Data-reduced training of neural networks for HVAC control.

Abstract

Disclosed is a system and method for controlling HVAC components of a building, the method including storing or accessing a temperature model for calculating a temperature rate of change of at least one of the building's zones; receiving sensor data from a plurality of temperature sensors within the building; receiving outdoor temperature data; calculating a temperature rate of change of the at least one zone by providing the sensor data and the outdoor temperature data to the temperature model; determining a predicted temperature based on the temperature rate of change; and generating a command signal to control at least one setpoint of the HVAC system, wherein the temperature model includes an explicitly defined model portion and an approximate model portion, the approximate model portion being solved using a function approximator trained to generate an output reflective of an amount of heat added to or removed from the at least one zone.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of and priority to U.S. Provisional Patent Application No. 63 / 397,537, filed August 12, 2022, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure relates generally to building management or building automation. More particularly, the present disclosure relates to systems and methods for controlling heating, ventilation, and air conditioning systems in buildings using neural networks. [Background technology]

[0003] A building management system (BMS) or building automation system (BAS) is a system used in a building to manage heating, ventilation or air conditioning (HVAC), lighting, power, security, elevators, and other building systems.

[0004] HVAC accounts for 30% to 60% (e.g., 40%) of total energy in commercial buildings, and achieving HVAC energy reductions is now a priority, but achieving or sustaining energy-efficient HVAC is not easy.

[0005] Traditionally, HVAC systems are thought of as a group of independent pieces of mechanical equipment. Each pump, chiller, tower, and air handling unit is designed to be turned on, run at a certain speed, and then turned off. This thinking leads to logic in the equipment components being designed to operate efficiently independently, and to a BMS that controls the equipment by automatically turning it on and off. The operational data typically present in a typical BMS is not easily accessible to building managers. If operational data is available, it is likely in the form of a stream of unformatted data points, a format incompatible with measuring performance or diagnosing problems.

[0006] Management of HVAC by a BMS or BAS typically involves the use of thermostats or other sensors located at various locations within a building, with each thermostat acting as a sensor to measure the temperature at that particular location. Typically, a thermostat includes a target temperature range (i.e., an associated value that is added or subtracted from the target temperature, thus defining what is called a tolerance temperature band). This temperature range can be changed by the room occupant or building manager. When the thermostat's measured temperature in the room is outside the target temperature range, the thermostat sends a command to the HVAC equipment to activate heating or cooling for the room. This reactive management of room temperature can be inefficient and not cost-effective. Taking room occupancy or other environmental variables that affect the building's thermal target into account in controlling the HVAC system can result in improved efficiency and cost-effectiveness. In addition to increased efficiency and cost savings, predictive, rather than reactive, HVAC control improves comfort and further saves energy.

[0007] To link temperature control to room occupancy, many existing BMS or BAS require replacing existing thermostats with thermostats that include occupancy or vacancy sensors, or adding occupancy or vacancy sensors to specific rooms and linking these additional sensors to HVAC controls to use this data. Large buildings can have hundreds of thermostats. Replacing hundreds of simple thermostats with occupancy sensors can be costly and labor intensive, presenting a significant obstacle to adopting this technology in large buildings.

[0008] Temperature predictions for HVAC systems are often based on a physics-based resistance-capacitance (RC) circuit model (or simply RC model), which can be formulated as a differential equation to simulate the thermal behavior of an HVAC system. RC models are usually used for temperature predictions in buildings with a small number of zones, typically less than 10.

[0009] As an alternative to RC models, black-box models such as neural networks may be used to predict the temperature of HVAC systems. However, using neural networks to simulate the behavior of large HVAC systems likely requires significant computational resources. For example, implementing a neural network to control the HVAC system of a large building likely requires a significant amount of input data to train the neural network, which could take days or even months to collect. Summary of the Invention

[0010] According to one aspect, a computer-implemented system for controlling a heating, ventilation, or air conditioning (HVAC) system of a building having multiple zones is provided, the system including: a processor; and a non-transitory memory device, wherein the memory device stores a temperature model for calculating a rate of temperature change of at least one of the building's zones; and a set of computer-readable instructions, the temperature model including an explicitly defined model portion and an approximate model portion, the approximate model portion being solved using a function approximator configured to generate an output reflective of an amount of heat added to or removed from the at least one zone; the set of instructions, when executed, causes the processor to receive sensor data from a plurality of temperature sensors within the building; receive outdoor temperature data; calculate a rate of temperature change of the at least one zone by providing the sensor data and the outdoor temperature data to the temperature model; determine a predicted temperature based on the rate of temperature change; and generate a command signal to control at least one setpoint of the HVAC system in response to the predicted temperature.

[0011] In some embodiments, the function approximator includes a neural network trained to generate at least a portion of the output.

[0012] In some embodiments, the output from the approximation model portion also reflects the total amount of temperature change over a given period based on the current setpoint.

[0013] In some embodiments, the output from the approximation model portion further reflects environmental thermal contributions.

[0014] In some embodiments, the environmental thermal contribution includes solar radiation.

[0015] In some embodiments, the temperature ramp rate of at least one zone i is:

number

number

[0016] In some embodiments,

number

[0017] In some embodiments, the temperature ramp rate of at least one zone i is:

number

[0018] In some embodiments, f i (U,SolRad,..)=NN(θ,U,Solrad,..)[i], where NN represents a function approximator of the approximate model part, U represents a vector of one or more HVAC controls of the HVAC system, SolRad is given data reflecting the contribution of solar radiation, and θ represents one or more parameters of the function approximator NN.

[0019] In some embodiments, formula f i (U, SolRad, ..) is determined for each zone in the plurality of zones.

[0020] According to another aspect, there is provided a computer-implemented method for controlling HVAC components of a building, the method including storing or accessing a temperature model for calculating a temperature rate of change of at least one of the building's zones; receiving sensor data from a plurality of temperature sensors in the building; receiving outdoor temperature data; calculating a temperature rate of change of the at least one zone by providing the sensor data and the outdoor temperature data to the temperature model; determining a predicted temperature based on the temperature rate of change; and generating a command signal for controlling at least one setpoint of the HVAC system responsive to the predicted temperature, wherein the temperature model includes an explicitly defined model portion and an approximate model portion, and the approximate model portion is solved using a function approximator configured to generate an output reflective of an amount of heat added to or removed from the at least one zone.

[0021] In some embodiments, the function approximator includes a neural network trained to generate at least a portion of the output.

[0022] In some embodiments, the output from the approximation model portion also reflects the total amount of temperature change over a given period based on the current setpoint.

[0023] In some embodiments, the output from the approximation model portion further reflects environmental thermal contributions.

[0024] In some embodiments, the environmental thermal contribution includes solar radiation.

[0025] In some embodiments, the temperature ramp rate of at least one zone i is:

number

number

[0026] In some embodiments,

number

[0027] In some embodiments, the temperature ramp rate of at least one zone i is:

number

[0028] In some embodiments, f i (U,SolRad,..)=NN(θ,U,Solrad,..)[i], where NN represents a function approximator of the approximate model part, U represents a vector of one or more HVAC controls of the HVAC system, SolRad is given data reflecting the contribution of solar radiation, and θ represents one or more parameters of the function approximator NN.

[0029] In some embodiments, formula f i (U, SolRad, ..) is determined for each zone in the plurality of zones.

[0030] According to yet another aspect, there is provided a non-transitory computer readable medium having stored thereon machine-interpretable instructions that, when executed by a processor, cause the processor to perform the computer-implemented method of any one of the preceding claims.

[0031] Upon reading this disclosure, those skilled in the art will recognize many additional features and combinations of the embodiments described herein. [Brief explanation of the drawings]

[0032] [Figure 1A] FIG. 1 is a schematic diagram illustrating an example system having a controller device for controlling HVAC components of a building, according to one embodiment. [Figure 1B] FIG. 1 is a schematic diagram illustrating another example system having a controller device for controlling HVAC components of a building, according to one embodiment. [Figure 2]FIG. 1 is a schematic diagram illustrating various parameters within a building that can be controlled by a controller device, according to one embodiment. [Figure 3] FIG. 1 is a schematic diagram illustrating an example architecture of a system including a controller device for controlling HVAC components of a building, according to one embodiment. [Figure 4] FIG. 2 is another schematic diagram illustrating an example architecture of a system including a controller device for controlling HVAC components of a building, according to one embodiment. [Figure 5] FIG. 1 is a schematic diagram of an exemplary architecture of a neural network maintained by a machine learning model of a controller device, according to one embodiment. [Figure 6] FIG. 1 illustrates an exemplary process performed by a controller device for dynamic management of HVAC components of a building, according to one embodiment. [Figure 7] FIG. 1 is a schematic diagram of a computing device that may be used to implement a controller device.

[0033] These drawings depict, for purposes of illustration, example embodiments to which variations, alternative configurations, alternative components, and modifications may be made. DETAILED DESCRIPTION OF THE INVENTION

[0034] Efficient management and operation of the heating, ventilation, and air conditioning (HVAC) system(s) in a building or network of buildings can reduce energy consumption, leading to lower building operating costs. Predictive or more finely tuned, responsive control of HVAC systems can improve efficiency as well as environmental conditions, comfort, and reduce management time.

[0035] The present disclosure provides a system and method for predicting future temperatures in a multi-zone building using machine learning to facilitate proactive control. Neural networks, a type of universal function approximator that allows for scalable solutions, can be utilized to determine the predicted temperatures. Based on the predicted temperatures, one or more command signals can be generated to control at least one setpoint of the HVAC system to achieve a target temperature.

[0036] In some embodiments, machine learning for HVAC temperature prediction simulates physics, such as environmental thermodynamics, in the form of differential equations. These embodiments can reduce the data and computational resource requirements for training neural networks; for example, while it may traditionally take months to gather a sufficient amount of training data to train a machine learning model for a large HVAC system without differential equations, some embodiments using differential equations in conjunction with neural networks can collect the necessary data in just days or hours. The use of differential equations in machine learning models improves model generalization, enhances performance, and increases human interpretability. Furthermore, by embedding physically meaningful parameters, this approach can provide information about the thermal properties of a building.

[0037] Neural ordinary differential equations (ODEs) generally embed neural networks into the mathematical structure of ordinary differential equations. A "generalized ordinary differential equation" is a neural ODE that allows the known parts of the equation to be made explicit, and the unknown components of the neural ODE to be modeled with a neural network.

[0038] Generalized ordinary differential equations (UODEs) have not been applied to the domain of HVAC temperature prediction. In some embodiments, the non-HVAC components of the RC model-based differential equations are explicitly implemented, and the HVAC contribution can be modeled using neural networks. Furthermore, neural network features can be used to extract building thermal characteristics and inform commercial HVAC control strategies.

[0039] As mentioned above, HVAC temperature predictions are typically either exclusively physics-based or exclusively neural network-based, each with its own advantages and disadvantages. The embodiments described herein integrate both physics-based differential equations and neural networks to improve the efficiency of HVAC control systems. For example, the flexibility and expressive power of neural networks can be leveraged to model various systems and learn directly from source data (e.g., temperature data). At the same time, some embodiments leverage prior physical knowledge by modeling the physical knowledge in differential equations (e.g., RC models), reducing the data and computational resource requirements for training the neural network. As a result, the system can provide meaningful information about the thermal characteristics of a building.

[0040] Traditionally, RC models used to simulate HVAC systems typically require knowledge of the spatial relationships between zones in order to design appropriate RC circuit analogs, which can be complex and time-consuming in buildings with many zones. In the embodiments described herein, the system makes no such assumptions and allows the learned parameters of a neural network to represent the likely neighborhoods of a given zone.

[0041] In some embodiments, the exemplary system disclosed herein may include a control module configured to extract and monitor all data items or "points," predict values ​​for all data items or "points" taking into account the thermodynamic properties of both the building as a whole and the zones therein, and control each component of the HVAC system to optimize HVAC performance. HVAC system components and operating parameters controlled by the exemplary system may include setpoints, fan speeds, valve openings, damper positions, chilled water temperatures, hot water temperatures, pipe pressures, pressure differentials, pump speeds, evaporator speeds, compressor pressures, etc.

[0042] In some embodiments, the example systems disclosed herein can be configured to send instructions to a BMS / BAS to adjust operating parameters of one or more or all HVAC components, or to send instructions directly to one or more specific HVAC devices to adjust operating parameters.

[0043] In some embodiments, the exemplary systems disclosed herein can provide and enable predictive HVAC control. The exemplary systems can utilize advanced data processing and / or artificial intelligence, including traditional linear regression models, nonlinear regression models, supervised learning, unsupervised learning, and deep learning techniques, to maintain or update control modules that can provide predictive HVAC control systems and modules that can learn. In some embodiments, the exemplary systems can integrate and interconnect devices, including sensors and HVAC components, within existing building infrastructure using wired or wireless networks. Some of the features described herein can utilize, for example, big data systems, machine learning and artificial intelligence, cloud computing technologies, and cloud services.

[0044] Predictive control of an HVAC system may include control of one or more HVAC systems based on forecasted future outdoor environmental conditions and past interior characteristics or behavior, including, but not limited to, temperature, wind speed, wind direction, percent cloud cover, percent eclipse, outdoor humidity, date, time of day, thermodynamic patterns derived from historical data sets, and predicted occupancy.

[0045] In some embodiments, the example systems disclosed herein can receive forecasts or other information from third-party providers, including government weather stations. Weather forecasts include short-term, medium-term, and long-term weather forecasts. Weather conditions and forecasts can be updated periodically. In some embodiments, weather information and tracking is updated at regular intervals, for example, every 1, 5, 10, 15, 20, 30, or 60 minutes. Optionally, in some embodiments, because zones or rooms with exterior walls or windows are more susceptible to external conditions than interior zones or rooms, the system can be configured to use historical data to predict the impact of specific weather conditions on the HVAC needs of specific parts of a building.

[0046] In some embodiments, when an exemplary system receives weather forecasts, it can adjust the operation of one or more HVAC components based on those forecasts. For example, in embodiments where the HVAC system includes multiple HVAC units, individual units can be turned on or off or adjusted based on predicted needs, optionally selecting heating or cooling units based on predicted conditions, and multiple units can operate at partial load if such operation is more efficient than a single unit operating at full load. In some embodiments, when multiple chillers serve a building, the system and method can provide chiller staging. Optionally, chiller staging takes into account equipment configuration and type, cooling capacity, chilled water flow rate, and power consumption by condensers and water tower fans.

[0047] In some embodiments, temperature changes can be predicted to allow buildings to be cooled or heated in advance, optionally allowing for off-peak power consumption.

[0048] In some embodiments, the exemplary system can be configured to control ventilation or airflow to redistribute heat or cold, for example, adjusting airflow to move cooler air from a basement floor to an upper floor. In some embodiments, the system can be configured to adjust outdoor air intake, optionally using outdoor air to cool a space, for example, by letting cool outdoor air flow into a building at night to reduce or avoid daytime cooling. The system can be configured to redistribute heat or cold only when it is cost-effective or advantageous to do so.

[0049] In some embodiments, the exemplary system can be configured to control HVAC components based on room occupancy, including controlling heating, cooling, and ventilation based on actual or predicted occupancy. Optionally, zones or rooms can be pre-cooled or pre-heated based on predicted occupancy. In some embodiments, the use of supply fans, hood exhaust, and make-up fans depends on actual or predicted occupancy, with use reduced in underutilized rooms or zones. Optionally, the system can be configured to automatically turn on ventilation systems or appliances (e.g., lights) when an occupant is detected in a room or zone. Optionally, the type of occupancy is considered in addition to occupancy; for example, an occupant engaged in strenuous activity will have a greater impact on environmental conditions than an inactive occupant.

[0050] In embodiments where room occupancy is predicted, HVAC settings can be adjusted to account for the expected number of occupants to compensate for or adjust for heat or CO2 generated by occupants in the space. Optionally, the system can control ventilation of a particular space based on controlling actual or predicted CO2 levels. In some embodiments, outdoor air changes are minimized to the extent necessary to maintain acceptable air quality, i.e., CO2 levels. In environments with extreme weather conditions, indoor air can optionally be filtered to minimize the amount of outdoor ventilation required.

[0051] In some embodiments, the exemplary system may be configured to manage a building's energy consumption, including managing power consumption during peak demand periods, thereby reducing costs.

[0052] In some embodiments, the exemplary system can be configured to manage HVAC components depending on the quality of outside air and / or the quality of air within a particular region or zone. For example, the system can be configured to control air change rates based on measured pollutant levels, including carbon monoxide and / or nitrous oxide levels, within a particular zone or area, such as a parking lot or loading dock. Optionally, the system can be further configured to filter the outside air or indoor air flow to improve air quality when needed.

[0053] Optionally, exemplary systems may be configured to adjust air flow or air intake depending on the quality of the outside air. For example, in some embodiments, if the outside air quality is poor, the outside air may be filtered or purified to remove airborne particles such as dust, pollen, and bacteria. Thus, in some embodiments, the air intake is selected based on the actual or expected air quality.

[0054] In some embodiments, multiple factors are considered in predicting HVAC requirements, and individual factors can be weighted, based on historical data or expected impact.

[0055] In some embodiments, the exemplary system may be configured to allow building managers to weight or prioritize one or more factors, for example, to maximize energy efficiency or cost savings. In such embodiments, a user interface may be provided that allows for the selection or ranking of priorities.

[0056] In some embodiments, the exemplary system can be configured to control HVAC components to minimize energy consumption while maintaining a minimum level of comfort in an occupant zone and / or area, taking into account factors that affect the occupant's perceived comfort level, including temperature and humidity. For example, both temperature and humidity setpoints can be adjusted to reduce energy requirements while maintaining comfort.

[0057] In some embodiments, the exemplary system can be configured for a single building or a network of buildings. In some embodiments, two or more buildings in a network of buildings are physically connected, for example, by a tunnel, interior corridor, or bridge. In such cases, the temperature and humidity setpoints in a first building can affect the temperature and humidity setpoints in a second connected building, and the exemplary system can be configured to operate the HVAC components of the buildings as a whole.

[0058] In some embodiments, the exemplary system is configured to be retrofitted into existing buildings and can be further configured to interface with existing HVAC components. In other embodiments, the system can be retrofitted into new buildings, where the system and method are optionally integrated into or are a module of a BMS or BAS.

[0059] In some embodiments, the exemplary system can be configured to interact directly with components of the HVAC system, e.g., communicate directly with sensors and actuators / controllers of the HVAC system. In alternative embodiments, the system can communicate directly with an existing master controller. Optionally, the system can perform a combination of direct interaction with the master controller and direct communication with components of the HVAC system.

[0060] In some embodiments, an exemplary system may generally include a controller device in communication with sensors and components of the HVAC system. The controller device may further be connected to a remote or cloud server. In some embodiments, both the controller device and the remote or cloud server include algorithm modules configured to analyze data.

[0061] The analysis of sensor data and / or calculation of changes in HVAC parameters can be performed on an on-site controller device, a remote server, or a cloud server, or a combination thereof. In some embodiments, the system is configured such that if a link to a remote server or cloud server does not exist, processing switches over to the on-site controller device.

[0062] In the illustrated embodiment, one or more neural networks are used for function approximation of certain portions of the ordinary differential equations. In other embodiments, other types of function approximators may be used in place of or in combination with one or more neural networks. In some embodiments, such function approximators may include regression-based function approximators, polynomial-based function approximators, etc. Generally speaking, function approximators may include classes of generic functions that can be parameterized to represent various input-output mappings.

[0063] 1A , a system is illustrated having a controller device 100 in communication with a building's HVAC components 400 and a remote server 300. More particularly, a system for managing a building's HVAC components 400 is described, the system including a processor and a memory storage device, the memory storage device storing a set of instructions and a temperature model for calculating a temperature rate of change of at least one of the building's zones, the temperature model including an explicitly defined model portion and an approximate model portion, the approximate model portion being solved using a neural network trained to generate an output reflecting an amount of heat added to or removed from the at least one zone, the set of instructions, when executed by the processor, causing the processor to receive sensor data from a plurality of temperature sensors within the building, receive outdoor temperature data, calculate a temperature rate of change of the at least one zone by providing the sensor data and the outdoor temperature data to the temperature model, determine a predicted temperature based on the temperature rate of change, and generate a command signal to control at least one setpoint of one or more components 400 of the HVAC system in response to the predicted temperature.

[0064] In some embodiments, the plurality of forecasts may include at least one of a predicted temperature value, a predicted water usage, a predicted power usage, a predicted gas usage, a predicted weather, and a predicted humidity level. Some forecasts, such as a predicted temperature value, a predicted power usage, or a predicted humidity level, may be determined based in part on a weather forecast from the remote server 300 or the third-party server 600.

[0065] Forecasts predicting building dynamic states can be obtained from one or more control modules or from the BAS / BMS, if present. Dynamic states can include, for example, predicted occupancy levels in one or more zones of the building.

[0066] In some embodiments, the plurality of current building conditions may include at least one of a number of zones, temperature readings, setpoints, sensor data, actuator data, occupancy schedules, and occupancy data. In some embodiments, the plurality of current building conditions are received from a building management system (BMS) 150.

[0067] In some embodiments, the system further comprises a display device for displaying one or more operating values ​​of the HVAC components, which may be, for example, display device 420 shown in FIG.

[0068] 1B , an exemplary system may include a controller device 100 in communication with an existing BMS or BAS 150 and a remote server 300. Forecasts may be provided directly to the controller device 100 from a third-party server 600 or via the remote server 300 as shown. More specifically, a system for managing a building's HVAC components 400 exists, the system including a controller device 100 in communication with an existing BMS or BAS 150. The BMS or BAS 150 communicates with sensors 200 and HVAC components 400 in the building and acts as an intermediary between the controller device 100 and the sensors 200 and HVAC components 400, such that the controller device 100 provides control instructions to the BMS or BAS 150, which in turn communicates with the necessary HVAC components 400. Similarly, the sensors 200 communicate with the BMS or BAS 150, which transmits data to the controller device 100.

[0069] In a further exemplary embodiment, the system includes a controller device 100 in communication with an existing BMS or BAS 150 , one or more sensors 200 or one or more HVAC components 400 .

[0070] This system embodiment allows for the system to be retrofitted using an existing building thermostat without the need to modify the thermostat. The controller device 100 can be advantageously configured and fabricated to communicate with an existing BMS / BAS 150, sensors 200, actuators, and / or individual controllers, and the Internet, for low-cost and easy retrofitting of existing thermostats and HVAC equipment in a building. The actuators in the HVAC system control the dampers. Using a low-voltage signal, the actuators position the dampers at any point between fully open and fully closed.

[0071] 1A and 1B, the system can be an Internet of Things (IoT) solution for individually controlling HVAC components 400 of a building management system (BMS) 150 to optimize overall system efficiency. The system can advantageously utilize existing sensors 200 (e.g., thermostats including thermometers) throughout the building and actuators or controllers on the various HVAC components 400 to individually control each at the individual component level. A controller device 100 communicates with both these sensors 200, actuators, and the Internet (i.e., a remote server 300). By converting data formats, the controller device 100 is adapted to receive information from the sensors 200 and send it to the actuators, even if the data is in a disparate format.

[0072] In some embodiments, the controller device 100 can execute one or more control modules that implement one or more algorithms to predict a building's needs for heating, cooling, and ventilation throughout the building. These predictions can be based on data obtained from sensors 200 (e.g., temperature, speed, humidity, and pressure, sensors for detecting human presence) and data obtained from the internet (e.g., local weather forecasts including sunlight intensity and direction, temperature, wind, or precipitation). By performing analysis of the building's historical data sets into the program, the controller device 100 can determine how to control each of the HVAC components 400 to optimize the system (i.e., minimize cost or energy consumption) based on one or more temperature settings.

[0073] In some embodiments, the controller device 100 can determine predicted temperatures for one or more zones within the building based on a neural network model of temperature change stored in a data storage device and real-time temperature data from one or more sensors 200, as described in more detail below.

[0074] Traditionally, HVAC control systems that rely narrowly or primarily on neural network systems tend to require large amounts of training data, necessitating implementation at a centralized (e.g., rooftop) HVAC system or remote implementation due to data and training requirements. In contrast, in some embodiments, controller device 100 is implemented entirely in a thermostat or IoT device, which may be a wall-mounted device. This is possible due to the reduced computational resources required by utilizing a generalized ODE to predict the temperature within a given room or building, which can use a function approximator (e.g., a neural network) to simulate environmental thermal factors and reduce the amount of training data required to train a control module to control the HVAC system. This provides more direct access to building-wide sensor data, better data security, and a faster training and deployment process for control modules that can be implemented and trained on-site at building 170.

[0075] Electrical equipment 500 can also be controlled in a similar manner by controller device 100. This is particularly true for lighting, where the need for lighting is highly dependent on measured occupancy, which affects the need for heating or air conditioning. This can also be true for other types of equipment with similar requirements and effects, such as industrial equipment, local servers, occupancy-dependent electrical loads (e.g., decorative equipment), power distribution equipment, power storage (i.e., battery banks, etc.), alarm systems, etc.

[0076] The control module and temperature model 110 may be stored and executed on either or both the controller device 100 and a remote server 300 (eg, the cloud), depending on bandwidth and time responsiveness needs.

[0077] Based on the analysis and decisions from the controller device 100, each HVAC component 400 is instructed through the controller device 100 to adjust its operation. Components include, but are not limited to, air system components (fresh air intakes, ducts, exhaust and supply fans, night purge, (dehumidification)), cooling system components (chillers, compressors, condenser water feed, head pressure, condensing pressure), and heating system components (boilers, steam pressure or delivery). A large number of components can be managed and controlled in real time or near real time by the controller device 100 based on various inputs including the current state of the building, various forecasts, and various user goals.

[0078] In some embodiments, the controller device 100 may be connected to an electric or thermal storage device, such as, for example, a battery system, an electric vehicle, or a fuel cell system. The controller device 100 may also be connected to a potential energy storage system, such as a flywheel, a compressed air tank, a water tank, underground heat, one or more storage media, a phase change material, etc.

[0079] In some embodiments, the controller device 100 may be connected to one or more district heating providers (eg, hot water or steam providers) or cooling providers (eg, chilled water providers).

[0080] Nowadays, building management systems (BMS) are commonly used in large buildings, making the operational status of equipment more visible to operators and easier to control. This visibility focuses all attention on monitoring tools, without considering the thermodynamic balance within the building. In fact, prior art BMSs focus on optimizing the performance of individual equipment, while at the same time wasting more than 50% of the building's thermal energy into the environment. Instead of managing the operational status of individual equipment as in a typical BMS, the disclosed system and controller device 100 is configured to manage the energy flow within the building by collecting real-time data from sensors 200 placed throughout the building and managing the overall heat flow.

[0081] This system can be configured to reduce thermal energy inefficiencies within a building. The controller device 100 can be used not only to collect data from the sensors 200 but also to connect (possibly through the BMS 150) to all chillers (or any other cooling system components), boilers (or any other heating system components), pumps, and the entire array of building control points. This infrastructure capability allows advanced algorithms or modules to be installed and deployed in the easily deployable controller device 100 to deliver the dynamic state of the building based on real-time environmental conditions and the building's internal loads, thus ensuring that all mechanical systems are optimally operating 24 / 7. This can be done by connecting the controller device 100 with existing equipment (e.g., sensors 200, controllers, actuators for HVAC components 400), possibly through the BMS 150. Building intelligence into the controller device 100 rather than the equipment (sensors, controllers, actuators) eliminates the need to replace existing equipment and allows for retrofitting. With this system in place, drift is eliminated, reducing occupant complaints, equipment alarms, and periodic commissioning (physical inspection and maintenance). HVAC systems achieve thermally balanced performance levels by generating and distributing, in real time, the precise amount of thermal energy needed to provide occupant comfort.

[0082] The potential energy savings realized by the disclosed embodiments, when retrofitted to existing HVAC equipment in a building, can easily exceed 30% of HVAC energy consumption. In some situations, even lower kW / ton ratios can be achieved by combining advanced control modules to match weather patterns to building heat load requirements in real time or near real time.

[0083] In some embodiments, the controller device 100 can implement relational control algorithms to optimize all equipment within all variable-flow HVAC systems (chillers, fans, pumps, etc.). Doing so allows the controller device 100 to use the minimum amount of power necessary to maintain occupant comfort levels. Control setpoints are automatically calculated based on real-time building load information inputs and weather conditions outside the building. This approach results in building-wide thermal load management rather than equipment-based management strategies.

[0084] The controller device 100 can implement relational control algorithms to monitor various parameters across the equipment and sensors 200 within a building and provide continuous automatic adjustments to the system based on building load, regardless of the type of equipment. In some embodiments, exemplary objectives of these systems and methods are to provide energy usage savings (kWh / year), demand savings (kW), heating / cooling load (therms), cooling tower water usage savings (gallons / year), carbon footprint reduction (lbs / year), and power usage effectiveness (PUE) reduction. In doing so, these systems and methods save significant amounts of money, reduce strain on the power grid, and improve tenant comfort.

[0085] This holistic view of the building monitored via controller device 100 provides real-time system adjustment recommendations, identifies operational inefficiencies, and provides real-time or near real-time building restart at any time of day. In some embodiments, this controller device 100 can manage energy demand by analyzing building occupancy, building thermodynamic patterns, and external weather to automatically adjust heating and cooling in various zones of the building to optimal levels in real time.

[0086] According to one embodiment, the system may include a remote server 300 connected to the controller device 100. The remote server 300 may be a single server, multiple servers each dedicated to a specific task, or multiple servers organized into a network to perform tasks in a distributed manner, e.g., a cloud.

[0087] According to one embodiment, referring to FIG. 3 , the controller device 100 locally collects and aggregates all data points from various sensors 200 and subsystems in a building, including, but not limited to, fans, electricity meters, elevators, occupancy counters, steam meters, interior or exterior air temperature, water, etc. To collect these data points, the controller device 100 can connect to an existing BMS 105 system via various protocols (e.g., BACnet, Modbus, LonWorks, etc.). This controller device 100 is compatible with all major BMSs on the market (e.g., JCI, Schneider, Honeywell, Siemens, Automated Logic, Panasonic, Legrand, Delta, IBM, Hitachi). An exemplary cloud solution server 300 network is connected to the controller device 100 by a data link 315 and collects data from one or more other data sources, including a weather station 310. The cloud solution server 300 includes one or more databases that can store data received from the controller device 100 and one or more other data sources. Data from edge buildings can be stored in a separate database, and the controller device 100 is optionally configured to update the information in the database at set intervals or when changing data is observed.

[0088] According to another embodiment, the cloud solution server 300 network collects data from multiple buildings, each building having its own controller device 100, 100 that communicates with respective electrical appliances such as lighting and with respective HVAC controllers.

[0089] 1B, the controller device 100 communicates with an HVAC controller or BMS / BAS 150, which in turn communicates with the components under its control. In one embodiment, the HVAC components 400 may be controlled by an HVAC controller that communicates directly with the controller device 100. In another embodiment, the HVAC controller that communicates directly with the controller device 100 controls both the HVAC components 400 and the sensors 200 to take all measurements from the sensors and send them to the controller device 100, and to communicate with the HVAC components after receiving instructions from the controller device 100.

[0090] According to one embodiment, data collected by device 100 from various sensors 200 located throughout the building is converted from the various protocols collecting the data into a universal open format and transmitted to a database on a remote server 300, such as a cloud computing database distributed across remote servers communicating over a network. This communication is preferably over a wireless connection, and more preferably over an encrypted connection due to the potential for personal information being collected by sensors 200.

[0091] According to one embodiment, a database, preferably located on a remote server 300 (e.g., located on the cloud), aggregates all data in the historical dataset and maintains fine granularity of the historical timeline for each data point. This database becomes a big data picture of the building's thermodynamic behavior over time and is used to extract additional value (e.g., trends) from the dataset. Specific algorithms use the dataset to build a real-time thermodynamic behavior model of the building, including predictions of optimal settings for various HVAC devices in real time.

[0092] The controller device 100 can execute a collection of control modules or algorithms in various combinations and sequences to utilize a thermodynamic model derived from a dataset containing real-time values ​​of data points, occupancy counts in each zone, and real-time external weather parameter conditions, along with forecasts for the next few hours. Based on the results of these control modules' execution, one or more control commands are generated and sent to various controllers within the building, which can direct adjustments to all HVAC devices to maintain desired temperature and humidity levels within the building at all times. These commands include optimal settings for the next time interval (e.g., five minutes), after which the controller device 100 initiates a new round of execution of the control modules and generates a new set of instructions. This process can continue and self-adjust based on the behavior of the HVAC equipment.

[0093] According to one embodiment, the control modules may be distributed across the controller device 100 and the remote servers 300 and 310. To optimize response times due to network communication latency, some of the control modules may run on the controller device 100 and some may run in the cloud.

[0094] For example, control modules executing on device 100 may include those involving real-time responses, such as sudden changes in zone occupancy, while control modules executing on remote server 300 may include those not related to real-time responses, such as detecting trends and correlations between events and power demand, or applying weather forecasts to create power demand forecasts. In other words, in some embodiments, real-time operations are preferably performed locally and analysis is preferably performed remotely. Data privacy may also be taken into consideration when determining whether to communicate data to remote server 300 over the Internet.

[0095] In one embodiment of the present invention, the control modules related to the dynamic heat balance process may be divided into various subgroups related to their functionality, for example, the control modules may be broadly divided into modules that specifically control the HVAC air components, modules that specifically control the water side components, modules that increase energy efficiency and / or reduce energy costs, and modules that optimize the control of the HVAC system. For example, in some embodiments, subgroups of modules may be specifically directed to particular HVAC components related to one or more of the following: Air systems that manage the ventilation, temperature, and humidity of the air flowing through the building A cooling system that generates and distributes the cooling heat load within a building, often supplying various air systems. Heating systems that generate and distribute the heating load within a building, often supplying various air systems Control system.

[0096] Each of these subgroups has its own set of algorithms that model the heat flow in a building and calculate the optimal running configuration of various HVAC devices. Below is an exemplary list of modules or algorithms for each subgroup:

[0097] With respect to the air system, the controller device 100 may configure and execute control modules to perform the following operations (without being limited to): Adjust fresh air intake based on the number of people in each zone of the building Adjusts fresh air intake based on outside temperature levels Adjusting airflow speed based on the number of people in each zone of the building Adjust the pressure in each zone Adjust the humidity control you need Adjusts duct static pressure based on fan speed and VAV adjustment Adjust supply air fans, hood exhaust fans, and make-up fans based on occupancy levels in each zone Adjusting the economy cycle (using outdoor air to cool the space) based on external parameters and internal load Adjusting the night purge (bringing cool outdoor air into the building at night to avoid mechanical cooling during start-up) based on weather parameters and forecast demand Adjusting the air velocity in the computer room based on the cooling load (using air movement to remove heat load instead of supplying cool air to the room) Adjusts humidification and dehumidification based on load.

[0098] With respect to a cooling system, the controller device 100 may configure and execute control modules to perform the following operations (without being limited to): Adjust chiller and compressor staging based on load balancing along the optimal chiller curve Adjust chilled water pump speed based on the number of people in the building Adjusting the water evaporator pump in the water tower based on the number of people in the building Adjusts chilled water temperature and flow rate based on number of people, outside temperature, and wet bulb temperature Adjusts the condenser water temperature to maintain the coldest possible CW for optimal cooling of the chiller Adjusts head pressure control to maintain optimal condenser fan energy consumption (only for air-cooled condensers and water towers) Adjust the condensing pressure based on the heat load being extracted (for water-cooled condensers) Adjust chiller and compressor staging based on load balancing along the optimal chiller curve Adjust chilled water pump speed based on the number of people in the building Adjusting the water evaporator pump in the water tower based on the number of people in the building Adjusts chilled water temperature and flow rate based on number of people, outside temperature, and wet bulb temperature Adjusts the condenser water temperature to maintain the coldest possible CW for optimal cooling of the chiller Adjusts head pressure control to maintain optimal condenser fan energy consumption (only for air-cooled condensers and water towers) · Adjust the condensing pressure based on the heat load to be extracted (for water-cooled condensers).

[0099] With respect to a heating system, the controller device 100 may configure and execute control modules to perform the following operations (without being limited to): Adjust hot water / steam delivery temperature based on heating demand load Adjust boiler sequencing based on load demand - Adjusting losses of a powered-down boiler ·Adjust steam pressure based on load demand.

[0100] With respect to the control system, the controller device 100 may configure and execute control modules to perform the following operations (including but not limited to): Adjusts space temperature setpoints and control bands based on optimal range trend and deadband (even with drift).

[0101] Referring now to FIG. 2 , the controller device 100, by executing various control modules, can generate control commands and parameters for operating the following HVAC components 400 within the building 170, without requiring any physical modification of the building 170: A. Optimization of cooling production 105 Variable head pressure control (air-cooled condensers and water towers) Adjust condensing pressure based on the heat load being extracted Variable head pressure control (water-cooled condenser) Adjust the condenser head pressure based on the heat load to be extracted Secondary chilled water pump optimization Balances the amount of chilled water circulated based on the needs of the Al-IU (reducing circulation during periods of low demand) Humidity control based on outside humidity level · Predictive cooling production based on weather forecast (6 hours) Balance chiller temperatures based on evaporator targets B. Distribution Optimization 120 Reset the cold water delivery temperature Set the warmest possible water temperature for optimal cooling Reset the condenser water temperature Set the coolest possible CW that is optimal for cooling the cooling device Reset the heating hot water delivery temperature Set the coldest possible water temperature, perfect for heating Heating water Delta T adjusted with pressure Delta P Chiller and compressor staging -Balancing chiller loads with optimal efficiency curves C. Ventilation and Airflow Optimization 130 Intake fans, hood exhaust fans, and make-up fans Reduce usage when not necessary (depending on room occupancy) Use outdoor air to cool the space when possible Nightly purge Demand-controlled ventilation based on CO2 control in occupied spaces Pre-cooling or heating based on the position of the sun Pre-cooling or heating based on the number of people in the zone Demand-controlled ventilation based on CO level control ·Resetting duct static pressure Reduces fan speed proportionally to VAV adjustment Uses air movement to remove heat load instead of supplying cool air to the room D. System Control Optimization 140 ·Occupancy control -Ventilation systems and lighting will automatically switch on when occupancy is detected in the area Optimal start / stop of heating / cooling in each zone Space temperature setpoints and control bands Set optimal range trend and deadband Master air handling unit supply air temperature signal Adjust master air delivery temperature to avoid simultaneous heating ·cooling.

[0102] Examples of sensors 200 provided within building 170 include, but are not limited to, sensors within HVAC systems such as temperature sensors (often within a thermostat), occupancy sensors, humidity sensors, pressure sensors, and air velocity sensors.

[0103] Other data not measured by the sensors 200 can be collected and used. For example, an online calendar can be queried to detect upcoming events at a particular location in the building. In another example, the remote server 300 can query a third-party server 600 to collect weather forecasts for the building's location, as shown in FIGS. 1A and 1B, and feed them into a thermodynamic model of the building being calculated by the remote server 300, taking into account sunlight, temperature, wind, humidity, pressure, and precipitation in the environment around the building. For example, a given zone in the building may receive more sunlight than another zone, and the remote server 300 may determine that the optimal strategy is to trigger ventilation to move air from one zone to another, warming the other zones without relying on a heating system and cooling the sunlit zone without requiring an air conditioning system.

[0104] 4 is a schematic diagram illustrating an example architecture of a system 480 including a controller device 100 for dynamic control of HVAC components 400 of a building 170, according to at least one embodiment. In some embodiments, the controller device 100 is configured to access and use a stored temperature model 110 to determine a predicted temperature for a given zone of the building 170.

[0105] In some embodiments, the rate of change of temperature is generated by controller device 100 based on a generic ODE, where some elements of the generic ODE are modeled using physics-based equations or representations and other elements are modeled using a function approximator such as a neural network. In some cases, the decision of whether to use physics-based equations or function approximators to represent one or more elements that affect the rate of change of temperature may be based on the amount of training data needed to train the function approximator to adequately model the elements. As an example, a physics-based representation, e.g.,

number

[0106] In some embodiments, the rate of change of temperature is generated by the controller device 100 based on a generalized ODE, non-HVAC components are explicitly defined based on RC model-based differential equations, and the HVAC contribution from the HVAC components 400 is modeled via a neural network. As shown in FIG. 5, the neural network 500 may have an architecture 500 including an input layer, multiple hidden layers, and an output layer. The input layer receives input features. The hidden layers map the input layer to the output layer. The output layer provides the neural network's predictions. In other embodiments, other suitable deep learning models and architectures may be used.

[0107] For example, the temperature prediction may be based on a temperature model 110 stored in a database connected to the controller device 100. Alternatively, the temperature model 110 may be stored in the database server 450 or as part of the BMS 150. The temperature model 110 may, in some embodiments, take the form of an ODE that models the rate of change of temperature in a given zone i of a building having multiple zones.

[0108] In some embodiments, a modular software framework may be implemented by the system for controlling the controller device 110. For example, various portions of the temperature model 110 (which may be a general-purpose ODE) may be implemented as separate software modules that can be dynamically swapped before or during operation. For example, after a certain time after an individual differential equation from the model 110 used to simulate a particular physical property of the environment (e.g., solar radiation) has accumulated a sufficient amount of raw data or reached a certain threshold, the individual differential equation used to simulate the physical property may be swapped with a corresponding function approximator element, which may be a neural network that can be trained fairly quickly based on the accumulated raw data, to better predict the physical property at future times.

[0109] The temperature model 110 includes an explicitly defined model portion and an approximate model portion, where the approximate model portion is modeled based on a neural network trained to generate an output that reflects the amount of heat added to or removed from at least one zone.

[0110] In some embodiments, the approximate model portion can be modeled using one or more function approximators, which can include a class of generic functions that are parameterized to represent various input-output mappings. The function approximators can be neural networks 500, as shown in FIG. 5.

[0111] In some embodiments, the output from the approximation model portion may also include a value reflecting the total amount of temperature change over a given period of time based on the current setpoint.

[0112] In some embodiments, the output from the approximation model portion may include values ​​that reflect solar radiation.

[0113] For example, the temperature model 110 may include an equation for determining the temperature change rate of at least one zone i:

number

number

[0114] In equation [1]

number

[0115] In another embodiment, the term

number

number

[0116] Equation [2] is parameterized by a neural network NN(θ,U,Solrad,..). U represents the vector of HVAC controls, SolRad is the solar radiation data, θ are the parameters of the function approximator (e.g., neural network), and other inputs to the function approximator can also be included. i (U,SolRad,..) models the environmental thermal contribution in zone i.

[0117] In some embodiments, U may represent a vector including one or more HVAC controls including values ​​reflecting the state of one or more HVAC components, such as, but not limited to, a temperature setpoint, an on / off switch, a fan, a heating coil, a cooling coil, a valve position, and / or a damper position.

[0118] The solar radiation data SolRad can be obtained from an external database, which can be a pyranometer or a third party database such as climacell, darksky, etc.

[0119] Other inputs to the neural network may include salient features such as temperature schedules, sun position, additional measurements of solar radiation data, etc. Additionally, further inputs to the neural network may include occupancy data from CO2 sensors, people counts from the building 170 access system or Wi-Fi system.

[0120] In some embodiments, the portion of the model approximated by the neural network may include other environmental heat contributing sources instead of or in combination with solar radiation.

[0121] For each zone in a multi-zone building 170, there is one such equation for each zone i, i=[1, 2, ... n], and all equations can be solved together as a system of equations. The index [i] in the right brackets indicates that the equations share a single neural network with multiple (e.g., n) outputs, each representing the temperature change rate for a respective zone i. For clarity, a zone in a building 170 may refer to a room or a group of interconnected rooms.

[0122] In the above equation [2], the following quantity τ i , τ ij , θ are unknown parameters that are learned from input data by the neural network 500. While θ is a parameter of the neural network, τ iand τ ij have physical meaning and their values ​​can be extracted to gain insight into the thermal properties of one or more zones of the building 170.

[0123] By including neural network 500 in differential equations [1] or [2], neural network 500 is asked to solve the practical problem, "Given the setpoint of the HVAC system, what is the contribution of the HVAC system to the rate of change of temperature in a given zone?" This is a much simpler and more structured problem than feeding all available data directly into the neural network and asking, "What will the temperature be at a discrete time in the future?", which leads to a much more efficient machine learning model that requires substantially less training time.

[0124] Our hybrid approach, which includes an explicitly defined model portion and an approximate model portion, can reduce the data required to train a neural network (potentially from months of data to days of data), improve the generalization of machine learning models, increase performance, and enhance human interpretability.

[0125] In some embodiments, explicitly defined portions can be used to model physical properties or characteristics of a room or zone when the relevant phenomenon or behavior occurs in a manner that facilitates explicit definition based on physical modeling and at least some inputs to the explicitly defined portions are available, whereas approximate model portions can be used when it is difficult to model the physical properties or characteristics of a room or zone based on existing physical behavior, or when it is more efficient to use a function approximator to model the physical properties or characteristics, for example, when a large amount of historical data exists to train a function approximator to simulate the physical properties or characteristics.

[0126] The reduced data requirements allow implementation on devices with limited computational resources for training the neural network (e.g., no GPUs), such as thermostats or other devices located in each zone (e.g., room).

[0127] In some embodiments, the thermal model 110 also provides a mechanism for extracting physically meaningful parameters from the sensor data to quantify the thermal properties of one or more zones directly from the sensor data.

[0128] Based on the rate of change of temperature generated based on the temperature model 110, the controller device 100 may determine a predicted temperature for one or more zones of the building 170.

[0129] For example, if the current temperature of a zone is currently measured at x degrees and the rate of temperature change is y degrees / minute, then the predicted temperature of the zone 2 hours from now is z=x+y * It will be 60 minutes.

[0130] In some embodiments, the temperature rate of change represented by equations [1] or [2] may be in the continuous time domain if dt is small enough, in which case equations [1] or [2] can be used to estimate the temperature at any time.

[0131] The controller device 100 can generate a command signal for controlling at least one setpoint of the HVAC system in response to the predicted temperature.

[0132] In some embodiments, generating the command signal may include generating one or more operating values ​​for the HVAC component 400 based on the predicted temperature and transmitting the one or more operating values ​​to the HVAC component.

[0133] In some embodiments, inputs for generating one or more command signals may include one or more of predicted space temperature forecasts for each zone, virtual electric or gas meter values, HVAC control inputs and outputs (from BMS 150), occupancy schedules (from BMS 150 or a third-party database), grid GHG intensity data (from an API or a third-party database), fee schedules (from an API or a third-party database), energy billing and consumption data (from an API or a third-party database), mechanical design information (from BMS 150 or engineer review), and occupancy status (from BMS 150 or sensors 200). Outputs from controller device 100 may include rooftop unit (RTU) and / or variable air volume (VAV) commands. RTU commands may include, for example, fan start / stop, fan speed, heating stage, cooling stage, and economizer damper. VAV commands may include, for example, space setpoint, fan start / stop, reheat adjustment, and damper adjustment.

[0134] The system 480 may include a database server 450 connected to the controller device 100. The database server 450 is connected to a dashboard display device 420 and a monitoring system 460. The controller device 100 can implement related control algorithms to monitor various parameters across the HVAC components 400 and sensors 200 in the building 170 via the monitoring system 460, providing continuous automatic adjustments to the system based on building load, regardless of equipment type. The goal of these systems and methods is to provide energy usage savings (kWh / year), demand savings (kW), heating / cooling load (therms), cooling tower water usage savings (gallons / year), carbon footprint reduction (lbs / year), and power usage effectiveness (PUE) reduction. In doing so, these systems and methods save significant amounts of money, reduce strain on the power grid, and improve tenant comfort.

[0135] In some embodiments, the control device 100 is configured to simulate the performance of the BMS system 150 (e.g., temperature / humidity, demand levels, emissions, etc., for a particular space) during multiple rounds of execution of one or more control modules. During each round, various combinations and sequences of control modules can be executed. In this manner, each "path to a desired outcome" can be evaluated, and the optimal path that brings the BMS closest to the client's desired outcome can be selected. The simulations are analyzed and compared by the control device 100, which can implement a powerful prediction engine that takes inputs (e.g., weather forecast for the next few hours), occupancy rates, emissions data, etc., to predict the performance of a particular path.

[0136] An example energy saving control module can be configured to save as much energy as possible by managing optimal start and stop times for HVAC components 400 within an HVAC system. The energy saving control module can use temperature forecasts, which may be generated by a separate temperature forecast module, and current outdoor temperature and building conditions, which may be transmitted from the BMS 150, to determine when it is optimal to start and stop one or more HVAC components 400.

[0137] Another example control module may be configured to adjust chiller or boiler delivery temperatures depending on predicted outdoor air temperatures, predicted building heat loads that may be generated by a separate heat load prediction module, and current outdoor air temperatures that may be transmitted from a third-party weather database or sensor 200. The control module may be configured to determine one or more set points for the chiller, boiler, or variable speed fan cooling tower.

[0138] The controller device 100 can be configured and installed to directly interface with the HVAC components 400, or when the controller device 100 is installed in an existing building, the controller device 100 can interface with an existing control system, such as the BMS 150, with existing control loops for the HVAC components 400. In the latter situation, the existing control loops are included in the simulation described above. When use of the existing control loops is predicted to be part of an optimal path, the controller device 100 can defer control of the HVAC components 400 to the existing control loops of the BMS 150. In some examples, control is deferred until it is no longer optimal. In other examples, the controller device 100 can take over control of the HVAC components 400, in which case it can completely overwrite or override the previous control loops of the BMS 150.

[0139] The controller device 100 can also be configured to allow clients (e.g., enterprise users or building managers) to create their own custom logic. A client may have multiple versions of the BMS system in their building, each with its own software, firmware, or control loops. Making changes to existing operating sequences can be tedious, costly, and vendor-dependent. The controller device 100 provides clients with a standard programming and vendor-independent platform to write their own proprietary logic and push it down to the underlying building BMS 150.

[0140] The system may include a display device 420, such as a dashboard, that may be configured to display multiple data items such as current and predicted temperatures, the control modules or temperature models installed in each BMS 150, and other metrics.

[0141] Leveraging the accuracy of various predictive models, such as temperature predictive models, the controller device 100 calculates a series of possible trajectories and returns an optimal combination with respect to one or more particular user goals. In some embodiments, the control device 100 can leverage machine learning models to drive control directly, for example, by using a reinforcement learning agent.

[0142] Buildings are typically designed with oversized HVAC equipment and programmed to overcool or overheat when high or low temperatures occur. In some embodiments, when detected by the controller device 100, a control module can be deployed depending on the building's HVAC equipment and operating parameters to minimize energy consumption and costs while maintaining / improving comfort (adherence to a setpoint, e.g., 21 degrees Celsius). This can be accomplished via a predictive control module by predicting that the HVAC system will need to activate heating or cooling in the morning based on today's weather forecast (e.g., at 6:30 AM for occupancy beginning at 7:00 AM, rather than the statically programmed 6:00 AM HVAC schedule). However, if the predicted temperature is too high or too low, the AI-based control module can relinquish the building back to the BMS 150 so that it can power on as soon as possible to meet the expected cooling or heating demand.

[0143] One or more inputs may be obtained or received from the building's BMS 150 or sensors 200, including, for example, actual temperature, mechanical equipment values, number of zones, and heat leaks. One or more inputs may be predicted values ​​obtained from a third party, such as an outside temperature forecast. One or more inputs, such as temperature fluctuations, may be predicted value(s) from one or more control modules configured to predict such values.

[0144] For example, one or more machine learning-based control modules may be implemented based on one or more machine learning models such as long-short-term memory, gated recurrent units, or multi-layer perceptrons.

[0145] One or more inputs may be obtained or received from the building's BMS 150 or sensors 200, including, for example, outside temperature, cloud cover, sun position, humidity, pollution data, zone number, and equipment associations. The one or more inputs may be forecast values ​​obtained from a third party, such as an outside temperature forecast.

[0146] In some embodiments, the controller device 100 can send commands directly to HVAC BMS output points (e.g., binary output, analog output) to dynamically control and manage the HVAC components 400.

[0147] FIG. 6 is an example process 650 performed by the controller device 100 for dynamically managing the HVAC components 400 of a building 170 according to one embodiment.

[0148] In step 602, the controller device 100 stores or accesses a temperature model 110 for calculating a temperature change rate for at least one of the zones of the building 170, the model 110 including an explicitly defined model portion and an approximate model portion.

[0149] In some embodiments, the approximate model portion is solved using a function approximator configured to generate an output that reflects the amount of heat added to or removed from at least one zone.

[0150] In some embodiments, the function approximator includes a neural network 500 that is trained to generate at least a portion of the output.

[0151] In some embodiments, the output from the approximation model portion also reflects the total amount of temperature change over a given period based on the current setpoint.

[0152] In some embodiments, the output from the approximation model portion further reflects environmental thermal contributions.

[0153] In some embodiments, the environmental thermal contribution includes solar radiation.

[0154] At step 604, the controller device 100 receives sensor data from multiple temperature sensors 200 within the building 170. The sensor data may include, for example, pressure sensors, temperature sensors, humidity sensors, light sensors, motion sensors, indoor air quality sensors, occupancy sensors, etc.

[0155] At step 606, the controller device 100 receives outdoor temperature data, which may be obtained from one or more external databases, such as a local weather station or weather database.

[0156] In step 608, the controller device 100 calculates the rate of temperature change for at least one zone by providing the sensor data and outdoor temperature data to a temperature model, where the approximate model portion is solved using a function approximator, such as a neural network, trained to generate an output reflecting the amount of heat added to or removed from the at least one zone.

[0157] In some embodiments, the temperature ramp rate of at least one zone i is:

number

number

[0158] In some embodiments,

number

[0159] In some embodiments, the temperature ramp rate of at least one zone i is:

number

[0160] In some embodiments, f i (U,SolRad,..)=NN(θ,U,Solrad,..)[i], where NN represents a function approximator of the approximate model part, U represents a vector of one or more HVAC controls of the HVAC system, SolRad is given data reflecting the contribution of solar radiation, and θ represents one or more parameters of the function approximator NN.

[0161] In some embodiments, formula f i (U, SolRad, ..) is determined for each zone in the plurality of zones.

[0162] In step 610, the controller device 100 determines a predicted temperature based on the temperature rate of change.

[0163] For example, if the current temperature of a zone is currently measured at X degrees and the rate of temperature change is Y degrees per unit of time (e.g., minute), the predicted temperature of the zone two hours from now is X+Y * It will be 60 minutes.

[0164] In some embodiments, the temperature rate of change represented by equations [1] or [2] may be in the continuous time domain if dt is small enough, in which case equations [1] or [2] can be used to estimate the temperature at any time.

[0165] In step 612, the controller device 100 generates a command signal for controlling at least one setpoint of the HVAC system in response to the predicted temperature.

[0166] In some embodiments, generating the command signal may include generating one or more operating values ​​for the HVAC component 400 based on the predicted temperature and transmitting the one or more operating values ​​to the HVAC component.

[0167] FIG. 7 is a schematic diagram of a computing device 700 that may be used to implement the controller device 100, according to one embodiment.

[0168] As shown, computing device 700 includes at least one processor 702 , memory 704 , at least one I / O interface 706 , and at least one network interface 708 .

[0169] Each processor 702 may be, for example, any type of general-purpose microprocessor or microcontroller, a digital signal processing (DSP) processor, an integrated circuit, a field programmable gate array (FPGA), a reconfigurable processor, a programmable read-only memory (PROM), or any combination thereof.

[0170] The memory 704 may include any suitable combination of any type of computer memory, whether internally or externally located, such as, for example, random access memory (RAM), read-only memory (ROM), compact disk read-only memory (CD-ROM), electro-optical memory, magneto-optical memory, erasable programmable read-only memory (EPROM), and electrically erasable programmable read-only memory (EEPROM), ferroelectric RAM (FRAM), etc.

[0171] Each I / O interface 706 allows computing device 700 to interconnect with one or more input devices, such as a keyboard, mouse, camera, touchscreen, and microphone, or one or more output devices, such as a display screen and speakers.

[0172] Each network interface 708 connects to a network (or multiple networks) capable of transmitting data, including the Internet, Ethernet, a plain old telephone service (POTS) line, a public switched telephone network (PSTN), an integrated services digital network (ISDN), a digital subscriber line (DSL), coaxial cable, optical fiber, satellite, mobile (e.g., 4G, 5G networks), wireless (e.g., Wi-Fi, WiMAX), an SS7 signaling network, a landline, a local area network, a wide area network, etc., and any combination thereof, thereby enabling computing device 700 to communicate with other components, exchange data with other components, access and connect to network resources, host applications, and run other computing applications.

[0173] For simplicity, one computing device 700 is shown, but system 100 may include multiple computing devices 700. The computing devices 700 may be the same type of device or different types of devices. The computing devices 700 may be connected in a variety of ways, including directly coupled, indirectly coupled via a network, and distributed over a wide geographic area and connected via a network (sometimes referred to as "cloud computing").

[0174] For example, without limitation, computing device 700 may be a server, a network appliance, a set-top box, an embedded device, a computer expansion module, a personal computer, a laptop, a personal data assistant, a cellular telephone, a smartphone device, a UMPC tablet, a video display terminal, a gaming console, or any other computing device that can be configured to perform the methods described herein.

[0175] In some embodiments, the computing device 700 may be adapted to function as the controller device 100 .

[0176] The description herein provides many exemplary embodiments of the inventive subject matter. While each embodiment represents a single combination of inventive elements, the inventive subject matter is considered to include all possible combinations of the disclosed elements. Thus, if one embodiment comprises elements A, B, and C, and a second embodiment comprises elements B and D, the inventive subject matter is considered to include any remaining combination of A, B, C, or D, even if not explicitly disclosed.

[0177] The device, system, and method embodiments described herein can be implemented in a combination of both hardware and software. These embodiments can be implemented on programmable computers, each computer including at least one processor, a data storage system (including volatile or non-volatile memory or other data storage elements or a combination thereof), and at least one communication interface.

[0178] Program code is applied to input data to perform the functions described herein and to generate output information. The output information is applied to one or more output devices. In some embodiments, the communication interface may be a network communication interface. In embodiments where elements may be combined, the communication interface may be a software communication interface, such as for inter-process communication. In still other embodiments, there may be a combination of communication interfaces implemented as hardware, software, and combinations thereof.

[0179] Throughout this description, numerous references are made to servers, services, interfaces, portals, platforms, or other systems formed from computing devices. It should be understood that use of such terms is intended to refer to one or more computing devices having at least one processor configured to execute software instructions stored on a tangible, non-transitory computer-readable medium. For example, a server may include one or more computers operating as a web server, database server, or other type of computer server in a manner that performs the described roles, responsibilities, or functions.

[0180] The technical solutions of the embodiments may be in the form of a software product. The software product may be stored in a non-volatile or non-transitory storage medium, which may be a compact disk read-only memory (CD-ROM), a USB flash disk, or a removable hard disk. The software product includes some instructions that enable a computer device (personal computer, server, or network device) to execute the method provided by the embodiments.

[0181] The embodiments described herein are implemented by physical computer hardware, including computing devices, servers, receivers, transmitters, processors, memory, displays, and networks. The embodiments described herein provide useful physical machines and specially configured computer hardware configurations.

[0182] Of course, the above-described embodiments are intended to be illustrative only and not limiting in any way. The described embodiments are susceptible to many modifications of form, arrangement of parts, details, and order of operation. The present disclosure is intended to encompass all such modifications within its scope as defined by the claims.

Claims

1. 1. A computer-implemented system for controlling a heating, ventilation, or air conditioning (HVAC) system in a building having multiple zones, the system comprising: a processor; 1. A non-transitory memory device, comprising: a temperature model for calculating a rate of change of temperature of at least one of the building's zones, the model including an explicitly defined model portion and an approximate model portion; A set of computer-readable instructions that, when executed, cause the processor to: receiving sensor data from a plurality of temperature sensors within the building; Receive outdoor temperature data, calculating a rate of temperature change of at least one zone by providing the sensor data and the outdoor temperature data to the temperature model, the approximation model portion being solved using a function approximator configured to generate an output reflecting an amount of heat added to or removed from at least one zone; determining a predicted temperature based on the temperature change rate; generating a command signal for controlling at least one setpoint of the HVAC system in response to the predicted temperature; a set of instructions that a non-transitory memory device storing the A system comprising:

2. The system of claim 1 , wherein the function approximator comprises a neural network trained to generate at least a portion of the output.

3. The system of claim 1 , wherein the output from the approximation model portion further reflects a total amount of temperature change over a given period based on a current setpoint.

4. The system of claim 1 , wherein the output from the approximate model portion further reflects environmental thermal contributions.

5. The system of claim 4 , wherein the environmental thermal contribution comprises solar radiation.

6. The temperature change rate of the at least one zone i is: [Equation 1] where T ｉ is the temperature of each zone i obtained from the sensor data, and T ｅｎｖ is the outdoor temperature data, and τ ｉ is the time scale associated with the thermal decay of zone i, and τ ｉｊ is the time scale representing the heat exchange between zone i and adjacent zone j, and C ＨＶＡＣ is the heat capacity of zone i, [Equation 2] 4. The system of claim 3, wherein x is the amount of heat added to or removed from at least one zone i. 【Request 7】 【Number 3】 The system of claim 6 , wherein ∑ m represents a total amount of temperature change calculated in the approximate model portion.

8. The temperature change rate of at least one zone i is [Equation 4] where T ｉ is the temperature of each zone i obtained from the sensor data, and T ｅｎｖ is the outdoor temperature data, and τ ｉ is the time scale associated with the thermal decay of zone i, and τ ｉｊ is the time scale representing the heat exchange between zone i and adjacent zone j, and f ｉ The system of claim 1 , wherein (U, SolRad, ..) models the environmental thermal contribution in zone i.

9. f ｉ 9. The system of claim 8, wherein (U, SolRad, ..) = NN(θ, U, SolRad, ..)[i], NN represents a function approximator of the approximate model portion, U represents a vector of one or more HVAC controls of the HVAC system, SolRad is given data reflecting the contribution of solar radiation, and θ represents one or more parameters of the function approximator NN.

10. The formula f ｉ The system of claim 9 , wherein (U, SolRad, ..) is determined for each zone in the plurality of zones.

11. 1. A computer-implemented method for controlling a heating, ventilation, or air conditioning (HVAC) system in a building having multiple zones, the method comprising: storing or accessing a temperature model for calculating a temperature change rate of at least one of the building's zones, the model including an explicitly defined model portion and an approximate model portion; receiving sensor data from a plurality of temperature sensors within a building; receiving outdoor temperature data; calculating a rate of change of temperature of at least one zone by providing sensor data and outdoor temperature data to a thermal model, the approximation model portion being solved using a function approximator configured to generate an output reflecting an amount of heat added to or removed from the at least one zone; determining a predicted temperature based on the temperature change rate; generating a command signal for controlling at least one setpoint of the HVAC system in response to the predicted temperature; A method comprising:

12. The method of claim 11 , wherein the function approximator comprises a neural network trained to generate at least a portion of the output.

13. The method of claim 11 , wherein the output from the approximation model portion further reflects the total amount of temperature change over a given period based on a current setpoint.

14. The method of claim 11 , wherein the output from the approximate model portion further reflects environmental thermal contributions.

15. The method of claim 14 , wherein the environmental thermal contribution comprises solar radiation.

16. The temperature change rate of the at least one zone i is: [Equation 5] where T ｉ is the temperature of each zone i obtained from the sensor data, and T ｅｎｖ is the outdoor temperature data, and τ ｉ is the time scale associated with the thermal decay of zone i, and τ ｉｊ is the time scale representing the heat exchange between zone i and adjacent zone j, and C ＨＶＡＣ is related to the heat capacity of zone i, [Equation 6] 14. The method of claim 13, wherein x is the amount of heat added to or removed from at least one zone i. 【Request 17】 【Number 7】 The method of claim 16 , wherein ∇ represents the total amount of temperature change calculated by the approximation model portion.

18. The temperature change rate of at least one zone i is [Equation 8] where T ｉ is the temperature of each zone i obtained from the sensor data, and T ｅｎｖ is the outdoor temperature data, and τ ｉ is the time scale associated with the thermal decay of zone i, and τ ｉｊ is the time scale representing the heat exchange between zone i and adjacent zone j, and f ｉ The method of claim 11 , wherein (U, SolRad, ..) models the environmental thermal contribution in zone i.

19. f ｉ 19. The method of claim 18, wherein (U, SolRad, ..) = NN(θ, U, SolRad, ..)[i], NN represents a function approximator of the approximate model portion, U represents a vector of one or more HVAC controls of the HVAC system, SolRad is given data reflecting the contribution of solar radiation, and θ represents one or more parameters of the function approximator NN.

20. The formula f ｉ 20. The method of claim 19, wherein (U, SolRad, ..) is determined for each zone in a plurality of zones.

Images