the expected indoor temperature of each of the one or more rooms, the expected indoor temperature of each room being determined by an estimated heat change in each room and the indoor temperature of each room at a previous time point. Preferably, the previous time point is at most one hour ago, more preferably at most 30 minutes ago. The existence of energy waste in a room is determined by comparing the expected indoor temperature of the room with the current indoor temperature.

[0046] As mentioned earlier, using more recent previous time points is preferred because it ensures reliability. While models and formulas may be reliable, as with all estimates, projections, and forecasts, the degree to which they may deviate from actual values ​​increases as the time point from which the estimate was made becomes more recent than previous time points. A more recent baseline reduces the amount of deviation that the results may have.

[0047] Thus, the previous time point is preferably at most 20 minutes past, more preferably at most 15 minutes past, or even at most 10 minutes or 5 minutes past.

[0048] In a preferred embodiment, if the indoor temperature of one of the one or more rooms is lower than the setpoint temperature of the one or more rooms by more than 0.5°C for at least 5 minutes, the controller confirms the existence of an energy waste state.

[0049] In an alternative embodiment, preferably in combination with the above embodiment, if the indoor temperature of one of the one or more rooms is higher than the setpoint temperature of the one or more rooms by more than 1.0°C for at least 5 minutes, the controller confirms the existence of an energy waste state.

[0050] In this embodiment, a rule-based approach is used to confirm the existence of energy waste states, with fixed thresholds for both minimum difference and minimum duration of that difference. This allows for a simple and straightforward way to determine the situation. However, it should be noted that the above rules can be complementary to other rules (dynamic or fixed).

[0051] Preferably, the minimum temperature difference between the room and the set point temperature is at least 0.75°C, more preferably at least 1.00°C, more preferably at least 1.25°C, more preferably at least 1.50°C, more preferably at least 1.75°C, even more preferably at least 2.00°C, even more preferably at least 2.25°C, even more preferably at least 2.50°C, even more preferably at least 2.75°C, even more preferably at least 3.00°C, even more preferably at least 3.50°C, even more preferably at least 4.00°C, even more preferably at least 4.50°C, even more preferably at least 5.00°C.

[0052] Preferably, the shortest time period during which the threshold is crossed is at least 6 minutes, preferably at least 7 minutes, preferably at least 8 minutes, preferably at least 9 minutes, preferably at least 10 minutes, more preferably at least 12 minutes, more preferably at least 14 minutes, more preferably at least 16 minutes, more preferably at least 18 minutes, more preferably at least 20 minutes, even more preferably at least 25 minutes, even more preferably at least 30 minutes, even more preferably at least 40 minutes, 50 minutes or 60 minutes.

[0053] In a preferred embodiment, the supervisory controller is configured to determine the actual heat change in at least one of the one or more rooms. This actual heat change is based on historical indoor temperature data, optionally on historical outdoor temperature data, and also on the current indoor temperature in the at least one of the one or more rooms. The actual heat change can be determined by using previous historical indoor temperature data (points) and the current indoor temperature. Using historical outdoor temperature data can improve its accuracy.

[0054] In a preferred embodiment, an energy waste state is detected if the actual heat change in a room exceeds or falls below the estimated heat change in the room by at least a predefined absolute or relative margin, preferably for at least a predefined shortest period of time.

[0055] Preferably, the presence of a sensor is detected only if the actual heat change exceeds or falls below the estimated heat change for a minimum time period (in the sense of confirming and taking action). This minimum time period is preferably as previously defined, i.e., at least 5 minutes, at least 6 minutes, preferably at least 7 minutes, preferably at least 8 minutes, preferably at least 9 minutes, preferably at least 10 minutes, more preferably at least 12 minutes, more preferably at least 14 minutes, more preferably at least 16 minutes, more preferably at least 18 minutes, more preferably at least 20 minutes, even more preferably at least 25 minutes, even more preferably at least 30 minutes, even more preferably at least 40 minutes, 50 minutes, or 60 minutes. This avoids false alarms, where sensor malfunctions, brief sudden temperature increases (e.g., human touch of the sensor), or similar temporary differences can be filtered out without intervention.

[0056] The predefined relative margin is preferably at least 1% (temperature in degrees Celsius), more preferably at least 1.5%, more preferably at least 2.0%, even more preferably at least 2.5%, even more preferably at least 3.0%, even more preferably at least 3.5%, even more preferably at least 4.0%, even more preferably at least 4.5%, even more preferably at least 5.0%, even more preferably at least 5.5%, even more preferably at least 6.0%, even more preferably at least 6.5%, even more preferably at least 7.0%, even more preferably at least 7.5%, even more preferably at least 8.0%, even more preferably at least 8.5%, even more preferably at least 9.0%, even more preferably at least 9.5%, even more preferably at least 10.0%, or even at least 11%, 12%, 13%, 14%, 15% or more, for example at least 17.5%, 20%, 22.5%, 25% or 30%.

[0057] In a preferred embodiment, the supervisory controller is configured to determine the actual heat change in each of the one or more rooms. The actual heat change is based on historical indoor temperature data, optionally historical outdoor temperature data, and also on the current indoor temperature in each of the one or more rooms.

[0058] In most cases, systems and methodologies will be applied to multi-room buildings to maximize their potential. However, this adds complexity, as heat variations are not always due to heat exchange with the external environment (for which outdoor temperature is used to determine heat exchange, as temperature difference strongly defines the conduction process). In multi-room configurations, heat variations may result from heat exchange with adjacent rooms, where heat loss may occur in some exchanges with other rooms, while heat gain may occur in others.

[0059] This effect can be considered by providing this data to the system (and the associated methodology). To consider this effect, it is not necessary to precisely model the exact physical processes (as that would require extensive knowledge, such as shared wall area, thickness, materials, etc.), but if the dataset is provided with sufficient data points (historical indoor temperature data for each room, and historical outdoor temperature data), formulas or models can infer the impact of different indoor temperatures from other rooms. This means that configuration (knowledge about which rooms are adjacent to which) does not necessarily need to be known or required as input, as the lack of influence will also be reflected in the historical data.

[0060] In another preferred embodiment for determining the actual heat change for each room, an energy waste state is detected if the actual heat change of one of the rooms exceeds or falls below at least a predefined absolute or relative margin for the estimated heat change of that room, and preferably continues for a predefined shortest period of time (as described above).

[0061] In a preferred embodiment, a supervisory controller is responsible for temperature control in multiple rooms, wherein the supervisory controller is configured to consider heat changes in each of the rooms when it detects the presence of energy waste, in order to identify heat transfer between two or more of the multiple rooms.

[0062] Allowing the monitoring controller to consider internal heat transfer between two or more rooms can further help clarify whether an actual state of energy waste is occurring. By being able to distinguish this internal heat transfer from actual heat loss to the outside or heat gain from the outside, the system and methodology allow for action to be taken only when necessary, rather than based on the perspective of isolated rooms.

[0063] However, this does not mean that heat transfer between rooms is always completely (or partially) ignored when detecting energy waste. In some cases, it can be ignored, such as when some rooms gain or lose more heat than others (e.g., if only some rooms have heating and / or cooling units while others do not), and heat transfer is actually designed to compensate for this and reach equilibrium. In such cases, only the heat changes that exchange heat with external locations outside these rooms are considered to determine the presence of energy waste and whether intervention is necessary.

[0064] In some implementations, heat transfer between rooms is also considered part of the room’s heat change, and the state of energy waste can be determined based on the room’s net heat change.

[0065] In some implementations, heat transfer with another room and heat changes in that room are considered separately, wherein such separate heat transfer with other rooms may lead to the determination of an energy waste state (even without indication of abnormal heat changes relative to the outside), thereby leading to intervention actions in the room itself and / or in the room where heat transfer leading to the determination of an energy waste state occurs.

[0066] As mentioned earlier, using machine learning can significantly improve reliability in determining the existence of energy waste. By training a model with a dataset for each room, heat transfer between rooms can be detected and taken into account in subsequent steps.

[0067] In a preferred embodiment, the intervention action is one or more of the following: alerting a human operator, disabling one or more temperature control units (preferably at least in the room where energy waste is detected), reducing the temperature control operation of one or more temperature control units, or suggesting to a human operator to disable or reduce the temperature control operation of one or more temperature control units for approval.

[0068] The aforementioned interventions can be linked together as follow-up actions in a specific order. Each follow-up intervention is then taken based on further circumstances, such as an increase in the difference between actual and estimated heat changes over a specific time period, lack of convergence, or slow convergence. In some actions, approval from a human operator is required; in such cases, follow-up actions can also be taken if the operator does not respond within a certain time period.

[0069] Further intervention could involve controlling certain home appliances or smart home devices, such as automatically closing windows and doors, or closing blinds or sunshades.

[0070] In a preferred embodiment, historical indoor and outdoor temperature data are collected over a single measurement period. Using a single measurement period ensures that the data points are connected and show the temperature variation over time, which is highly relevant for constructing models / formulas to estimate changes in heat over time. In this sense, it is important that the data is registered over a sufficiently long measurement period to ensure that there are enough data points and sufficient variations to allow for the construction of models / formulas.

[0071] In an alternative preferred embodiment, historical indoor temperature data and / or historical outdoor temperature data are collected over multiple time-separated measurement periods. It should be noted that the dataset therefore includes subsets for individual measurement periods, where each subset will still be a collection of multiple data points over time within said individual measurement period in order to determine the dynamics of the room temperature.

[0072] Using multiple measurement cycles offers the following advantages: the conditions within each measurement cycle may differ (different outdoor temperatures, operating heating units, room occupancy, abundant outdoor sunlight, etc.), thus providing a broader perspective on temperature dynamics and the impact of these different conditions. This allows the constructed models or formulas to handle estimated heat changes under different conditions with greater reliability (not only from different conditions in different measurement cycles, but also by handling many variations through extrapolation and interpolation).

[0073] In this implementation, historical indoor temperature data and / or historical outdoor temperature data are periodically updated or replaced. Preferably, both are updated or replaced consistently with each other, as ideally these are linked data points (indoor and outdoor temperatures at essentially the same time). Consistent updates or replacements mean that linked data points are updated / replaced together, or removed. Periodically updating, replacing, and / or removing data points ensures that the retained data points are relevant to the situation, and that old, outdated data points no longer contaminate the dataset, for example, in cases where structural changes modify rooms to the point that the old data is no longer representative.

[0074] In other implementations, historical indoor and / or outdoor temperature data are retained, ensuring that the estimated heat variation is based on a continuously growing dataset. This guarantees that the estimation is performed with a detailed understanding of various scenarios, thus ensuring accurate estimation.

[0075] Having a large dataset representing a high degree of diversity is particularly advantageous when training models using machine learning. This is often the result of collecting data points over a long period (as long as possible). While data can be pruned periodically, this does not necessarily need to be performed as a standard operation and can be the result of, for example, human action (operator "resetting" the system; automated reset based on an anomalous difference between estimated and actual heat changes; system prompting operator to initiate a reset based on such an anomalous difference).

[0076] In a preferred embodiment, the supervisory controller detects the presence of energy waste based on an error function defined by the following formula: In this formula, It refers to the actual energy changes in each room. It is the estimated energy change in each room, further taking into account the heat generated from the temperature control unit in each room, and wherein the cumulative sum control chart (CUSUM) value based on the error function determines the existence of an energy waste state in the room, wherein the existence of an energy waste state is confirmed if the CUSUM value exceeds a predetermined upper limit or if a predetermined lower limit exceeds the CUSUM value.

[0077] Most preferably, the actual and estimated energy changes ( , ) is the actual and estimated heat change ( , ).

[0078] Using the cumulative sum of the differences between estimated and actual values ​​ensures that temporary fluctuations do not trigger interventions (e.g., hot or cold objects pressing on temperature sensors, airflow obstruction, etc.). Instead, such temporary fluctuations may flatten out in the cumulative sum (the opposite fluctuation partially cancels out the original fluctuation), or they will be placed in the context of the cumulative sum, where a single high deviation results in a small or medium average deviation over the entire cumulative sum.

[0079] According to a second aspect, the present invention relates to a method for temperature control in one or more rooms, each of the one or more rooms including at least one temperature control unit, the method comprising the following steps: a. Estimate the heat change in each of the one or more rooms, wherein the estimated heat change is based on historical outdoor temperature data indicating outdoor temperature over a longer period of time and historical indoor temperature data indicating indoor temperature in each of the one or more rooms over the longer period of time. b. Detect the presence of an energy waste state in at least one of the one or more rooms based on the estimated heat change in each of the one or more rooms and the current indoor temperature in each of the one or more rooms; c. Taking an intervention action when an energy waste state is detected in at least one room, preferably wherein the intervention action is carried out in or for the at least one room, more preferably wherein the intervention action includes at least shutting down the temperature control unit in the at least one room where the energy waste state is detected.

[0080] The advantages mentioned regarding the system according to the invention also apply to the methods described above and subsequent embodiments. Furthermore, it should be noted that any embodiment mentioned regarding the system can also be converted into an embodiment of the method and should be considered as implicitly disclosed herein.

[0081] In a preferred embodiment, the step of estimating heat changes is performed via a formula or model based on the historical outdoor temperature data and the historical indoor temperature data, wherein the formula or model receives the indoor temperature at a previous time point as input. Preferably, the previous time point is at most 1 hour ago, more preferably at most 30 minutes ago.

[0082] Preferably, the indoor temperature used as input is even more recent, and is at most 20 minutes past, more preferably at most 15 minutes past, or even at most 10 minutes or 5 minutes past.

[0083] In a particularly preferred embodiment, the formula or model is developed or created using machine learning by training the model on an available dataset. The use of machine learning is particularly valuable here because of the large amount of data points that can be obtained and stored in historical indoor and outdoor temperature datasets, and used to train the model, making it more accurate and robust, and suitable for working in any relevant new situations.

[0084] As will be discussed further in the specification, the machine learning model can be further extended by training it on other data that the machine learning model considers.

[0085] In a preferred embodiment, the formula or model receives the outdoor temperature at a previous time point as further input. Preferably, this time point is at most 1 hour ago, more preferably at most 30 minutes ago. Preferably, the previous time point is the same as or close to the previous time point of the indoor temperature input above. As mentioned regarding the current indoor temperature, the outdoor temperature used as input is preferably more recent, and is at most 20 minutes ago, more preferably at most 15 minutes ago, or even at most 10 minutes or 5 minutes ago, for the same reasons as above. In that case, the substantial change is more related to sudden weather changes, but it may also be due to human influence.

[0086] To ensure the accuracy of the formula or model, it is periodically updated and / or regenerated, preferably at fixed intervals, to accommodate changes in the environment, the room, room-related characteristics, etc. The regeneration of the formula or model can be performed, for example, every minute, every 2 minutes, every 5 minutes, every 10 minutes, every 20 minutes, every 30 minutes, every hour, every 2 hours, every 3 hours, every 6 hours, or longer. The formula or model is then typically regenerated based on a more recent subset of historical indoor and outdoor temperature data, which may partially overlap with or may not overlap with the subset used in previously generated formulas or models.

[0087] In a preferred embodiment, the formula or model is also based on historical operating data of the temperature control units, which defines the operating status of each temperature control unit over the longer time period.

[0088] As mentioned earlier, using machine learning can significantly improve reliability in identifying states of energy waste. By training a model with historical data from temperature control unit operations, heat gain or loss as a result of temperature control unit operations can be detected and taken into account in subsequent steps.

[0089] In a preferred embodiment, at least one temperature control unit is a heat medium circulation system comprising at least one radiator and a heat source unit, and further includes a valve for shutting off heat medium circulation in the heat medium circulation system, wherein the valve is controlled based on the detection of an energy waste state in the room containing the radiator, thereby controlling the operation of the heat medium circulation system. One of the main applications of the system and methodology of the present invention is to take action on an energy waste state, wherein the temperature control unit is used to heat via the radiator through heat medium circulation. In this case, an energy waste state is detected, and based on the current indoor temperature (and potentially other characteristics, such as the current outdoor temperature and historical operating data of the temperature control unit, ...), the estimated heat change represents a smaller heat loss or a larger heat gain (relative to a certain tolerance) than actually measured. If there is no cause for such a difference, the present invention will consider the existence of unexpected heat loss, such as due to an open window, etc., and will take action accordingly via an intervention action. In particular, one such action would involve shutting off heating in the room where the energy waste state is detected, which can be performed by simply operating the valve to (completely or partially) stop the circulation of the heat medium to the radiator in the room where the energy waste state is detected. In this way, heat medium circulation can still reach other rooms where no energy waste state has been determined.

[0090] In a preferred embodiment, at least one temperature control unit is an air conditioning system comprising at least one compressor and at least one air conditioning unit, wherein the compressor is controlled based on energy waste detected in the room of the air conditioning unit in order to control the operation of the air conditioning system.

[0091] One of the main applications of the system and methodology of this invention is to take action on energy waste states, where a temperature control unit is used for cooling via a compressor and an air conditioning unit. In this case, an energy waste state is detected, and based on the current indoor temperature (and potentially other characteristics, such as the current outdoor temperature and historical operating data of the temperature control unit, ...), the estimated heat change represents a greater heat loss or a smaller heat gain (relative to a certain tolerance) than actually measured. If there is no cause for such a difference, the invention will consider the presence of unexpected heat input, such as due to an open window, and will take action accordingly via an intervention. In particular, one such action would involve closing (or reducing) the cooling in the room where the energy waste state is detected.

[0092] In a preferred embodiment, the method includes the steps of: estimating the expected indoor temperature of each of the one or more rooms, the expected indoor temperature of each room being determined by an estimated heat change in each room and the indoor temperature of each room at a previous time point. Preferably, the time point is at most one hour past, more preferably at most 30 minutes past. The existence of an energy waste state in a room is determined by comparing the expected indoor temperature of the room with the current indoor temperature. If the current indoor temperature of a room exceeds or falls below at least a predefined absolute or relative margin of the room's expected indoor temperature, preferably at least for a predefined minimum time period (discussed previously), the existence of an energy waste state is detected.

[0093] The predefined absolute margin is preferably at least 0.5°C, preferably at least 0.75°C, more preferably at least 1.00°C, more preferably at least 1.25°C, more preferably at least 1.50°C, more preferably at least 1.75°C, even more preferably at least 2.00°C, even more preferably at least 2.25°C, even more preferably at least 2.50°C, even more preferably at least 2.75°C, even more preferably at least 3.00°C, even more preferably at least 3.50°C, even more preferably at least 4.00°C, even more preferably at least 4.50°C, even more preferably at least 5.00°C.

[0094] The predefined relative margin is preferably at least 1% (temperature in degrees Celsius), more preferably at least 1.5%, more preferably at least 2.0%, even more preferably at least 2.5%, even more preferably at least 3.0%, even more preferably at least 3.5%, even more preferably at least 4.0%, even more preferably at least 4.5%, even more preferably at least 5.0%, even more preferably at least 5.5%, even more preferably at least 6.0%, even more preferably at least 6.5%, even more preferably at least 7.0%, even more preferably at least 7.5%, even more preferably at least 8.0%, even more preferably at least 8.5%, even more preferably at least 9.0%, even more preferably at least 9.5%, even more preferably at least 10.0%, or even at least 11%, 12%, 13%, 14%, 15% or more, for example at least 17.5%, 20%, 22.5%, 25% or 30%.

[0095] As mentioned earlier, using more recent previous time points is preferred because it ensures reliability. While models and formulas may be reliable, as with all estimates, projections, and forecasts, the degree to which they may deviate from actual values ​​increases as the time point from which the estimate was made becomes more recent than previous time points. A more recent baseline reduces the amount of deviation that the results may have.

[0096] Thus, the previous time point is preferably at most 20 minutes past, more preferably at most 15 minutes past, or even at most 10 minutes or 5 minutes past.

[0097] In a preferred embodiment, if the indoor temperature of one of the one or more rooms is lower than the setpoint temperature of that one of the one or more rooms by more than 0.5°C for at least 5 minutes, the existence of an energy waste state is confirmed.

[0098] In an alternative embodiment, preferably in combination with the above embodiment, if the indoor temperature of one of the one or more rooms is higher than the setpoint temperature of the one or more rooms by more than 1.0°C for at least 5 minutes, the existence of energy waste is confirmed.

[0099] In this embodiment, a rule-based approach is used to confirm the existence of energy waste states, with fixed thresholds for both minimum difference and minimum duration of that difference. This allows for a simple and straightforward way to determine the situation. However, it should be noted that the above rules can be complementary to other rules (dynamic or fixed).

[0100] Preferably, the minimum temperature difference between the room and the set point temperature is at least 0.75°C, more preferably at least 1.00°C, more preferably at least 1.25°C, more preferably at least 1.50°C, more preferably at least 1.75°C, even more preferably at least 2.00°C, even more preferably at least 2.25°C, even more preferably at least 2.50°C, even more preferably at least 2.75°C, even more preferably at least 3.00°C, even more preferably at least 3.50°C, even more preferably at least 4.00°C, even more preferably at least 4.50°C, even more preferably at least 5.00°C.

[0101] Preferably, the shortest time period during which the threshold is crossed is at least 6 minutes, preferably at least 7 minutes, preferably at least 8 minutes, preferably at least 9 minutes, preferably at least 10 minutes, more preferably at least 12 minutes, more preferably at least 14 minutes, more preferably at least 16 minutes, more preferably at least 18 minutes, more preferably at least 20 minutes, even more preferably at least 25 minutes, even more preferably at least 30 minutes, even more preferably at least 40 minutes, 50 minutes or 60 minutes.

[0102] In a preferred embodiment, the method includes the step of determining the actual heat change in at least one of the one or more rooms. This actual heat change is based on historical indoor temperature data, optionally on historical outdoor temperature data, and also on the current indoor temperature in the at least one of the one or more rooms. The actual heat change can be determined by using previous historical indoor temperature data (points) and the current indoor temperature. Using historical outdoor temperature data can improve its accuracy.

[0103] In a preferred embodiment, an energy waste state is detected if the actual heat change in a room exceeds or falls below at least a predefined absolute or relative margin (as previously discussed) of the estimated heat change in the room.

[0104] As previously mentioned, it is preferred that presence is detected only when the actual heat change in the room exceeds or falls below the estimated heat change for at least a predefined minimum period of time (as previously discussed).

[0105] In a preferred embodiment, the method includes the step of determining the actual heat change in each of the one or more rooms. The actual heat change is based on historical indoor temperature data, optionally historical outdoor temperature data, and also on the current indoor temperature in each of the one or more rooms.

[0106] In most cases, systems and methodologies will be applied to multi-room buildings to maximize their potential. However, this adds complexity, as heat variations are not always due to heat exchange with the external environment (for which outdoor temperature is used to determine heat exchange, as temperature difference strongly defines the conduction process). In multi-room configurations, heat variations may result from heat exchange with adjacent rooms, where heat loss may occur in some exchanges with other rooms, while heat gain may occur in others.

[0107] This effect can be considered by providing this data to the system (and the associated methodology). To consider this effect, it is not necessary to precisely model the exact physical processes (as that would require extensive knowledge, such as shared wall area, thickness, materials, etc.), but if the dataset is provided with sufficient data points (historical indoor temperature data for each room, preferably along with historical outdoor temperature data), formulas or models can infer the impact of different indoor temperatures from other rooms. This means that configuration (knowledge about which room is adjacent to which) does not necessarily need to be known or required as input, as the lack of influence will also be reflected in the historical data.

[0108] In a preferred embodiment, the intervention action is one or more of the following: alerting a human operator, disabling one or more temperature control units (preferably at least in rooms where energy waste is detected), reducing the heating operation of one or more temperature control units, or suggesting to a human operator to disable or reduce the temperature control operation of one or more temperature control units for approval.

[0109] The aforementioned interventions can be linked together as follow-up actions in a specific order. Each follow-up intervention is then taken based on further circumstances, such as an increase in the difference between actual and estimated heat changes over a specific time period, lack of convergence, or slow convergence. In some actions, approval from a human operator is required; in such cases, follow-up actions can also be taken if the operator does not respond within a certain time period.

[0110] Further intervention could involve controlling certain home appliances or smart home devices, such as automatically closing windows and doors, or closing blinds or sunshades.

[0111] In a preferred embodiment, historical indoor and outdoor temperature data are collected over a single measurement period. Using a single measurement period ensures that the data points are connected and show the temperature variation over time, which is highly relevant for constructing models / formulas to estimate changes in heat over time. In this sense, it is important that the data is registered over a sufficiently long measurement period to ensure that there are enough data points and sufficient variations to allow for the construction of models / formulas.

[0112] In an alternative preferred embodiment, historical indoor temperature data and / or historical outdoor temperature data are collected over multiple time-separated measurement periods. It should be noted that the dataset therefore includes subsets for individual measurement periods, where each subset will still be a collection of multiple data points over time within said individual measurement period in order to determine the dynamics of the room temperature.

[0113] In this implementation, historical indoor temperature data and / or historical outdoor temperature data are periodically updated or replaced. Preferably, both are updated or replaced consistently with each other, as ideally these are linked data points (indoor and outdoor temperatures at essentially the same time). Consistent updates or replacements mean that linked data points are updated / replaced together, or removed. Periodically updating, replacing, and / or removing data points ensures that the retained data points are relevant to the situation, and that old, outdated data points no longer contaminate the dataset, for example, in cases where structural changes modify rooms to the point that the old data is no longer representative.

[0114] In other implementations, historical indoor and / or outdoor temperature data are retained, ensuring that the estimated heat variation is based on a continuously growing dataset. This guarantees that the estimation is performed with a detailed understanding of various scenarios, thus ensuring accurate estimation.

[0115] Having a large dataset representing a high degree of diversity is particularly advantageous when training models using machine learning. This is often the result of collecting data points over a long period (as long as possible). While data can be pruned periodically, this does not necessarily need to be performed as a standard operation and can be the result of, for example, human action (operator "resetting" the system; automated reset based on an anomalous difference between estimated and actual heat changes; system prompting operator to initiate a reset based on such an anomalous difference).

[0116] In a preferred embodiment, the presence of energy waste is detected based on an error function defined by the following formula: In this formula, It refers to the actual energy changes in each room. This is the estimated energy change in each room. The formula further considers the heat generated from the temperature control unit in each room. The method also includes the step of determining the existence of an energy waste state in a room based on the cumulative sum control chart (CUSUM) value of an error function, wherein the existence of an energy waste state is confirmed if the CUSUM value exceeds a predetermined upper limit or if a predetermined lower limit is exceeded.

[0117] Using the cumulative sum of the differences between estimated and actual values ​​ensures that temporary fluctuations do not trigger interventions (e.g., hot or cold objects pressing on temperature sensors, airflow obstruction, etc.). Instead, such temporary fluctuations may flatten out in the cumulative sum (the opposite fluctuation partially cancels out the original fluctuation), or they will be placed in the context of the cumulative sum, where a single high deviation results in a small or medium average deviation over the entire cumulative sum. Attached Figure Description

[0118] Figure 1 A flowchart illustrating an embodiment of the present invention is shown, illustrating the use of the CUSUM method to detect energy waste status via an application of available data.

[0119] Figure 2A and Figure 2B Methods for training and deploying a machine learning model (in this case, a neural network) for detecting wasted energy states according to embodiments of the present invention are shown. Detailed Implementation

[0120] The invention is further described by way of the following non-limiting examples, which are intended to further illustrate the invention and are not intended to, nor should they be construed as, limiting the scope of the invention.

[0121] Unless otherwise defined, all terms used in this disclosure, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art as described herein. Further guidance includes terminology definitions to better understand the teachings of this invention.

[0122] As used herein, the following terms have the following meanings: As used herein, "temperature control unit" refers to one or more units (of the same type, operating in conjunction) suitable for heating and / or cooling a room or location. The methods of cooling and heating can vary and can take the form of introducing heated or cooled air (or other media) via thermal radiation (radiators / convection devices).

[0123] As used herein, "room" refers to a largely spatially separated volume that is substantially sealed off from its surroundings by insulated physical barriers (such as walls, windows, doors, etc.). The qualifiers "first" or "second" or any other word are used only to distinguish rooms, but any statements or implementations describing a first room apply equally to any other room.

[0124] The "historical indoor / outdoor temperature data" used in this article refers to a dataset comprising multiple data points representing past indoor / outdoor temperature measurements and associated time periods (relative or absolute). These measurements are preferably obtained periodically at fixed intervals, but this is not necessarily required. Data points can be processed to remove outliers, which can be rule-based. The dataset can be a rolling dataset, discarding the oldest data points as new data points are added, or it can cover a fixed time period (e.g., all data points from the past X hours). Most preferably, the historical indoor and outdoor temperature data are linked, with each data point in one set linked to a data point in the other set, and linked to an associated time period (relative or absolute).

[0125] "Data" or "dataset" as a shorthand refers to one of the datasets mentioned, or may refer to a combination of multiple said datasets, as will be clear in the context in which it is used.

[0126] As used in this article, “one” and “the” refer to the singular and plural referents, respectively, unless the context clearly specifies otherwise. As an example, “a compartment” refers to one or more compartments.

[0127] As used herein, "comprising" and "consisting of" are synonymous with "including" or "containing" and are inclusive or open-ended terms that specify the presence of subsequent content (e.g., components) and do not exclude or prevent the presence of additional, unlisted components, features, elements, components, or steps known in the art or disclosed herein.

[0128] Furthermore, the terms first, second, third, etc., used in the specification and claims are used to distinguish similar elements and are not necessarily used to describe order or chronological order, unless otherwise stated. It should be understood that such terms are interchangeable where appropriate, and the embodiments of the invention described herein can operate in a different order than those described or shown herein.

[0129] The range of values ​​listed by endpoints includes all numbers and fractions that fall within that range, as well as the listed endpoints.

[0130] The terms "one or more" or "at least one" (e.g., one or more or at least one of a set of components) are self-evident; however, by further example, the term specifically covers a reference to any one of the components, or a reference to any two or more of the components, such as any ≥3, ≥4, ≥5, ≥6 or ≥7 of the components, and up to all of the components.

[0131] Unless otherwise defined, all terms used in this disclosure, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. Definitions of terms used in this specification are included as further guidance to better understand the teachings of this invention. The terms or definitions used herein are for illustrative purposes only and are intended to aid in understanding the invention.

[0132] Throughout this specification, the reference to "one embodiment" or "implementation" means that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of the invention. Therefore, the phrases "in one embodiment" or "in an embodiment" appearing throughout this specification do not necessarily refer to the same embodiment, but may refer to the same embodiment. Furthermore, particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments, as will be apparent to those skilled in the art based on this disclosure. Moreover, while some embodiments described herein include some features included in other embodiments but not others, combinations of features from different embodiments are intended to fall within the scope of the invention and form different embodiments, as will be understood by those skilled in the art. For example, in the appended claims, any claimed embodiment may be used in any combination.

[0133] The supervisory controller may include one or more processing units or modules, such as a central processing unit (CPU), a microprocessor, or a suitably programmed field-programmable gate array (FPGA) or application-specific integrated circuit (ASIC). Additionally or alternatively, the supervisory controller may include any memory segments necessary to perform its functions of controlling the operation of the heat pump system. Such memory segments may be part of (e.g., integrally formed or located on the same chip) the supervisory controller, or provided separately but electrically connected to the supervisory controller. As an example, memory segments may include volatile and non-volatile memory resources, including, for example, working memory (e.g., random access memory). Additionally, memory segments may include an instruction storage section (e.g., ROM in the form of electrically erasable programmable read-only memory (EEPROM) or flash memory) storing a computer program containing computer-readable instructions that, when executed by the supervisory controller, cause the supervisory controller to perform the various functions described herein.

[0134] The supervisory controller may be wired to the temperature control unit (or its sub-segments, for example, to a valve that connects the boiler to the radiator, but not necessarily to each sub-segment), and / or may be wirelessly connected to the temperature control unit, wherein the local controller on the temperature control unit (or again, on its sub-segments) communicates wirelessly with the supervisory controller.

[0135] Example

[0136] Figure 1 A flowchart of an implementation method is shown, in which energy waste status is detected by using the CUSUM method on available data. Figure 1 In this context, the energy waste state is described as "an open window is detected," but those skilled in the art will understand that this is merely an example.

[0137] The CUSUM method is a statistical process control technique used to monitor process behavior over time and detect when a process deviates from its expected behavior. In the context of this system used to detect energy waste, the CUSUM method is used to monitor the prediction error of energy loss in each room. The prediction error can be calculated as the difference between the following: The predicted energy loss in each room is estimated based on a one-step advance prediction of the indoor temperature via a simple model. The actual energy loss in each room is estimated based on the actual indoor temperature measured by the sensors. Based on room temperature, the actual energy loss can be calculated as follows: ,in It is the estimated thermal resistance of the room. It's the outdoor temperature. It refers to the indoor temperature.

[0138] The CUSUM method works by summing the prediction errors for energy loss in each room. The prediction error increases when energy-wasting conditions exist (e.g., open windows, where outdoor and indoor temperatures differ), because the simple model is fitted to represent the thermal behavior of rooms without energy-wasting conditions.

[0139] The calculation of CUSUM itself can be performed as follows: For time series, the following prediction errors are periodically measured and stored: The estimated average is The estimated standard deviation is .

[0140] The sum of the upper and lower cumulative processes is defined using the following formula:

[0141] if ,but:

[0142] if ,but:

[0143] Then, an offset point is detected when the following occurs (i.e., an energy waste state exists): or

[0144] In the above formula, It is a time step It is an estimate versus the actual. The estimated "average" of the difference (which can be calculated over at least 50 time steps when we know that the energy waste state does not exist) compensates for systematic biases in the prediction error (e.g., constant biases in the prediction error that are independent of the presence of the energy waste state).

[0145] It is an estimate versus the actual. The estimated standard deviation of the difference (which can be calculated over at least 50 time steps when we know that the energy waste state does not exist). This parameter affects the values ​​of USUM and LSUM, as well as the offset standard.

[0146] This is the minimum average offset to be detected (default: 1). This is the control limit (default: 5). This parameter determines the offset point threshold. Lower values ​​result in a more sensitive detector, while higher values ​​result in a more robust detector.

[0147] In some variations, heuristic rules can be added, such as allowing energy waste states to be identified / confirmed only when the indoor temperature moves in the opposite direction to the set point. More preferably, it should be limited to cases where the indoor temperature drops, as the most common energy waste states are related to heat loss, and this is the most important situation to address due to the large amount of energy wasted.

[0148] Other heuristics could be a predetermined minimum time period (e.g., at least 5 minutes) within which energy waste is identified.

[0149] Figure 2A and Figure 2B The training of a machine learning model (in this case, a neural network) for detecting wasted energy states, according to an embodiment of the present invention, is shown. Figure 2A ) and deployment ( Figure 2B )method.

[0150] exist Figure 2A In the first phase described, indoor and outdoor temperature data are collected for a large number (10,000) of residences (although this could also be performed for other types of buildings, or in combination with other types of buildings such as offices, shops, etc.), with data collected for at least one year for each residence. For the relevant dataset to be received, it is important that the residences for which data is collected are diverse in terms of location, orientation, insulation, window size, number of windows, shading, etc. Windows (or at least some windows) of the residences are equipped with sensors to determine their status (open or closed, as 1 or 0), and each window has a sensor associated with the room for which the indoor temperature is determined. Data from these sensors is also collected and linked to the indoor and outdoor temperature data collected throughout the year. In this case, the sampling time can be selected as 1 second, although other options remain open. The available data is compiled into tables, where indoor temperature data (for the room where the window is located) and outdoor temperature data are provided for each point in time and for each window.

[0151] Based on this table, a first neural network (f1) is trained to estimate the energy / heat changes in the room. A second neural network (f2) is trained to determine the presence (or absence) of energy waste, in this case, an open window (W). Window sensor data is used as the ground truth for both networks. Typically, further details can be added to the networks (via tables, or separately), particularly information about the operating status of the room's temperature control unit, and, if available, its output (and optionally, the type of energy source).

[0152] The types of neural networks can vary, such as recurrent neural networks (RNNs), like long short-term memory (LSTM) networks, or convolutional neural networks (CNNs), generative adversarial networks (GANs), transformer neural networks, like time fusion transformers (TFTs), etc.

[0153] Subsequently, the neural network parameters are optimized, such as the backtracking time (H, indicating the time period of historical indoor and outdoor temperature data and other data used by the model), sampling time (N), the number of layers used, and the number of neurons in each layer.

[0154] Ultimately, an acceptable model is achieved and deployed with optimized parameters, such as... Figure 2B As shown. Of course, even during deployment, further optimization based on new inputs continues. After deployment, the trained first and second neural networks receive indoor temperature data and outdoor temperature data (outside the rooms) for each room, with a sampling rate N (optimized) and a backtracking time (H), and based on this received data, the energy waste state (whether the windows are open or closed) can be estimated, and the energy / heat change can be estimated.

[0155] In the above examples, the main reference is to a situation where heat is lost to the external environment, but it is clear to those skilled in the art that the principle can be extended to heat gain from the external environment and its impact on the above examples.

[0156] It should be understood that the invention is not limited to any of the previously described implementations, and that some modifications may be made to the presented manufacturing examples without re-evaluating the appended claims.

Claims

1. A system for temperature control in one or more rooms, each of the one or more rooms including at least one temperature control unit, the temperature control unit including one or more heating and / or cooling units, the system including a supervisory controller configured to control the operation of the temperature control unit. in, The supervisory controller is configured to estimate the heat variation in each of the one or more rooms. The estimated heat change is based on historical outdoor temperature data indicating outdoor temperature over a longer period of time and historical indoor temperature data indicating indoor temperature in each of the one or more rooms over the longer period of time. The supervisory controller is configured to detect the presence of energy waste in at least one of the one or more rooms based on the estimated heat change in each of the one or more rooms and the current indoor temperature of each of the one or more rooms. The monitoring controller is configured to take intervention action when an energy waste state is detected in at least one room. Preferably, the intervention action is carried out in or for the at least one room. More preferably, the intervention action includes at least shutting down the temperature control unit in the at least one room where the energy waste state is detected.

2. The system for temperature control according to claim 1, wherein, The estimated heat change is estimated via a formula or model based on the historical outdoor temperature data and the historical indoor temperature data, wherein the formula or model receives the indoor temperature at a previous time point as input, preferably at most 1 hour ago, more preferably at most 30 minutes ago.

3. The system for temperature control according to claim 2, wherein, The formula or model receives the outdoor temperature at a previous time point as further input, preferably at least 1 hour ago, more preferably at least 30 minutes ago, and preferably at the same time point as or close to the time point of the input indoor temperature.

4. The system for temperature control according to claim 2 or 3, wherein, The formula or model is also based on historical operating data of the temperature control units, which defines the operating status of each temperature control unit over the longer time period.

5. The system for temperature control according to any one of claims 1 to 4, wherein, The supervisory controller is configured to estimate the expected indoor temperature of each of the one or more rooms, the expected indoor temperature of each room being determined by the estimated heat change of each room and the indoor temperature of each room at a previous time point, preferably at least one hour ago, more preferably at least 30 minutes ago. The existence of energy waste in a room is determined by comparing the expected indoor temperature with the current indoor temperature.

6. The system for temperature control according to any one of claims 1 to 5, wherein, The supervisory controller is also configured to determine the actual heat variation in each of the one or more rooms. The actual heat change is based on the indoor temperature history data, optionally the outdoor temperature history data, and also on the current indoor temperature in each of the one or more rooms.

7. The system for temperature control according to any one of claims 1 to 6, wherein, The supervisory controller is also configured to determine the actual heat variation in each of the one or more rooms. The actual heat change is based on the historical indoor temperature data, optionally the historical outdoor temperature data, and also on the current indoor temperature in each of the one or more rooms. The supervisory controller is responsible for temperature control in multiple rooms, and when an energy waste condition is detected, the supervisory controller is configured to consider the heat changes in each of the rooms to identify heat transfer between two or more of the multiple rooms.

8. The system for temperature control according to any one of claims 1 to 7, wherein, The supervisory controller detects the presence of energy waste based on an error function defined by the following formula: ； in, This refers to the actual change in heat. It is the estimated heat change in each room, also taking into account the heat generated from the temperature control unit in each room, and wherein the existence of an energy waste state is determined based on the cumulative sum control graph (CUSUM) value of the error function, wherein the existence of an energy waste state is confirmed if the CUSUM value exceeds a predetermined upper limit or if a predetermined lower limit exceeds the CUSUM value.

9. A method for temperature control in one or more rooms, each of the one or more rooms including at least one temperature control unit, the method comprising the steps of: a. Estimate the heat change in each of the one or more rooms, wherein the estimated heat change is based on historical outdoor temperature data indicating outdoor temperature over a longer period of time and historical indoor temperature data indicating indoor temperature in each of the one or more rooms over the longer period of time. b. Detect the presence of an energy waste state in at least one of the one or more rooms based on the estimated heat change in each of the one or more rooms and the current indoor temperature in each of the one or more rooms; c. Taking an intervention action when an energy waste state is detected in at least one room, preferably wherein the intervention action is carried out in or for the at least one room, more preferably wherein the intervention action includes at least shutting down the temperature control unit in the at least one room where the energy waste state is detected.

10. The method according to claim 9, wherein, The step of estimating heat changes is performed via a formula or model based on the historical outdoor temperature data and the historical indoor temperature data, wherein the formula or model receives the indoor temperature at a previous time point as input, preferably at least 1 hour ago, more preferably at least 30 minutes ago.

11. The method according to claim 10, wherein, The formula or model receives the outdoor temperature at a previous time point as further input, preferably at least 1 hour ago, more preferably at least 30 minutes ago, and preferably at the same time point as or close to the time point of the input indoor temperature.

12. The method according to claim 10 or 11, wherein, The formula or model is also based on historical operating data of the temperature control units, which defines the operating status of each temperature control unit over the longer time period.

13. The method according to any one of claims 10 to 12, the method comprising the following steps: Estimate the expected indoor temperature of each of the one or more rooms, the expected indoor temperature of each room being determined by the estimated heat change of each room and the indoor temperature of each room at a previous time point, preferably at least one hour ago, more preferably at least 30 minutes ago. The existence of energy waste in a room is determined by comparing the expected indoor temperature with the current indoor temperature.

14. The method according to any one of claims 10 to 13, further comprising the following steps: Determine the actual heat change in each of the one or more rooms, the actual heat change being based on the indoor temperature history data, optionally the outdoor temperature history data, and also based on the current indoor temperature in each of the one or more rooms.

15. The method according to any one of claims 10 to 14, wherein, The existence of energy waste is detected based on an error function, which is defined by the following formula: ; in, This refers to the actual change in heat. It is the estimated heat change in each room. The formula also takes into account the heat generated by the temperature control unit in each room, and The method includes the following steps: The existence of an energy waste state in a room is determined based on the cumulative sum control chart (CUSUM) value of the error function, wherein the existence of an energy waste state is confirmed if the CUSUM value exceeds a predetermined upper limit or if a predetermined lower limit exceeds the CUSUM value.

Images