Model prediction based building environmental equipment control method

Abstract

The application relates to the field of building equipment control and discloses a building environment equipment control method based on model prediction. The application comprises the following steps: collecting real-time data and constructing a state space vector according to the real-time data; constructing a search space according to the state space vector and the collaborative relationship weight, the conflict relationship weight and the predicted energy consumption of a plurality of equipment combinations; the priority of the equipment combinations in the search space is positively correlated with the collaborative relationship weight and is negatively correlated with the conflict relationship weight and the predicted energy consumption; a plurality of candidate control sequences oriented to a prediction time window are generated in the search space; the candidate control sequences are subjected to thermal verification based on the building thermal characteristics; a target control sequence is determined from the candidate control sequences that pass the thermal verification, and the target control sequence is issued to corresponding equipment for execution. The building equipment control effect is improved.

Description

Technical Field

[0001] This invention relates to the field of building equipment control, and in particular to a model-based prediction method for controlling building environment equipment. Background Technology

[0002] Building environmental control involves the operation and regulation of various devices such as air conditioning, fresh air systems, underfloor heating, humidity control equipment, shading devices, and electric windows to maintain indoor environmental parameters such as temperature, humidity, and air quality within target ranges. Related technologies typically adjust equipment based on feedback from deviations between indoor environmental parameters and target values, or adjust equipment operating parameters based on load forecasting results.

[0003] As the types and quantities of environmental control equipment in buildings continue to increase, the above-mentioned control methods have the following problems in actual operation: when multiple devices operate simultaneously, their effects may cancel each other out, such as cooling and heating or dehumidification and humidification occurring simultaneously, resulting in energy waste; the large number of devices that can be called upon and the variety of combinations lead to high computational overhead and low decision-making efficiency in control decisions; the overall energy consumption of the equipment is relatively high; and the control strategy may deviate from the actual thermal behavior of the building during actual execution, resulting in large fluctuations in indoor environmental parameters and unstable control effects. Summary of the Invention

[0004] The purpose of this invention is to provide a model-based prediction-based building environment equipment control method to solve the problems in related technologies, such as energy waste caused by the mutual cancellation of the effects of multiple devices operating simultaneously, large computational overhead for control decisions, high overall operating energy consumption, and fluctuations in indoor environmental parameters caused by control strategies deviating from building thermal behavior.

[0005] To address the aforementioned technical problems, this invention provides a model-based prediction-based building environment equipment control method, comprising: collecting real-time data and constructing a state space vector based on the real-time data; constructing a search space based on the state space vector and the respective collaborative relationship weights, conflict relationship weights, and expected energy consumption of multiple equipment combinations; the priority of an equipment combination in the search space is positively correlated with the collaborative relationship weights and negatively correlated with the conflict relationship weights and expected energy consumption; generating multiple candidate control sequences oriented towards the prediction time window within the search space; performing thermal verification on the candidate control sequences based on building thermal characteristics; determining the target control sequence from the candidate control sequences that have passed the thermal verification, and distributing the target control sequence to the corresponding equipment for execution.

[0006] The present invention also provides an electronic device, comprising: a processor; and a memory storing a computer program, wherein the computer program, when executed by the processor, implements the above-described model-based prediction-based building environment equipment control method.

[0007] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described model-based prediction-based building environment equipment control method.

[0008] In this invention, real-time data collected is uniformly encoded into control decision-making basis through state space vectors. When constructing the search space, the priority of each equipment combination is determined based on the weights of its cooperative relationship, conflict relationship, and expected energy consumption. This prioritizes equipment combinations with high cooperativeness, low conflict, and low expected energy consumption in sequence generation, eliminating the need to exhaustively list all equipment combinations for candidate control sequences. After verification by building thermal characteristics, the target control sequence is determined and executed. As a result, equipment combinations with mutually canceling effects are suppressed at the decision-making stage, reducing energy waste; the computational overhead of control decisions is reduced; the use of high-energy-consuming equipment combinations is constrained, resulting in lower overall operating energy consumption; and the control arrangements executed conform to building thermal behavior, leading to more stable indoor environmental parameters. Attached Figure Description

[0009] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0010] Figure 1 A flowchart illustrating a model-predictive-based building environment equipment control method provided in one embodiment of the present invention; Figure 2 A schematic diagram of the overall architecture of a model-predictive-based building environment equipment control method provided in one embodiment of the present invention; Figure 3 A schematic diagram illustrating the composition of a state space vector according to an embodiment of the present invention; Figure 4 A schematic diagram illustrating the process of constructing a search space according to an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the process of determining a target control sequence and issuing it for execution, according to one embodiment of the present invention. Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation

[0011] Building environmental control systems typically require the operation and regulation of various devices, such as air conditioning, fresh air systems, underfloor heating, humidity control equipment, shading devices, and motorized windows, to maintain indoor environmental parameters such as temperature, humidity, and air quality within target ranges. Traditional building environmental control solutions generally employ feedback control or predictive control: feedback control adjusts the operating parameters of individual devices based on deviations between indoor temperature, humidity, or air quality and target values, such as proportional-integral-derivative control, rule-based control, or fuzzy control; predictive control adjusts the operating parameters of devices over a future period based on load forecasts or model predictions; some solutions also utilize building thermal inertia models to predict the building's thermal response process and optimize equipment operating strategies accordingly.

[0012] The implicit assumption of the above control scheme is that the adjustment processes of each device are independent of each other, and the optimal adjustment of a single device for its own control objective can be superimposed to achieve a good control effect for the entire building environment. This assumption is approximately valid in scenarios with few types of equipment and a single control objective.

[0013] However, analysis of building environments with multiple devices operating simultaneously reveals that environmental parameters such as indoor temperature, humidity, and air quality are coupled with each other, and the regulatory effects of different devices superimposed within the same physical field. Simultaneous operation of air conditioning and underfloor heating results in thermal offsetting, while simultaneous operation of dehumidification and humidification results in humidity offsetting. These offsetting effects cause the output of some devices to be mutually consumed, resulting in ineffective energy consumption.

[0014] Furthermore, buildings have significant thermal inertia, and the environmental parameter deviations caused by the offsetting effect often appear with a lag. This can easily be misinterpreted by the control system as an increase in load, leading to increased equipment output and exacerbating the mutual increase in energy consumption and fluctuations in environmental parameters.

[0015] The deeper mechanism lies in the fact that the operational relationships between devices, such as coordination, conflict, substitution, and stability, lack explicit expression in traditional control architectures, and control decisions can only be made at the level of single device parameters. As the types and number of devices increase, the decision space of device combinations expands dramatically, and it is computationally difficult to evaluate all device combinations one by one within the real-time control cycle. At the same time, comfort control and energy consumption budget are disconnected, and the control process is difficult to dynamically adjust under daily, weekly, monthly budgets or time-of-use electricity pricing conditions.

[0016] The essence of the aforementioned problems lies not in the adjustment precision of individual devices, but in the lack of a unified expression of equipment operating relationships, building thermal status, and energy consumption budget status in control decisions. Related technologies often employ reactive, post-event measures to address equipment conflicts. This means that adjustments are made after a conflict occurs and causes abnormal environmental parameters or energy consumption, relying on alarm prompts or manual intervention. It is difficult to proactively avoid mutually offsetting equipment combinations during the control decision generation phase.

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of the present invention to facilitate a better understanding of the invention. However, the technical solutions claimed in the present invention can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.

[0018] The terms "first" and "second" used in this invention are for descriptive purposes only, to distinguish similar objects, and should not be construed as indicating or implying relative importance, nor should they be construed as implicitly indicating the number or order of the indicated technical features.

[0019] In this invention, "multiple" means two or more; "at least one" means any one or any combination of the listed items. For example, "at least one of A, B and C" can mean: A alone, B alone, C alone, a combination of A and B, a combination of A and C, a combination of B and C, or a combination of A, B and C.

[0020] The model-based prediction-based building environment equipment control method provided in this invention can be applied to environmental control scenarios of various buildings such as office buildings, commercial complexes, hospitals, schools, data centers, and residences. It can be deployed in building automation systems, building energy management systems, or smart home control systems, and the control range can cover various levels of a single room, a single floor, the entire building, or even a building complex.

[0021] This invention identifies the lack of unified expression and equipment relationship modeling in the aforementioned background problems, and proposes a model-based prediction-based building environment equipment control method: Real-time collected data is uniformly encoded into a state-space vector. A search space is constructed based on the cooperative relationship weights, conflict relationship weights, and expected energy consumption of equipment combinations. Multiple candidate control sequences are generated within the search space, oriented towards the prediction time window. After verification by building thermal characteristics, the target control sequence is determined and executed. This scheme replaces post-generation step-by-step evaluation with a pre-generation combination priority ranking, maintaining the existing equipment and execution links in the control architecture and ensuring compatibility with existing building automation systems.

[0022] The equipment control method for building environments provided in this invention is an intelligent building environment control scheme: multi-source real-time data is uniformly encoded through state space vectors, enabling control decisions to be based on autonomous perception of the current operating status of the building environment; the coordination, conflict, substitution, and stability relationships between equipment are structurally expressed and reasoned through an equipment coordination relationship graph; candidate control sequences are autonomously generated and selected within a priority-organized search space, achieving autonomous collaborative decision-making among multiple equipment oriented towards the prediction time window; the execution results of the target control sequence are further fed back to continuously correct the relationship weights in the equipment coordination relationship graph, allowing the graph to continuously evolve with the actual operation of the building, and the control strategy to possess self-learning and adaptive capabilities. Thus, this scheme organizes perception, decision-making, execution, and feedback updates into a closed loop, enabling autonomous adjustment of control arrangements according to changes in building operating status, equipment operating relationships, and energy consumption budget status. Under the dual constraints of building thermal characteristics and energy consumption budget, it continuously outputs target control sequences that conform to the actual thermal behavior of the building, achieving intelligent control of the building environment.

[0023] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention. The following implementation details are provided only for ease of understanding and are not essential for implementing this solution.

[0024] Figure 1 A flowchart illustrating a model-predictive-based building environment equipment control method according to an embodiment of the present invention. See also... Figure 1 The method includes steps S100 to S500.

[0025] See Figure 2 , Figure 2 This is a schematic diagram of the overall architecture of a model-based prediction-based building environment equipment control method according to an embodiment of the present invention. The overall processing of this method can be divided into a digital twin layer, a state space construction layer, an equipment collaboration relationship graph layer, an equipment collaboration search space layer, a generative decision layer, a physical constraint correction layer, an energy consumption budget constraint layer, a collaborative execution layer, and a feedback update layer. These layers are sequentially connected: building environment data, meteorological data, equipment data, energy consumption data, and user feedback data are accessed through the state space construction layer; step S100 corresponds to the digital twin layer and the state space construction layer; step S200 corresponds to the equipment collaboration relationship graph layer and the equipment collaboration search space layer; step S300 corresponds to the generative decision layer; step S400 corresponds to the physical constraint correction layer; and step S500 corresponds to the energy consumption budget constraint layer, the collaborative execution layer, and the feedback update layer. The feedback update layer uses the execution results for graph weight updates and search space updates for use in the next control cycle.

[0026] In one example, the method can be executed by a control device with data processing capabilities, such as a building automation host, an edge computing gateway, a local server, or a cloud server. The present invention does not limit the specific form of the executing entity.

[0027] Step S100: Collect real-time data and construct a state space vector based on the real-time data.

[0028] Real-time data refers to multi-source data that reflects the current operating status of the building environment, which may include building environment data, equipment operation data, and energy consumption data. Optionally, real-time data may also include outdoor meteorological data and user feedback data. The types of data can be expanded according to the building's sensing and metering conditions. This invention does not limit the specific composition of real-time data.

[0029] In one example, building environment data includes indoor temperature, indoor humidity, carbon dioxide concentration, particulate matter concentration, etc.; outdoor meteorological data includes outdoor temperature, outdoor humidity, solar radiation intensity, etc.; equipment operation data includes equipment operating status, power status, load rate, response delay time, number of start-stop cycles, etc.; energy consumption data includes current cumulative energy consumption, energy consumption of each device, and historical energy consumption data. Among them, the energy consumption of each device can be collected by the metering circuit or smart meter of each device, and the historical energy consumption data is obtained by accumulating and storing energy consumption data over time; user feedback data includes user feedback information on comfort levels such as temperature, humidity, etc.

[0030] The state-space vector is a state representation used for control decision-making, obtained by uniformly encoding the aforementioned multi-source data. (See also...) Figure 3 In one example, the state space vector can be represented as S(t) = [Xe(t), Xt(t), Xd(t), Xc(t), Xp(t)], where S(t) is the state space vector at time t; Xe(t) is the environmental state vector, whose components may include indoor temperature, indoor humidity, carbon dioxide concentration, particulate matter concentration, outdoor temperature, outdoor humidity, solar radiation intensity, etc.; Xt(t) is the thermal state vector, whose components may include building thermal inertia, thermal equilibrium state, spatial thermal coupling relationship, thermal response time constant, etc.; Xd(t) is the equipment state vector, whose components may include equipment operating status, power status, load rate, response delay time, number of start-stop cycles, etc.; Xc(t) is the equipment coordination state vector; and Xp(t) is the energy consumption state vector, whose components may include current cumulative energy consumption, daily budget remaining value, weekly budget remaining value, monthly budget remaining value, real-time electricity price, time-of-use electricity price, etc. All of the above components can be normalized and then encoded into the state space vector.

[0031] The composition of the state space vector can be adjusted according to control requirements. In one embodiment, the state space vector includes a device cooperative state vector; the device cooperative state vector includes a cooperative gain coefficient and a conflict coefficient; the cooperative gain coefficient characterizes the energy consumption gain obtained by multiple devices operating cooperatively relative to the independent operation of each device; the conflict coefficient characterizes the degree of thermal offset, humidity offset, or energy consumption offset generated during the operation of multiple devices.

[0032] In a specific example, the device collaboration state vector can also include collaboration priority and collaboration stability index. Collaboration priority is dynamically calculated based on energy consumption per unit of environmental improvement, current operating costs, device response speed, comfort contribution, user preferences, and current energy budget status. For example, in a summer cooling scenario, collaboration priority can decrease sequentially from shading, natural ventilation, fresh air cooling, to air conditioning; in a winter heating scenario, collaboration priority can decrease sequentially from passive solar heat gain, to fresh air heating, to underfloor heating; in a transitional season scenario, a combination of low-energy-consumption devices such as fresh air, shading, and nighttime ventilation can be prioritized to meet temperature and humidity targets. Collaboration stability index is used to evaluate the ability of the device combination to maintain environmental stability over the long term, and can be determined jointly by temperature fluctuation amplitude, humidity fluctuation amplitude, carbon dioxide concentration fluctuation amplitude, device start-stop frequency, power fluctuation frequency, and user comfort change rate.

[0033] Therefore, the synergistic benefits and conflict levels between devices are quantified and enter the state space vector. The subsequent construction of the search space can directly use the quantified synergistic gain coefficient and conflict coefficient to sort and filter device combinations, and the determination of device combination priority has an objective data basis.

[0034] There are various ways to calculate the collaborative gain coefficient, such as calculation based on energy consumption comparison or calculation based on environmental improvement comparison. In one embodiment, the collaborative gain coefficient is the ratio of the difference between the first energy consumption and the second energy consumption to the first energy consumption; the first energy consumption is the total energy consumption when multiple devices independently complete the target task, and the second energy consumption is the total energy consumption when multiple devices work collaboratively to complete the target task.

[0035] In a specific example, the synergistic gain coefficient = (E1 - E2) / E1, where E1 is the first energy consumption, i.e., the total energy consumption when multiple devices independently complete the target task; E2 is the second energy consumption, i.e., the total energy consumption when the multiple devices work together to complete the same target task; the units of E1 and E2 are the same, for example, kilowatt-hours. The target task refers to the task of adjusting the environmental parameters of a specified space to a target range, such as adjusting the indoor temperature of a room from 28 degrees Celsius to 26 degrees Celsius. The first and second energy consumptions can be obtained from energy consumption metering data statistics, or from simulations using equipment energy efficiency models or building digital twin models.

[0036] Furthermore, a time-corrected cooperative gain coefficient K = (E1 - E2) / E1 × (T1 / T2) can be introduced, where T1 is the time required for multiple devices to reach the target state independently, and T2 is the time required for multiple devices to reach the target state collaboratively. The units for T1 and T2 are the same, for example, hours. The time-corrected cooperative gain coefficient reflects the benefits of collaborative operation in terms of both energy consumption and adjustment speed: the faster the collaborative operation reaches the target state, the larger T1 / T2, and the larger the time-corrected cooperative gain coefficient.

[0037] Therefore, the synergistic gain coefficient is directly calculated from the total measurable energy consumption under the two operating modes, quantifying the synergistic benefits by the relative saving ratio, so that the synergistic benefits of different equipment combinations can be compared under a unified standard; the time-corrected synergistic gain coefficient further incorporates the adjustment speed into the quantification range, making the evaluation of synergistic benefits more comprehensive.

[0038] Optionally, the state space vector can be constructed based on the building digital twin model, or based on historical operating data, real-time building environment data, or a combination of the above data. This invention does not limit the specific construction method of the state space vector.

[0039] In a specific example, a building digital twin model is a virtual mapping model of the built environment, which can include building and space models, thermal models, equipment models, and human behavior models. The building and space model includes building geometry, room zoning, building envelope, orientation, window-to-wall ratio, shading conditions, etc., used to determine heat exchange boundaries and spatial coupling relationships. The thermal model includes thermal resistance, heat capacity, thermal inertia, thermal equilibrium state, thermal response time constant, etc., used to predict building thermal response and verify control results. The equipment model describes the capacity, energy consumption, response delay, and direction of action of equipment such as air conditioning, fresh air systems, underfloor heating, humidity control equipment, shading devices, electric windows, and energy storage devices. The human behavior model includes occupancy rate, usage habits, time patterns, comfort preferences, etc., used to modify comfort targets and equipment call-up strategies. The building digital twin model can also incorporate outdoor meteorological input. The thermal inertia parameters, thermal response time constant, spatial thermal coupling matrix, and equipment response model parameters calculated by the digital twin model can be encoded into the state space vector as components of the thermal state vector.

[0040] Step S200: Construct the search space based on the state space vector and the respective cooperative relationship weights, conflict relationship weights, and expected energy consumption of multiple device combinations.

[0041] Among them, the priority of equipment combinations in the search space is positively correlated with the weight of cooperative relationships, and negatively correlated with the weight of conflict relationships and the expected energy consumption.

[0042] See Figure 4In one example, constructing the search space may include steps S210 to S230. Step S210 involves obtaining the collaboration relationship weights, conflict relationship weights, and expected energy consumption for each of the multiple device combinations. Step S220 involves determining the priority of each device combination in the search space based on the state space vector and the collaboration relationship weights, conflict relationship weights, and expected energy consumption of each device combination. The priority is positively correlated with the collaboration relationship weights and negatively correlated with the conflict relationship weights and expected energy consumption. Step S230 involves constructing the search space based on the priorities of each device combination. Steps S210 to S230 are explained below.

[0043] An equipment combination refers to a combination of two or more devices that participate in coordinated operation. Equipment may include air conditioning systems, fresh air systems, underfloor heating systems, humidity control systems, shading systems, electric window systems, energy storage systems, photovoltaic systems, heat pump systems, and other building environment control equipment. This invention does not limit the specific types of equipment.

[0044] Collaboration relationship weights and conflict relationship weights are attributes of an equipment portfolio, reflecting the intensity of benefits and conflicts arising from the collaborative operation of the equipment within the portfolio. The sources of collaboration relationship weights and conflict relationship weights can be various: they can be pre-configured based on equipment type and operational experience, provided by external systems such as building automation platforms or energy management platforms, or determined based on an equipment collaboration relationship graph. This embodiment uses determination based on an equipment collaboration relationship graph as an example for illustration.

[0045] In one embodiment, the weights of collaborative relationships and conflict relationships are determined based on a device collaborative relationship graph. The device collaborative relationship graph uses devices as nodes and collaborative relationships, conflict relationships, substitution relationships, and stability relationships between devices as edges, with each edge configured with a corresponding relationship weight. The relationship weights include collaborative relationship weights, conflict relationship weights, substitution relationship weights, and stability relationship weights.

[0046] In a specific example, the equipment collaboration graph can be represented as G=(V,E,W), where V is the set of equipment nodes, E is the set of equipment relationship edges, and W is the set of equipment relationship weights. Collaboration relationships indicate that the combined operation of multiple devices can achieve higher environmental improvement or energy-saving effects than individual operation; conflict relationships indicate that multiple devices experience thermal offsetting, humidity offsetting, energy consumption offsetting, or control direction conflicts during operation; substitution relationships indicate that multiple devices can achieve the same or similar environmental regulation goals; stability relationships indicate the ability of the equipment combination to maintain stable environmental operation. Each relationship edge is assigned a corresponding relationship weight, the weight value reflecting the strength of the equipment relationship. Each relationship weight can be a normalized dimensionless quantity, for example, taking a value in the range of 0 to 1.

[0047] In one example, the weights for the synergistic relationship between shading equipment and air conditioning equipment could be 0.9, the conflict relationship 0.1, the substitution relationship 0.2, and the stability relationship 0.8; the weights for the synergistic relationship between air conditioning equipment and underfloor heating equipment could be 0.1, the conflict relationship 0.9, the substitution relationship 0.2, and the stability relationship 0.3; and the weights for the synergistic relationship between air conditioning equipment and fresh air system could be 0.8, the conflict relationship 0.2, the substitution relationship 0.4, and the stability relationship 0.7. These values ​​are merely examples; the specific weights should be determined based on the actual operating characteristics of the equipment in the building.

[0048] In one example, the collaborative relationship weights of a device portfolio can be obtained by aggregating the collaborative relationship weights on the relationship edges between each pair of devices in the portfolio, such as by averaging, minimizing, or weighted summing; the aggregation method for the conflict relationship weights of a device portfolio is similar. The above aggregation methods can be used individually or in combination, and this invention does not limit the specific aggregation method.

[0049] Alternatively, the device collaboration relationship graph can be implemented using methods such as rule base, matrix structure, knowledge graph, graph neural network or machine learning model automatic generation. Other implementation methods that can express device nodes, relationship edges and relationship weights are also applicable.

[0050] Therefore, the collaborative relationship weights and conflict relationship weights on which the search space is constructed have a unified and structured source. The relationship strength of any combination of devices can be obtained from the relationship weights of the corresponding edges in the graph, avoiding the need to set weights for each combination of devices individually. The graph uses four types of relationship edges to completely depict the operational relationships between devices, providing a more comprehensive basis for the selection of device combinations.

[0051] The relationship weights can be configured statically or updated dynamically as the building actually operates. In one embodiment, the model-based prediction-based building environment equipment control method further includes: obtaining the execution result of the target control sequence after execution; updating the relationship weights of each edge in the equipment coordination relationship graph according to the execution result; wherein, the coordination relationship weights increase with the increase of comfort improvement and energy saving in the execution result, and decrease with the increase of conflict degree in the execution result.

[0052] In one example, the execution results can be organized into an execution result vector R(t), which includes the actual temperature changes, actual humidity changes, air quality changes, actual energy consumption, equipment operating status, and user feedback collected during the execution of the target control sequence. Based on the execution results, the following can be statistically obtained: Comfort Gain, representing the degree of comfort improvement brought about by the combined operation of equipment; Energy Saving, representing the degree of energy saving compared to single-equipment operation or historical baseline operation; Conflict Penalty, representing the degree of thermal offsetting, humidity offsetting, energy offsetting, or control direction conflict during the combined operation of equipment; Energy Waste, representing the additional energy consumption caused by equipment conflict or unreasonable coordination; Substitution Success, representing the degree to which one equipment successfully replaces another equipment to achieve the same control objective; Substitution Failure, representing the degree to which substitution control fails to achieve the target state; Stability Gain, representing the degree to which the combined equipment improves the stability of the indoor environment; Fluctuation Penalty, representing the penalty for fluctuations in temperature, humidity, air quality, or power; and Switch Penalty, representing the penalty for frequent start-stop operations of equipment. All of the above quantities can be normalized and are dimensionless.

[0053] In a specific example, the multidimensional relationship weights of the edge between device i and device j can be expressed as Wij(t) = {Ws,ij(t), Wc,ij(t), Wr,ij(t), Wt,ij(t)}, where Ws,ij(t) is the cooperative relationship weight between device i and device j at time t, Wc,ij(t) is the conflict relationship weight, Wr,ij(t) is the substitution relationship weight, and Wt,ij(t) is the stability relationship weight. The weights of various relationships can be updated according to the execution results: Ws,ij(t+1)=Ws,ij(t)+α1×ComfortGain+α2×EnergySaving-α3×ConflictPenalty;Wc,ij(t+1)=Wc,ij(t)+β1×ConflictPenalty+β2×EnergyWaste-β3×ComfortGain;Wr,ij(t+1)=Wr,ij(t)+γ1×SubstitutionSuccess-γ2×SubstitutionFailure;Wt,ij(t+1)=Wt,ij(t)+δ1×StabilityGain-δ2×FluctuationPenalty-δ3×SwitchPenalty. Wherein, α1, α2, α3, β1, β2, β3, γ1, γ2, δ1, δ2, and δ3 are all weight adjustment coefficients, which are dimensionless quantities. Their values ​​can be configured according to the needs of update sensitivity, for example, they can be in the range of 0 to 1, such as 0.05, 0.1, 0.2, or 0.5. The larger the value of the weight adjustment coefficient, the greater the correction of the corresponding relationship weight by the single execution result. After updating in the above manner, the collaborative relationship weight increases with the increase of comfort improvement and energy saving, and decreases with the increase of conflict degree, which is consistent with the update direction in the above embodiment.

[0054] As a result, the relationship weights are continuously adjusted according to the actual execution results of the building, and the equipment coordination relationship map gradually approaches the actual operating characteristics of the equipment in the building, so that the accuracy of control decisions is continuously improved with the accumulation of the operation process.

[0055] In one embodiment of the present invention, the single update magnitude of the relationship weights can be adjusted based on the sufficiency of the execution result sample and the degree of deviation from the operating conditions. Execution results may be affected by occasional disturbances, such as temporary opening of doors and windows, short-term gathering of people, or extreme weather. A single execution result under such conditions may not represent the true operating relationship of the equipment combination. Therefore, when the execution result sample is small or the operating conditions significantly deviate from normal, the single update magnitude of the relationship weights is reduced; when the execution result sample is sufficient and the operating conditions are stable, the normal update magnitude is restored. This avoids excessive correction of the equipment collaboration relationship graph by a single abnormal execution result, ensuring a smooth and reliable evolution process of the graph weights.

[0056] Projected energy consumption is another attribute of an equipment portfolio. In one example, the projected energy consumption of an equipment portfolio refers to the energy consumption that the portfolio is expected to consume within a forecast time window under a predetermined operating mode, or it can be converted into projected energy consumption per unit time. Projected energy consumption can be estimated based on the rated power, projected load rate, and projected operating time of each device in the portfolio, or it can be predicted by the equipment energy efficiency model or building digital twin model, and the unit can be kilowatt-hours.

[0057] The search space is a decision space that can be used to generate control sequences after uniformly mapping equipment collaboration relationships, building operating states, and operating constraints. In a specific example, the search space is constructed from the equipment collaboration relationship graph through constraint mapping and budget state mapping. The search space can be represented as GS={V, E, W, C, P}, where V is the set of equipment nodes, E is the set of equipment relationship edges, W is the set of equipment relationship weights, C is the set of constraints, and P is the set of energy consumption budget states.

[0058] In addition to device nodes, relation edges, and relation weights, the search space can also include operational constraints and budget states. In one embodiment, the search space includes a set of constraints and a set of energy consumption budget states; the set of constraints includes at least one of building thermal inertia constraints, thermal balance constraints, spatial thermal coupling constraints, device response delay constraints, comfort boundary constraints, and device conflict constraints; the set of energy consumption budget states includes at least one of current cumulative energy consumption, daily budget remaining value, weekly budget remaining value, monthly budget remaining value, real-time electricity price, and time-of-use electricity price.

[0059] In a specific example, building thermal inertia constraints are used to ensure that control arrangements are adapted to the building's heat storage and release processes; thermal balance constraints are used to ensure that control arrangements meet the balance between heat gain and loss; spatial thermal coupling constraints are used to ensure that control arrangements consider the heat transfer effects between adjacent spaces; equipment response delay constraints are used to ensure that control arrangements are adapted to the delay time from receiving instructions to generating regulatory effects for each device; comfort boundary constraints are used to ensure that environmental parameters are within the allowable comfort range; and equipment conflict constraints are used to exclude known mutually canceling equipment combinations. Current cumulative energy consumption, daily budget surplus, weekly budget surplus, and monthly budget surplus reflect the remaining energy consumption within each budget period; real-time electricity price and time-of-use electricity price reflect the changes in energy costs over time.

[0060] Optionally, the constraint set may also include safety operation constraints, such as upper power limits and lower start-stop interval limits for equipment, to ensure safe equipment operation; the energy consumption budget state set may also include demand response states to adjust energy consumption arrangements during grid demand response periods. The above extensions can be selected according to the building's operation and management needs.

[0061] Thus, building physics constraints and energy consumption budget status are explicitly incorporated into the search space. Candidate control sequences are constrained and guided by building physics laws and energy consumption budget status during the generation stage, reducing the possibility that the generated results violate building operation laws or exceed budget conditions.

[0062] In one example, the search space is dynamically constructed in real time based on the state space vector, device collaboration state vector, device collaboration relationship graph, and energy budget status. During the search space construction process, device combinations with higher collaboration relationship weights receive higher priority; device combinations with higher conflict relationship weights receive lower priority; device combinations with higher expected energy consumption receive further lower priority when the energy budget is tight, while device combinations with lower expected energy consumption receive higher priority. For example, in a summer cooling scenario, when the outdoor temperature is lower than the indoor temperature and the remaining energy budget is low, the priority of combinations related to shading devices, natural ventilation devices, and fresh air devices in the search space increases, while the priority of combinations related to air conditioning devices decreases.

[0063] Therefore, the generation of candidate control sequences takes place within a priority-organized search space. Device combinations with high coordination, low conflict, and low expected energy consumption are given priority in sequence generation. The sequence generation process does not require exhaustive search of all device combinations, thus reducing the computational overhead of control decision-making and improving decision-making efficiency.

[0064] Optionally, the search space is not a fixed topology, but is dynamically reconstructed based on the state space vector, device collaboration state vector, device collaboration relationship graph, energy consumption budget status, and real-time environmental changes, so that the search space always matches the current operating status of the building.

[0065] In one embodiment of the present invention, the adjustment of device combination priority can also be smoothed or hysteresis-processed. Relationship weights and energy consumption budget states fluctuate briefly during operation. If the priority ranking responds immediately to each fluctuation, the search space will be frequently reconstructed, and candidate control sequences generated in adjacent control cycles may switch back and forth, causing frequent device start-ups and shutdowns. Therefore, when the change in priority ranking is small or short-lived, the current search space structure is maintained; when the change is persistent or significant, the search space is reconstructed. This avoids frequent search space reconstruction caused by short-term fluctuations in weights or budget states, prevents back-and-forth switching of control arrangements in adjacent control cycles, and helps maintain the continuity of the control process and the stability of indoor environmental parameters.

[0066] Step S300: Generate multiple candidate control sequences oriented towards the prediction time window within the search space.

[0067] The forecast time window refers to a future period of time that the control decision is directed towards. In one example, the duration of the forecast time window can be set according to the control scenario, for example, it can be from 0.5 hours to 24 hours, such as 0.5 hours, 1 hour, 2 hours, 4 hours, 8 hours or 24 hours; the specific value of the duration should be adapted to the building thermal response time constant and the control cycle.

[0068] In one example, the process of generating candidate control sequences includes: first, generating a set of candidate control actions within the search space, which can be represented as {a1, a2, ..., an}, where each control action represents a device collaborative control unit, which may include the calling device, start-up time, operating power, control objective, and duration; for example, the control action could be turning on the shading, turning on natural ventilation, turning on fresh air cooling, turning on air conditioning cooling, or turning on underfloor heating, etc. Then, based on the collaborative relationships, substitution relationships, conflict relationships, and stability relationships between devices, path reasoning and combination expansion are performed on the candidate control actions to generate multiple candidate control sequences within the prediction time window. The candidate control sequence represents a device collaborative control path, rather than a control command for a single device. For example, a candidate control sequence could be {a1, a3, a5, a8}, indicating that control actions a1, a3, a5, and a8 are executed sequentially.

[0069] In a specific example, when generating candidate control sequences, the operational experience of historical high-yield control paths, equipment coordination relationship graphs, and the characteristics of the current state space vector can be combined to perform combined reasoning, path expansion, and prediction generation of control paths, thereby forming new equipment coordination control paths. This allows candidate control sequences to not only cover existing effective control experience but also explore new coordination methods.

[0070] Optionally, the generation of candidate control sequences can be achieved using at least one of graph search algorithms, reinforcement learning algorithms, generative neural networks, and large language models. Multiple algorithms can also be used in combination, and other generation methods that can generate control sequences within the search space are also applicable.

[0071] Optionally, based on the building's digital twin model and candidate control sequences, changes in the building's environmental state within a prediction time window can be predicted to form a predicted state vector. This predicted state vector serves as the basis for subsequent thermal verification and determination of the target control sequence. This approach bases the selection of candidate control sequences on predictions of future environmental states, which is beneficial for identifying deviations in control arrangements in advance.

[0072] Step S400: Perform thermal verification on the candidate control sequence based on the building's thermal characteristics.

[0073] Building thermal properties refer to the inherent thermal characteristics of a building, which may include building thermal inertia, thermal balance characteristics, spatial thermal coupling relationships, and thermal response time constant.

[0074] In one example, thermal verification includes thermal consistency verification, which determines whether a candidate control sequence conforms to the building's thermal inertia, thermal balance, and spatial thermal coupling relationship: if the adjustment rhythm of the candidate control sequence is consistent with the building's heat storage and release process, the adjustment target is consistent with the balance relationship between the building's heat gain and loss, and the control arrangement takes into account the heat transfer effect between adjacent spaces, then the candidate control sequence is determined to pass the thermal consistency verification.

[0075] In one example, thermal verification may also include response delay verification, which determines whether a candidate control sequence can reach the target state within the expected time based on equipment response delay and building thermal response time; thermal verification may also include comfort verification, which determines whether environmental parameters are within the comfort boundary constraints after the candidate control sequence is executed. Each verification can be implemented individually or in combination. Candidate control sequences that fail verification are excluded or enter the correction process.

[0076] Therefore, the candidate control sequence is checked against the actual thermal behavior of the building before it is issued and executed. Control arrangements that deviate from the building's thermal characteristics are eliminated in advance, the fluctuation of indoor environmental parameters is reduced, and the control effect is more stable.

[0077] Step S500: Determine the target control sequence from the candidate control sequences that have passed thermal verification, and send the target control sequence to the corresponding equipment for execution.

[0078] See Figure 5In one example, the process of determining the target control sequence from the candidate control sequences that have passed thermal verification and issuing it for execution may include steps S510 to S540. Step S510 involves identifying the degree of conflict among the device combinations in the candidate control sequences to obtain a conflict coefficient, and performing conflict resolution on the candidate control sequences when the conflict coefficient is higher than a preset conflict threshold; Step S520 involves determining the expected total energy consumption of the candidate control sequences within the prediction time window, and performing budget correction on the candidate control sequences when the expected total energy consumption is higher than the current remaining energy consumption budget; Step S530 involves determining the target control sequence from the candidate control sequences that have passed thermal verification and issuing the target control sequence to the corresponding devices for execution; Step S540 involves collecting the execution results after the target control sequence is executed, and updating the relationship weights of the device collaboration relationship graph based on the execution results, as the basis for constructing the search space in the next control cycle. Steps S510 to S540 are explained below.

[0079] Before the target control sequence is sent to the corresponding device for execution, conflict checks can be performed on the candidate control sequences. There are various methods for conflict checking, such as rule-based conflict comparison and quantization identification based on conflict coefficients. This embodiment uses quantization identification based on conflict coefficients as an example.

[0080] In one embodiment, before sending the target control sequence to the corresponding device for execution, the method further includes: identifying the degree of conflict in the device combination in the candidate control sequence and obtaining a conflict coefficient; if the conflict coefficient is higher than a preset conflict threshold, performing conflict resolution on the candidate control sequence; conflict resolution includes at least one of shutting down some devices in the device combination, reducing the operating power of the devices, delaying the startup time of the devices, and regenerating the candidate control sequence.

[0081] In a specific example, when identifying the degree of conflict among device combinations in a candidate control sequence, combinations of devices that mutually cancel each other out at the thermal level can be identified, such as simultaneous operation of air conditioning and underfloor heating, simultaneous operation of air conditioning and low-temperature fresh air supply, and simultaneous operation of dehumidification and humidification. The degree of conflict covers types such as thermal cancellation, humidity cancellation, energy consumption cancellation, and control direction cancellation. The conflict coefficient here is the same concept as the conflict coefficient in the device cooperative state vector, and is calculated by device conflict identification. The preset conflict threshold can be set according to operational experience. For example, when the conflict degree is normalized to 0 to 1, the preset conflict threshold can be 0.6 to 0.9, such as 0.6, 0.7, 0.8, or 0.9. The specific value of the preset conflict threshold should be such that conflict resolution is triggered before significant conflict causes energy waste.

[0082] In one example, when shutting down some devices in a device portfolio, they can be shut down in ascending order of coordination priority, for example, prioritizing the shutdown of low-priority devices to resolve conflicts while preserving the regulation capabilities of high-priority devices. Reducing device operating power and delaying device startup time are suitable for situations with relatively mild conflicts, mitigating the mutual cancellation effect through peak shifting or power reduction. Regenerating candidate control sequences can be achieved through control path reconstruction, i.e., re-executing the processes of state space vector construction, device coordination graph reasoning, candidate control sequence generation, and budget constraint verification until the candidate control sequences satisfy the conflict constraints. The above resolution methods can be implemented individually or in combination.

[0083] As a result, operational conflicts between devices are identified and resolved before execution, and combinations of devices with mutually canceling effects are excluded from actual operation, thus reducing ineffective energy consumption and fluctuations in indoor environmental parameters caused by device conflicts.

[0084] Before the target control sequence is sent to the corresponding equipment for execution, an energy consumption budget check can be performed on the candidate control sequences. There are various ways to check the budget, such as overall comparison by prediction time window or rolling comparison by budget period. This embodiment uses overall comparison by prediction time window as an example.

[0085] In one embodiment, before the target control sequence is sent to the corresponding device for execution, the method further includes: determining the expected total energy consumption of the candidate control sequence within the prediction time window; if the expected total energy consumption is higher than the current remaining energy consumption budget, performing budget correction on the candidate control sequence; the budget correction includes relaxing the environmental control target, calling devices in order of energy consumption from low to high, reducing the priority of high-energy-consuming device combinations in the search space, and regenerating at least one of the candidate control sequences.

[0086] In one example, the estimated total energy consumption can be obtained by summing the estimated energy consumption of each control action in the candidate control sequence, or by predicting the candidate control sequence as a whole using an equipment energy efficiency model or a building digital twin model. The unit can be kilowatt-hours. It should be noted that the estimated total energy consumption and the estimated energy consumption of the equipment combination when constructing the search space are different levels of quantities: the estimated total energy consumption is for the entire candidate control sequence, while the estimated energy consumption is for a single equipment combination. The current remaining energy consumption budget is determined based on the daily, weekly, or monthly budget remaining values, and can also be discounted in conjunction with real-time electricity prices, time-of-use electricity prices, or demand response requirements. If the estimated total energy consumption is not higher than the current remaining energy consumption budget, the candidate control sequence passes the budget verification.

[0087] In a specific example, relaxing environmental control targets means adjusting the target values ​​of environmental parameters within permissible limits. For instance, adjusting the target temperature for summer cooling from 26 degrees Celsius to 27 degrees Celsius, or the target temperature for winter heating from 22 degrees Celsius to 21 degrees Celsius, to reduce the depth of regulation and energy demand. Prioritizing equipment based on energy consumption (from low to high) means prioritizing low-energy-consuming equipment such as shading devices, natural ventilation, and fresh air systems, reducing the involvement of high-energy-consuming equipment like air conditioners. Lowering the priority of high-energy-consuming equipment combinations in the search space involves reassessing the equipment collaboration graph, reducing the priority of high-energy-consuming collaboration paths, and increasing the priority of energy-saving collaboration paths. Regenerating candidate control sequences can also be achieved through control path reconstruction until the candidate control sequences meet budget constraints. These corrective measures can be implemented individually or in combination.

[0088] Therefore, the estimated total energy consumption of the candidate control sequence is compared with the current remaining energy consumption budget before execution. Control arrangements that exceed the budget are corrected before issuance. Equipment operation is constrained within the energy consumption budget, and the control process can be dynamically adjusted between environmental control objectives and energy consumption budget.

[0089] There are several ways to determine the target control sequence from candidate control sequences that have passed thermal verification and completed conflict resolution and budget correction. For example, it can be selected based on the comprehensive evaluation value, the first candidate control sequence that satisfies all constraints can be selected according to the candidate generation order, or a preset priority rule can be used. This embodiment will use the selection based on the comprehensive evaluation value as an example.

[0090] In a specific example, the comprehensive evaluation value of each candidate control sequence that passes thermal verification is determined, and the candidate control sequence with a comprehensive evaluation value no lower than that of other candidate control sequences is determined as the target control sequence. The comprehensive evaluation value is determined by synergistic benefits, expected total energy consumption, comfort achievement level, environmental stability, and equipment conflict level. The comprehensive evaluation value increases with the increase of synergistic benefits, comfort achievement level, and environmental stability, and decreases with the increase of expected total energy consumption and equipment conflict level. For example, the comprehensive evaluation value = w1 × synergistic benefits - w2 × expected total energy consumption + w3 × comfort achievement level + w4 × environmental stability - w5 × equipment conflict level, where w1, w2, w3, w4, and w5 are dimensionless weighting coefficients, and their values ​​can be configured according to control requirements, for example, they can be in the range of 0 to 1, such as 0.1, 0.2, 0.3, 0.5, or 1; the synergistic benefits, expected total energy consumption, comfort achievement level, environmental stability, and equipment conflict level are normalized and then substituted into the input.

[0091] In one example, synergistic gain is a sequence-level evaluation factor, which can be obtained by summing or weighting the synergistic gain coefficients of the various equipment combinations involved in the candidate control sequence; comfort achievement refers to the degree to which environmental parameters reach the comfort target after the execution of the candidate control sequence, which can be determined based on the deviation between the predicted state vector and the comfort target; environmental stability refers to the degree to which the candidate control sequence keeps the environmental parameters stable within the prediction time window, which can be determined based on the predicted fluctuation range of the environmental parameters; and equipment conflict degree can be determined based on the conflict coefficient of each equipment combination in the candidate control sequence.

[0092] It should be noted that the stability relationships in the equipment coordination relationship graph, the coordination stability index in the equipment coordination state vector, and the environmental stability in the comprehensive evaluation value are three different concepts: the stability relationship is the relationship edge in the graph describing the ability of the equipment combination to maintain stable environmental operation; the coordination stability index is a quantitative component in the state space that evaluates the long-term ability of the equipment combination to maintain environmental stability; and environmental stability is an evaluation factor for a single candidate control sequence. Similarly, coordination benefits and coordination gain coefficients are concepts at different levels: the coordination gain coefficient is a quantitative index at the equipment combination level, while the coordination benefits are a sequence-level evaluation factor obtained by summing the coordination gain coefficients.

[0093] In one example, when the target control sequence is sent to the corresponding device for execution, each control action in the target control sequence can be sent to the controller of the corresponding device through the building automation bus or IoT communication protocol. The controller then performs the corresponding adjustment action according to the start time, operating power, control target and duration. This invention does not limit the communication method used for sending the sequence.

[0094] In one example, during the execution of the target control sequence, actual temperature changes, actual humidity changes, air quality changes, actual energy consumption, equipment operating status, and user feedback are continuously collected to form the execution result. The execution result is fed back to the state space vector construction stage, the equipment coordination relationship map stage, and the candidate control sequence generation stage to correct the coordination gain coefficient, conflict coefficient, coordination priority, coordination stability index, and model parameters, so that the equipment coordination relationship continues to evolve with the building operation process.

[0095] In a specific application example, under a summer cooling scenario, the indoor temperature is higher than the comfort target and the solar radiation intensity is high. The state space vector shows that the current energy consumption budget is limited. The device coordination relationship graph shows that there is a high coordination relationship weight between shading and natural ventilation, and a certain substitution relationship weight between fresh air cooling and air conditioning cooling. In the generated candidate control sequence, the priority sequence is: adjust the shading opening to reduce solar heat gain, turn on natural ventilation or nighttime ventilation, turn on fresh air cooling when necessary, and turn on air conditioning cooling at low power when the comfort target still cannot be met. The thermal verification step verifies whether the outdoor heat and humidity load introduced by ventilation has a reverse effect, and the budget verification step verifies whether the expected total energy consumption is not higher than the current remaining energy consumption budget. Finally, the target control sequence that meets the comfort and budget constraints is determined and issued.

[0096] In another specific application example, during winter heating, the indoor temperature is lower than the target value and the outdoor temperature is also low. The state space vector shows that underfloor heating has a longer response delay but higher stability, while fresh air heating has a faster response but higher energy consumption. Candidate control sequences including passive solar heat gain, fresh air heating, and underfloor heating are generated according to collaborative priorities. The early start-up time of underfloor heating is determined by building thermal inertia constraints. If simultaneous operation of air conditioning heating and low-temperature fresh air leads to thermal conflicts, the operating power of lower-priority equipment is reduced or its start-up time is delayed to reduce energy consumption offsetting.

[0097] In another specific application example, during the transitional season, when outdoor temperature and humidity conditions are suitable, a combination of low-energy-consuming equipment such as fresh air systems, sunshades, and nighttime ventilation is prioritized to meet the temperature and humidity targets. High-energy-consuming equipment such as air conditioners are kept off or in low-power standby mode, thus keeping energy consumption at a low level while meeting comfort targets.

[0098] It should be noted that the steps in the above method are only for clarity of description. In practice, they can be combined into one step or some steps can be broken down into multiple steps. As long as they include the same logical relationship, they are all within the protection scope of this invention. Adding insignificant modifications or introducing insignificant designs to the process, without changing the core design of the process, are also within the protection scope of this invention.

[0099] It should be noted that the actions described in the claims can be performed in different orders and still achieve the desired result; provided there is no contradiction, the execution order of the above steps and examples is not limited by the order of description.

[0100] Figure 6 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. See also... Figure 6The electronic device includes a processor and a memory; the memory stores a computer program, which, when executed by the processor, implements the model-predictive building environment equipment control method provided in the above embodiments.

[0101] The electronic device provided in this embodiment is a hardware carrier embodiment corresponding to the above method embodiments. It implements the model-predictive building environment equipment control method provided in the above method embodiments by loading and running a computer program. The parameter selection, optional extensions, and technical effects of each step in the above method also apply to this electronic device and will not be repeated.

[0102] In one example, the processor can be a data processing device such as a central processing unit, microprocessor, digital signal processor, field-programmable gate array, or application-specific integrated circuit; the memory can include non-volatile memory and volatile memory, such as read-only memory, flash memory, hard disk, solid-state drive, and random access memory. This invention does not limit the specific types of processors and memory.

[0103] Optionally, the electronic device may also include a communication interface and a bus. The communication interface is used to interact with controllers, sensors, and external platforms of various devices in the building. Data transmission between the processor, memory, and communication interface is accomplished via the bus. The specific form of the communication interface can be adapted to the fieldbus or network standard to which it is connected.

[0104] In a specific example, the electronic device can take the form of a building automation host, an edge computing gateway, a local server, or a cloud server; the electronic device can be deployed on-site in the building, or in a remote data center or in the cloud, to control one or more buildings via a network. This invention does not limit the deployment form of the electronic device.

[0105] This embodiment provides a computer-readable storage medium on which a computer program is stored; when the computer program is executed by a processor, it implements the model-predictive building environment equipment control method provided in the above embodiment.

[0106] The computer-readable storage medium provided in this embodiment is a carrier embodiment corresponding to the above method embodiment. The computer program carried thereon implements the model prediction-based building environment equipment control method provided in the above method embodiment when executed by a processor.

[0107] In one example, all or part of the steps in the methods of the above embodiments can be implemented by hardware related to program instructions. This program is stored in a storage medium and includes several instructions to cause a device or processor to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory, random access memory, magnetic disk, or optical disk. The present invention does not limit the specific type of storage medium.

[0108] Furthermore, the examples mentioned in the above embodiments and examples can be freely combined, and any combination can be understood as an embodiment; without contradiction, the embodiments described herein can be combined with other embodiments.

[0109] Through the model-based prediction-based building environment equipment control method, electronic equipment, and computer-readable storage medium provided in the above embodiments, building environment control is expanded from single-point parameter adjustment of a single device to multi-device collaborative control oriented to the prediction time window: the collaboration, conflict, substitution, and stability relationships between devices are explicitly expressed in the device collaboration relationship map, the control decision is organized according to the priority of the device combination before it is generated, and is subject to the dual constraints of building thermal characteristics and energy consumption budget status, and the execution result is fed back for continuous correction of relationship weights, forming a closed loop of decision-making, execution, and updating.

[0110] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing the present invention, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of the present invention.

Claims

1. A model-based prediction-based method for controlling building environment equipment, characterized in that, include: Collect real-time data and construct a state space vector based on the real-time data; A search space is constructed based on the state space vector and the respective collaboration weights, conflict weights, and expected energy consumption of multiple device combinations. The priority of the device combination in the search space is positively correlated with the cooperative relationship weight and negatively correlated with the conflict relationship weight and the expected energy consumption. Multiple candidate control sequences oriented towards the prediction time window are generated within the search space; The candidate control sequences are thermally verified based on the building's thermal characteristics. The target control sequence is determined from the candidate control sequences that have passed the thermal verification, and the target control sequence is sent to the corresponding equipment for execution.

2. The model-based prediction-based building environment equipment control method according to claim 1, characterized in that, The collaborative relationship weights and conflict relationship weights are determined based on the device collaborative relationship graph; The device collaboration relationship graph uses devices as nodes and collaboration, conflict, substitution, and stability relationships between devices as edges, with each edge configured with a corresponding relationship weight. The relationship weights include the collaborative relationship weights, conflict relationship weights, substitution relationship weights, and stability relationship weights.

3. The model-based prediction-based building environment equipment control method according to claim 2, characterized in that, Also includes: Obtain the execution result of the target control sequence after execution; Based on the execution results, the relationship weights of each edge in the device collaboration relationship graph are updated; wherein, the collaboration relationship weights increase with the increase of comfort improvement and energy saving in the execution results, and decrease with the increase of conflict degree in the execution results.

4. The model-based prediction-based building environment equipment control method according to claim 1, characterized in that, The state space vector includes a device cooperative state vector; The device cooperative state vector includes a cooperative gain coefficient and a conflict coefficient; The collaborative gain coefficient represents the energy savings gained from the collaborative operation of multiple devices relative to the independent operation of each device. The conflict coefficient characterizes the degree of thermal offsetting, humidity offsetting, or energy offsetting that occurs during the operation of multiple devices.

5. The model-based prediction-based building environment equipment control method according to claim 4, characterized in that, The synergistic gain coefficient is the ratio of the difference between the first energy consumption and the second energy consumption to the first energy consumption. The first energy consumption is the total energy consumption when multiple devices independently complete the target task, and the second energy consumption is the total energy consumption when the multiple devices work together to complete the target task.

6. The model-based prediction-based building environment equipment control method according to claim 1, characterized in that, The search space includes a set of constraints and a set of energy consumption budget states; The set of constraints includes at least one of the following: building thermal inertia constraints, thermal balance constraints, spatial thermal coupling constraints, equipment response delay constraints, comfort boundary constraints, and equipment conflict constraints. The energy consumption budget status set includes at least one of the following: current cumulative energy consumption, daily budget remaining value, weekly budget remaining value, monthly budget remaining value, real-time electricity price, and time-of-use electricity price.

7. The model-based prediction-based building environment equipment control method according to claim 1, characterized in that, Before sending the target control sequence to the corresponding device for execution, the following steps are also included: The degree of conflict of the device combinations in the candidate control sequence is identified to obtain the conflict coefficient; If the conflict coefficient is higher than a preset conflict threshold, conflict resolution is performed on the candidate control sequence; The conflict resolution includes at least one of the following: shutting down some devices in the device combination, reducing the operating power of the devices, delaying the startup time of the devices, and regenerating the candidate control sequence.

8. The model-based prediction-based building environment equipment control method according to claim 1, characterized in that, Before sending the target control sequence to the corresponding device for execution, the following steps are also included: Determine the total energy consumption of the candidate control sequence within the prediction time window; If the projected total energy consumption is higher than the current remaining energy consumption budget, a budget adjustment is performed on the candidate control sequence; The budget adjustment includes relaxing environmental control objectives, calling up devices in order of energy consumption from low to high, reducing the priority of high-energy-consuming device combinations in the search space, and regenerating at least one of the candidate control sequences.

9. An electronic device, characterized in that, include: processor; A memory storing a computer program that, when executed by the processor, implements the model-predictive building environment equipment control method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the model-based prediction-based building environment equipment control method as described in any one of claims 1 to 8.

Images