Building energy consumption control method

Abstract

The application discloses a kind of building energy consumption control methods, comprising: obtaining the thermal state data of equipment;Quantitative thermal urgency index that fuses static temperature difference, dynamic change and immediate demand dimension is generated based on thermal state data;According to quantitative thermal urgency index, the communication resources distributed to equipment are cooperatively scheduled.Scheduling is scored to candidate communication link by multidimensional evaluation system at least containing thermal coordination dimension, under the double budget constraints of communication energy consumption and equipment power, target link is determined by marginal utility evaluation, and communication resource scheduling scheme containing time slot allocation and power authorization is generated and executed.The application deeply couples the physical state demand of equipment and communication resource scheduling, realizes the cooperative optimization of cross-domain resources, improves the response speed and overall energy efficiency of building energy consumption control.

Description

Technical Field

[0001] This invention relates to the field of energy consumption control, and in particular to a method for controlling building energy consumption. Background Technology

[0002] In the context of the current global energy transition and carbon neutrality strategy, the refined management and optimization of energy consumption in buildings, as a major source of urban energy consumption and carbon emissions, has become a crucial research topic. Statistics show that the energy consumption of building heating, ventilation, air conditioning (HVAC), and hot water supply systems accounts for a significant proportion of total building energy consumption. Therefore, introducing advanced IoT communication, intelligent sensing, and optimized control technologies to achieve intelligent and collaborative management of the numerous energy-consuming devices in buildings can not only improve energy efficiency but also effectively alleviate the load pressure on the power grid.

[0003] Currently, intelligent control technologies for building energy consumption have made some progress. Existing solutions generally deploy intelligent sensors and control terminals (such as smart sockets and thermostats) on the energy-consuming equipment (such as electric water heaters and air conditioners), and collect the equipment's status data (such as temperature, on / off status, and instantaneous power) to a central control platform or cloud server via wireless communication technologies (such as Wi-Fi, ZigBee, or LoRa). The central platform remotely controls the start / stop or adjusts the power of the equipment based on preset rules or simple scheduling algorithms. For example, some systems can start thermal storage equipment during off-peak electricity periods according to time-of-use pricing strategies; other systems allow users to remotely control the equipment manually or set scheduled tasks via mobile applications. At the communication level, these systems typically use standardized network access protocols, such as Carrier Sense Multiple Access (CSMA) mechanisms, to manage equipment data reporting and solve channel contention problems when a large number of devices are connected.

[0004] However, with the surge in the number of intelligent devices in buildings and the increasing sophistication of control requirements, existing technologies still face some challenges in achieving deep coupling between communication resource scheduling and device physical status, as well as cross-domain collaborative optimization under multiple resource constraints. For example, existing methods for allocating communication resources usually rely solely on general network layer indicators, neglecting the urgency of the physical tasks carried by the data. This can lead to resource mismatch: devices with stable thermal states and no urgent control needs may gain priority in communication due to their favorable channel conditions, while devices that urgently need to respond to user commands may be stuck in a long wait due to channel conflicts, which reduces the response efficiency of the control system. Summary of the Invention

[0005] Purpose of the invention: In order to solve the above-mentioned problems in the existing technology, a building energy consumption control method is provided.

[0006] Technical solution: According to one aspect of this application, it is applied to at least one device connected to a communication network, comprising: acquiring thermal state data of at least one device; generating a quantitative thermal urgency index based on the thermal state data; scheduling communication resources allocated to at least one device according to the quantitative thermal urgency index to generate a communication resource scheduling scheme; and executing the communication resource scheduling scheme to control the device.

[0007] Beneficial effects: This invention deeply couples the physical state requirements of equipment with communication resource scheduling, realizes cross-domain resource collaborative optimization, and improves the response speed and overall energy efficiency of building energy consumption control. Attached Figure Description

[0008] Figure 1 A flowchart of a building energy consumption control method provided in an embodiment of this application.

[0009] Figure 2 This is a flowchart illustrating how communication resources allocated to at least one device are scheduled based on a quantified thermal urgency index, as provided in an embodiment of this application.

[0010] Figure 3 This is a flowchart illustrating the determination of a target communication link based on a comprehensive score, provided as an embodiment of this application.

[0011] Figure 4 This is another flowchart for scheduling communication resources allocated to at least one device based on a quantified thermal urgency index, as provided in an embodiment of this application.

[0012] Figure 5 This is a flowchart illustrating the comprehensive quantification of immediate demand dimensions provided in this application embodiment. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] In order to solve the aforementioned problems in the existing technology, the applicant conducted in-depth research and discovered that, in addition to the problems in the background technology, the following problems also exist:

[0016] Existing solutions lack a unified, multi-dimensional framework for resource coordination and evaluation. Decision-making often involves viewing problems from isolated perspectives; for example, communication link selection may only consider communication quality, while power allocation may only consider the thermal state of the devices themselves. The availability of communication resources, the power carrying capacity of network links, and the physical requirements of the devices are not considered within a unified framework for joint evaluation and optimization. This fragmented approach prevents the system from making globally optimal decisions. For instance, it might select a communication link with excellent signal strength for a device, but the gateway to which it belongs may be saturated with power, rendering subsequent heating commands ineffective due to insufficient power and thus rendering the decision invalid.

[0017] To solve these problems, combined with Figures 1 to 5 The present invention will be specifically described through the following embodiments.

[0018] Example 1: A building energy consumption control method is provided, applicable to, for example, intelligent centralized hot water supply systems in large commercial buildings. This system comprises dozens to hundreds of intelligent electric water heaters (hereinafter referred to as devices) distributed across different floors. Each device is connected to the power grid via a smart socket with power data acquisition and on / off control functions. These smart sockets are connected to several LoRa communication gateways deployed within the building via LoRa wireless communication technology. The gateways then aggregate the data to a cloud-based building energy consumption control platform via wired or cellular networks. This platform is the application entity that executes the control method of this invention.

[0019] In this embodiment, the (building energy consumption control method) is applied to at least one device connected to a communication network, and includes the following steps:

[0020] Step 1.1: Obtain thermal state data for at least one device;

[0021] Specifically, acquiring thermal state data is not limited to the direct reading of values ​​from a single sensor, but rather achieves precise sensing through multi-source data fusion and modeling. For example, the control platform periodically gathers information from multiple data streams: firstly, the thermal sensing data stream reported by the devices, containing first-hand physical data such as the current water temperature and power status of the devices; secondly, the measurement stream related to the communication network, such as the terminal wireless signal strength (RSSI) and signal-to-noise ratio (SNR) reported by the gateway; and thirdly, the local energy consumption sampling stream, namely the instantaneous power and cumulative power consumption of the devices collected by the smart socket. To process these data from different sources and with different sampling rates, the system uses methods such as Precise Time Protocol (PTP) to timestamp and align each data stream.

[0022] Building upon this foundation, the system goes beyond simply using current temperature readings. It can also acquire more comprehensive thermal state data through predictive methods based on physical models. For instance, based on a first-order thermal capacity-thermal resistance (RC) physical model, combined with current power and ambient temperature, it can estimate and predict the device's temperature, rate of temperature change, and remaining thermal buffer energy in the next time period, collectively forming a complete thermal state vector. This data acquisition approach, integrating real-time measurement and model prediction, provides a more forward-looking and stable data foundation for subsequent precise control, far exceeding that of single-sensor readings.

[0023] Step 1.2: Based on thermal state data, generate a quantitative thermal urgency index;

[0024] The system transforms the multi-dimensional thermal state data obtained in step 1.1 into a single, floating-point numerical thermal urgency index U through a quantization model. _thermal This indicator clarifies the urgency of the device's current need for energy and communication resources. For example, its calculation can comprehensively consider at least three dimensions: the difference between the current water temperature and the user-set target temperature, the rate of temperature change over time, and the intensity of immediate demand triggered by the user's real-time or scheduled actions. If the device's current water temperature is significantly lower than the set value, the water temperature is rapidly decreasing, and it coincides with a peak water usage period scheduled by the user, then its Ui... _thermal The value will be very high. Conversely, for equipment that has reached the target temperature and is in a heat preservation state, its U... _thermal The value will be very low. Based on this, the system abstracts the complex physical state and user needs into numerical values ​​that can be compared and sorted horizontally across different devices to achieve differentiated and intelligent scheduling.

[0025] Step 1.3: Based on the quantitative thermal urgency index, schedule the communication resources allocated to at least one device to generate a communication resource scheduling scheme;

[0026] Communication resources are multi-dimensional and can include communication links (such as which frequency band or gateway to use for communication), communication time slots (when data can be transmitted on the selected link), and communication power. Based on U... _thermal The scheduling of indicators is reflected in at least two levels:

[0027] Firstly, regarding the selection of macro-level communication links, U _thermal It participates in the comprehensive scoring of multiple candidate communication links; even if the link itself has excellent communication quality, if the gateway it is connected to is currently close to its power load limit and cannot support the start-up heating of equipment with high thermal urgency, then the comprehensive score of the link will be reduced by U. _thermal The relevant evaluation dimensions are lowered; this ensures that the selection of links not only considers the quality of communication itself, but also is closely coupled with the physical service requirements of the equipment.

[0028] Secondly, in terms of micro-level task scheduling, U _thermal This directly affects the priority of communication tasks initiated by the device (such as status reporting and control command confirmation). When faced with multiple devices simultaneously requesting access, U... _thermal Higher-value devices will have their communication requests given higher scheduling priority, resulting in more priority transmission opportunities or shorter access latency.

[0029] Through scheduling at these two levels, the system ultimately generates a detailed communication resource scheduling scheme. This scheme may specify which communication link each device should use within a future time window, in which alternative time slots it should attempt to access the network, and the power budget corresponding to its communication behavior.

[0030] Step 1.4: Execute the communication resource scheduling scheme to control the equipment.

[0031] Furthermore, the cloud platform distributes the generated communication resource scheduling scheme to relevant network infrastructure (such as LoRa gateways) and terminal devices. The gateway manages downlink communication windows and access permissions according to the scheme, while the devices interact with the network following the communication strategies specified in the scheme (such as using specific time slots or specific power). Communication scheduling is for the purpose of achieving physical control of devices. For example, devices with high thermal urgency are given priority communication rights, their heating requests are transmitted to the platform first and confirmed, and the platform sends a closing command to the corresponding smart socket via the network to start the device's heating program. Correspondingly, devices with low urgency may be guided to postpone reporting non-critical data during specific periods, saving communication and energy costs for the entire system. During execution, the system continuously monitors the execution effect and uses the results as feedback to initiate the next round of optimization cycles, achieving adaptive closed-loop management of the entire building's energy consumption control.

[0032] Example 2 describes the method for constructing the thermal urgency index, specifically as follows:

[0033] Step S2.1: Quantify the thermal urgency index, including: the static temperature difference dimension, which represents the difference between the current temperature and the set temperature; the dynamic change dimension, which represents the rate of temperature change over time; and the dimension, which represents the immediate demand caused by the user's immediate behavior.

[0034] Specifically, the three dimensions are weighted and summed to obtain the thermal urgency index U. _thermal Among them, the static temperature difference dimension is used to quantify the current heating deficit of the equipment, and its value can be obtained by (T... _set -T * ) + Calculation yielded T _set For the target temperature set by the user (e.g., 95°C), T * (x) represents the current temperature of the device, determined either by sensor measurement or by prediction using a physical model. + The function is positive, i.e., max(x, 0); this dimension intuitively reflects the distance between the device and the target state.

[0035] The dynamic change dimension is used to capture the trend of temperature change, and its value can be expressed as the absolute value of the rate of temperature change |dT|. * / dt|. This dimension is predictive; even if the temperature has not yet fallen below the threshold, if its |dT|... * If the value of / dt| is large and negative, then its state is deteriorating rapidly and needs to be given higher urgency for early intervention.

[0036] The immediate demand dimension is used to quantify sudden, event-driven thermal demand. For example, an immediate demand event occurs when a user issues a command to use water immediately via a mobile app, or when the system receives a pre-set timed heating task. The value D of this dimension can be calculated by summing all active demand events, such as D = Σw _i ·exp(-Δt _i / λ); where w _i A preset weight for the i-th demand event (e.g., immediate water use has a higher weight than scheduled water use); Δt _i λ is the difference between the time the event occurred and the current time; λ is the time decay constant, used to make the demand intensity decay over time.

[0037] Step S2.2, the process of comprehensively quantifying the immediate demand dimension, includes: generating predicted demand intensity based on historical demand data using a prediction model; calculating immediate demand intensity based on real-time user demand event data; and combining the predicted demand intensity with the immediate demand intensity to obtain the final quantification result of the immediate demand dimension.

[0038] Among them, the predicted demand intensity D is generated through a prediction model. _predicted Optionally, a time-series-based seasonal autoregressive moving average (SARIMA) model is employed. This model learns historical water or electricity consumption data of the equipment (e.g., hourly electricity consumption over the past few weeks) and combines it with time feature vectors (e.g., whether it is a weekday or holiday, and which time of day) to predict the probability distribution P of demand within a specific future time window (e.g., the next hour). _demand (t+Δt). Predicted demand intensity D _predicted This is the weighted integral ∫P of the probability distribution. _demand (t+τ)×exp(-τ / λ _pred )dτ;Calculate the immediate demand intensity D based on real-time user demand event data _immediate The calculation method is the same as D in step S2.1; the two are weighted and combined to obtain D. _combined =w _immediate ×D _immediate +w _predicted ×D _predicted ;where w _immediate and w _predicted These are weighting coefficients, for example, they can be set to 0.7 and 0.3 respectively, so that the system's response to real-time instructions takes precedence over its response to predicted demands.

[0039] Based on this, the final calculation formula for the preferred thermal urgency index is as follows:

[0040] U _thermal =α·(T _set -T * )+β·|dT * / dt|+γ·D _combined +δ·uncertainty;

[0041] Among them, U _thermal The final output is a quantitative thermal urgency index, which is a dimensionless floating-point number; α is the weighting coefficient for the static temperature difference dimension, with an example value range of [0.4, 0.6]; T _set Set the target temperature for the equipment, in degrees Celsius (°C); T * dT represents the measured or predicted temperature of the device, in degrees Celsius (°C); β is the weighting coefficient for the dynamic dimension, with an example value range of [0.1, 0.3]; * / dt| represents the absolute value of the rate of temperature change, in degrees Celsius per second (°C / s); γ is the weighting coefficient for the immediate demand dimension, with an example value range of [0.2, 0.4], and the sum of α, β, and γ can be normalized to 1; D _combinedThis represents a combination of forecast and real-time demand intensity; δ is the weighting coefficient for the forecast uncertainty dimension, with an example value range of [0, 0.1]; uncertainty is the standard deviation (std(P)) of the demand probability distribution output by the forecast model. _demand This is used to appropriately increase the urgency when the uncertainty of prediction is high, so as to make the system decision more conservative and safe.

[0042] This step integrates static temperature difference, dynamic changes, and immediate demand dimensions to construct a quantitative thermal urgency index; transforming the originally abstract and multi-dimensional physical state of equipment and user needs into a standardized, calculable single numerical value U. _thermal Its input parameters (such as the current temperature T) * Target temperature T _set Both user command events and physical data are closely related to physical data and business events in the field of building energy consumption control.

[0043] This eliminates the need for the system to process and parse the raw state data; instead, it can directly rely on U. _thermal The value is used to determine which device's control task is more important or urgent. For example, in a building hot water supply scenario, when a water heater triggers a U-wave response due to an emergency water usage command from a user... _thermal During periods of high demand, subsequent communication and energy requests will receive higher priority, enabling the control system to respond quickly to users' immediate needs. Furthermore, introducing demand intensity prediction based on models such as SARIMA allows for advance resource reservation and scheduling preparations for foreseeable demand peaks (such as peak water usage periods in the evening), improving the overall stability and predictability of system operation and eliminating the blind spot of physical demand in communication scheduling.

[0044] Example 3 describes a preferred implementation of multi-dimensional collaborative scoring and communication link evaluation, including:

[0045] Step S3.1: Based on the quantitative thermal urgency index, schedule the communication resources allocated to at least one device, including: establishing a multi-dimensional evaluation system for at least one device; the multi-dimensional evaluation system includes at least a thermal synergy dimension, and its evaluation result is determined by the quantitative thermal urgency index; using the multi-dimensional evaluation system to evaluate each candidate communication link and generate its own comprehensive score; based on the comprehensive score, determine the target communication link from multiple candidate communication links to form a communication resource scheduling scheme.

[0046] Among them, the multidimensional evaluation system is a four-dimensional collaborative scoring model. ， Used to calculate the comprehensive score U for each candidate device-band-gateway combination (k, g). _total (k, g). These four evaluation dimensions are:

[0047] Communication quality dimension Q(k, g): This dimension evaluates the quality of the link itself as a communication channel. Its score is obtained by normalized weighting of multiple indicators such as signal-to-noise ratio, received signal strength, and channel idle rate.

[0048] Load balancing dimension L(k, g): This dimension evaluates the impact of connecting the device to the link on the overall network load balancing. Its score is derived by inverse mapping of indicators such as the gateway's load factor, uplink acknowledgment latency (ACK latency), and queue length of pending messages.

[0049] Energy efficiency dimension E(k, g): This dimension evaluates the energy cost required to complete a unit communication task on the link, and its score is derived by normalizing the inverse of the historical energy consumption per bit of the link.

[0050] Thermal Coordination Dimension T(k, g): This dimension strongly couples the physical thermal requirements of the device with the selection of the communication link, and its score is determined by U. _thermal Determined by factors such as;

[0051] Final overall score U _total (k, g) is the weighted sum of the scores across the four dimensions:

[0052] U _total (k, g) = w _q ·Q(k, g)+w _l ·L(k, g)+w _e ·E(k, g)+w _t ·T(k, g);

[0053] Among them, U _total (k, g) is the combined score of candidate link (k, g); w _q w _l w _e w _t These are the weight coefficients corresponding to the four dimensions, and these weights can be dynamically and adaptively adjusted using a multi-objective optimization algorithm. ， Example initial value: w _q =0.3, w _l =0.2, w _e =0.3, w _t =0.2.

[0054] Step S3.2: The multi-dimensional evaluation system evaluates each candidate communication link in the thermal synergy dimension, which can be achieved by generating a thermal synergy score. The generation of the thermal synergy score includes: comprehensively considering at least three evaluation factors related to the candidate communication link, wherein: the first evaluation factor represents the urgency of the task, and its magnitude is positively correlated with the quantitative thermal urgency index; the second evaluation factor represents the power adequacy, and its magnitude is positively correlated with the available power margin of the candidate communication link; the third evaluation factor represents the switching cost, and a switching penalty is applied when switching from the current communication link is required to select a candidate communication link.

[0055] Preferably, the three evaluation factors can be quantified and multiplied to obtain the thermal synergy score T(k, g): First evaluation factor (task urgency): U is expressed using the sigmoid function. _thermal Mapped to a urgency score T between 0 and 1 _urgency =sigmoid(U _thermal / U _ref );U _thermal The higher, T _urgency The closer the score is to 1, the stronger the correlation between the two; the second evaluation factor (power adequacy): the current power P through the gateway where the link is located. _current_kg With power limit P _max_kg The power margin score T is calculated using the ratio of the two values. _capacity =(1-P _current_kg / P _max_kg The larger the available power margin, the more T _capacity The higher the score, the stronger the positive correlation between the two; the third evaluation factor (switching cost): a switching penalty T is imposed through an exponential decay function. _continuity =exp(-λ×I _switch ); where I _switch As an indicator variable, T is 1 if a handover is required when this link is selected, and 0 otherwise; when a handover occurs, T... _continuity The score will be lowered as a penalty to the total score.

[0056] Alternatively, when evaluating the third evaluation factor, namely the handover cost, a frequency band stability quantification model can be used instead of a simple handover indicator variable. This scheme calculates a comprehensive stability-weighted score Δ. _stable This score considers not only whether a switch has occurred, but also multiple historical and dynamic indicators. For example, Δ _stable The calculation can incorporate: dwell time score S _duration The longer the current link is held, the higher the score; the handover frequency score is S. _frequency The fewer recent handovers, the higher the score; and the channel quality fluctuation score S _varianceThe more stable the historical channel quality, the higher the score; based on this, the system can more accurately and comprehensively assess the potential risks that handover may bring, and make more robust decisions.

[0057] This step establishes a four-dimensional evaluation system incorporating a thermal synergy dimension, elevating the communication link selection decision from a simple network layer problem to a collaborative optimization problem across the physical, network, and application layers, thus addressing the lack of a cross-domain collaborative perspective in decision evaluation. Specifically, by introducing the thermal synergy dimension, link evaluation no longer relies solely on traditional Quality of Service (QoS) metrics; now, the link's overall score is directly influenced by the power adequacy (physical constraints) of the connected gateway and the thermal urgency (application requirements) of the devices it serves; the thermal urgency U... _thermal This input parameter, along with characteristics such as link power margin and switching costs, is nonlinearly combined to generate a thermal synergy score. This ensures that the final decision simultaneously satisfies communication reliability, network balance, energy efficiency, and the executability of the physical task. In practical building energy control scenarios, the system can avoid selecting invalid links where the signal is good but the gateway is unable to handle the power of new heating equipment. This guarantees that every communication decision is physically feasible and meaningful, improving the success rate of control command execution and the overall system resource utilization efficiency.

[0058] Example 4 provides a specific implementation of incremental scanning and token generation with dual budget constraints. It explains how, after generating the comprehensive score of each candidate communication link, the system determines the target communication link set through incremental scanning within a preset budget framework and generates an authorization token.

[0059] Step S4.1, the process of determining the target communication link based on the comprehensive score, includes: initializing the cumulative communication cost and cumulative power cost; progressively evaluating multiple candidate communication links in descending order of comprehensive score; during the evaluation, if including the next candidate communication link would cause the cumulative communication cost to exceed the preset communication energy consumption budget, or cause the cumulative power cost to exceed the preset device power budget, then the evaluation process is terminated; the candidate communication links included in the selection before the termination of the evaluation process are determined as the target communication links.

[0060] Preferably, candidate communication links (e.g., device A-band 1-gateway X) are ranked according to their comprehensive score U. _total Sort the results in descending order to obtain the evaluation list; initialize the system with two accumulators: one for accumulating communication cost C. _RF Set to 0, cumulative power cost P _allocated Set to 0; the system evaluates each link in the list one by one, starting from the top (the link with the highest score); for each link to be evaluated, the system estimates the communication cost increment ΔC required to accept it. _RF(e.g., the energy required to perform channel scanning and access signaling, in millijoules (mJ)) and power cost increment ΔP (e.g., the additional operating power required by the equipment to support services on this link, in watts (W); the system makes a preliminary judgment before adding the link to the final selection set: if C _RF +ΔC _RF This will exceed the preset total communication energy consumption budget B. _RF , or P _allocated +ΔP will exceed the preset total power budget for the equipment. _P If the incremental scan process terminates immediately, then the link is added to the selection set and the cumulative cost is updated: C _RF =C _RF +ΔC _RF P _allocated =P _allocated +ΔP, continue evaluating the next candidate link.

[0061] Step S4.2, the conditions for terminating the evaluation process also include: when evaluating the next candidate communication link, determining the ratio between the overall score gain that can be obtained by accepting the link and the incremental cost of communication energy consumption or equipment power required, as a marginal utility indicator; when the marginal utility indicator is lower than the preset utility threshold, the evaluation process is terminated.

[0062] Furthermore, a marginal utility indicator is introduced, representing the increase in overall score obtained per unit of resource input; the marginal utility ΔU / ΔC in the communication dimension is calculated separately. _RF The marginal utility ΔU / ΔP is calculated along with the power dimension. Here, ΔU represents the increase in the average score of the entire selected link set after accepting a new link. The system sets a utility threshold θ. _margin For example, 0.01. In each iteration, if the maximum marginal utility of the two dimensions is max(ΔU / ΔC)... _RF If both ΔU / ΔP are below this threshold, the return that can be obtained by continuing to invest resources is negligible, and the system will actively terminate the scan, so that resources are prioritized for the most cost-effective link selection.

[0063] Step S4.3, the method also includes determining the communication energy consumption cost used in the evaluation process, including: synchronously acquiring a local energy consumption sampling stream that records the instantaneous power of the device, and an event log that records the communication protocol behavior; using a change point detection algorithm to identify energy consumption mutation events with power changes in the local energy consumption sampling stream; matching and associating the energy consumption mutation events with specific communication protocol events recorded in the event log based on timestamps; and quantifying and attributing the energy consumption cost caused by the specific communication protocol events based on the matching association.

[0064] Specifically, the Cumulative Sum (CUSUM) change-point detection algorithm can be used to process instantaneous power sequences uploaded from high frequencies (e.g., 1Hz) by smart sockets. When a device performs a communication action (such as initiating channel scanning or data transmission), its instantaneous power typically undergoes a step change. The CUSUM algorithm can sensitively capture these inflection points on the power curves, i.e., energy consumption abrupt changes. Based on this, the communication gateway records detailed communication protocol event logs (e.g., time T1, device A begins scanning frequency band F). By precisely matching the timestamps of the power abrupt change events with the timestamps of the communication event logs, the system can bind a specific power curve (and its integral, i.e., energy consumption) to a specific communication action (such as scanning frequency band F), accurately calculating ΔC. _RF .

[0065] Optionally, to achieve more refined cost accounting and attribution analysis, a multi-scale energy consumption decomposition mapping method can be employed. This method sets multiple CUSUM detection thresholds (e.g., a lower power threshold for scanning events and a higher threshold for data transmission) to account for the power characteristics differences of different communication events, thus decomposing the total energy consumption E... _total Decomposed into multiple independent components, such as scan energy consumption E _scan beacon energy consumption E _beacon Data transmission energy consumption E _data Idle listening power consumption E _idle Frequency band switching energy consumption E _switch And so on. This fine-grained energy consumption profile can not only provide ΔC _RF The estimation provides higher accuracy and can also provide deeper insights for subsequent system energy efficiency optimization.

[0066] Step S4.4 After the evaluation process is terminated, it also includes: generating a communication token for authorizing communication access opportunities based on the difference between the communication energy consumption budget and the final cumulative communication cost; and generating a power token for authorizing device power usage based on the difference between the device power budget and the final cumulative power cost.

[0067] This step transforms the abstract budget surplus into a concrete, manageable, and allocable scheduling credential.

[0068] Specifically, the system calculates the total amount of two types of tokens: Token _RF_total =floor((B _RF -C _RF ) / C _unit ); Token _P_total =floor((B _P -P _allocated ) / P _step ); where Token _RF_totalThe total number of communication tokens generated, each token representing the right to access communication once or several times; Token _P_total The total number of power tokens generated, each token representing the right of a device to use a certain amount of power within a certain period of time; B _RF and B _P The preset total budget; C _RF and P _allocated C represents the total cost incurred after the scan was completed. _unit and P _step These represent the unit granularity or face value of the two types of tokens, for example, C. _unit It can be set to 5mJ, P _step It can be set to 100mW; floor() is the floor function.

[0069] Furthermore, after generating tokens, the system can also employ a strategy based on comprehensive priority to allocate these tokens. Specifically, the system calculates a comprehensive allocation priority for each device i to be allocated a token. _i This priority takes into account the thermal urgency U of the device. _thermal_i The length of its pending data queue, QueueLength _i and its historical communication success rate. _rate_i Based on this, the system follows the Priority _i The tokens in the token pool are distributed to each device in descending order of priority and in proportion to their value. To prevent a single high-priority device from monopolizing all token resources, the system also imposes fairness constraints. For example, it stipulates that the number of any type of tokens obtained by a single device cannot exceed 30% of the total number of tokens of that type. The system prioritizes devices that need resources the most (high thermal urgency, large amount of data to be sent) and perform the best (high success rate), while taking into account overall fairness.

[0070] By introducing incremental scanning and marginal utility evaluation mechanisms under dual budget constraints, this invention establishes a refined, efficient, and economically-minded resource allocation framework for building energy consumption control systems. Firstly, the dual budget setting for communication energy consumption and equipment power defines a clear hard cap on system resource consumption, ensuring that the overall operating cost of the system (both communication overhead and total power load) remains within a controllable range. This is crucial for the economic operation of large-scale buildings and meeting the load management requirements of the power grid. Secondly, the introduction of marginal utility evaluation—calculating the ratio of comprehensive score gain to incremental resource cost to determine whether to continue scanning—includes a cost-benefit analysis in the resource allocation process. It intelligently weighs the performance improvement brought by acquiring a better link against the additional scanning and signaling overhead incurred, decisively terminating when the return becomes negligible. This avoids resource waste caused by pursuing ultimate performance and improves the overall efficiency of resource allocation. Third, the budget surplus is transformed into tangible communication tokens and power tokens, providing manageable and allocable tokens for subsequent fine-grained scheduling, so that macro-level budget constraints can be effectively transmitted and implemented to the micro-behavior of each device.

[0071] Example 5 describes an optional implementation scheme for the hysteresis decision mechanism with enhanced thermal inertia, illustrating how the system determines the target communication link from multiple candidate communication links. In some application scenarios, even if the system evaluates a new candidate communication link with a higher overall score, switching between different links too frequently may introduce additional signaling overhead, data transmission interruptions, or even disturbances to the physical state of the device (such as thermal equilibrium).

[0072] Step S5.1: Determine the target communication link from multiple candidate communication links. A hysteresis decision mechanism is applied to decide whether to switch from the current communication link to a new candidate communication link with a higher comprehensive score. The operation of the hysteresis decision mechanism depends on the dynamic switching threshold. The setting of the dynamic switching threshold is positively correlated with the thermal inertia index calculated based on the thermal state data of the device. That is, the greater the thermal inertia of the device, the higher the dynamic switching threshold.

[0073] Preferably, after the incremental scan process is completed, the system will obtain the current optimal candidate link (k). * g * ) and its score U _total (k * g * The system will connect it to the link (k) that the device is currently bound to. - g - ) and its score U _total (k - g - The comparison is made. Switching is not solely based on U. _total (k* g * Slightly higher than U _total (k - g - It will happen, but it needs to satisfy U. _total (k * g * )-U _total (k - g - )>η _hysteresis , where η _hysteresis This is the hysteresis threshold for switching.

[0074] In this embodiment, the hysteresis threshold η _hysteresis It is not a fixed value, but rather dynamically adjustable. Its value is set in relation to what is known as the thermal inertia index. _inertia The physical quantities are positively correlated. Thermal inertia index. _inertia The aim is to quantify the ability of a device to maintain its current thermal state, and its calculation method can be: thermal _inertia =C _th ×|T * -T _ambient | / P _current Among them, C _th The heat capacity of the device can be obtained using the physical model estimation method of Example 8; |T * -T _ambient | This represents the difference between the current temperature of the equipment and the ambient temperature, reflecting the thermal energy currently stored in the equipment; P _current This refers to the current operating power of the equipment. This indicator is the ratio of the thermal energy stored in the equipment to the current input power; a large-capacity water heater filled with hot water and in a low-power heat preservation state has a much greater thermal inertia than an empty water heater that is heating at high power.

[0075] Based on this indicator, the dynamic switching threshold η _hysteresis The calculation formula can be: η _hysteresis =η _0 ×(1+k _inertia ×thermal _inertia / τ _th ); where η _0 The basic threshold; k _inertia τ is the inertial influence coefficient. _th This is the thermal time constant of the device, used for thermal... _inertia Normalize.

[0076] Accordingly, for devices with high thermal inertia, the switching threshold is adjusted accordingly; the system requires a new link with a sufficiently high performance advantage before it is willing to disturb a device that is operating stably. Conversely, for devices with low thermal inertia, the system can more flexibly switch links to pursue optimal communication performance. This mechanism allows the optimization decisions of the communication network to intelligently adapt to the inherent characteristics of the physical objects it serves, improving the stability and efficiency of the entire control system.

[0077] Example 6 describes the specific implementation of hybrid priority scheduling and power-aware adaptive backoff, and elaborates on task scheduling and access conflict management. The execution of this example relies on the authorization token generated in Example 4.

[0078] Step S6.1 In this embodiment, the step of scheduling communication resources allocated to at least one device based on the quantitative thermal urgency index includes: determining a mixed priority for each task in a series of tasks to be scheduled; the determination of the mixed priority integrates the service level factor characterizing the service level of the task and the thermal state factor determined by the quantitative thermal urgency index; and scheduling the series of tasks to be scheduled based on the mixed priority to generate a communication resource scheduling scheme.

[0079] Specifically, every task to be scheduled in the system, whether it is a periodic status report initiated by the device or a remote control command issued by the platform, will be assigned a dynamically calculated hybrid priority. _hybrid The formula for calculating this priority is shown in the example: Priority _hybrid =w _sla ·P _sla +w _thermal ·sigmoid(U _thermal )+w _token ·(Token _RF +Token _P ) / 2; where Priority _hybrid For the final mixed priority score; w _sla w _thermal w _token P represents the adjustable weights of each factor, and the sum of the three can be normalized to 1; _sla This is the service level factor, whose value is mapped according to the service level of the task. For example, an emergency power outage protection command P... _sla The value is 1.0, while the value for ordinary user temperature adjustment commands is 0.6, and the value for routine status data reporting is 0.3; sigmoid(U _thermal U is the thermal state factor. _thermal This refers to the thermal urgency index calculated in Example 2, which is then smoothed and normalized using the sigmoid function; (Token)_RF +Token _P ) / 2 is the token holding factor, representing the number of communication and power tokens the device holds as allocated by Example 4. Devices holding more tokens score higher in this category. Based on the calculated Priority... _hybrid The system ranks all tasks to be scheduled in descending order of their scores. High-priority tasks will be assigned to protected, low-collision-probability communication slots, while low-priority tasks may be scheduled in more contentious slots, or be appropriately delayed or merged for transmission, thus generating the final scheduling scheme.

[0080] Step S6.2: During the execution of the communication resource scheduling scheme, communication conflicts are inevitable due to the openness of the wireless channel. When a communication conflict is detected, an adaptive backoff time is calculated for the scheduling tasks involved in the conflict. The determination of the backoff time includes: a system energy state consideration factor, wherein the backoff time increases accordingly as the available remaining power budget of the system decreases; and a thermal state factor of the scheduling task, wherein the higher the quantitative thermal urgency index of the task, the shorter the maximum allowable backoff time for it (the scheduling task).

[0081] Furthermore, the calculation of adaptive backoff time is based on a combination of at least three technical components: an exponential backoff component that is exponentially related to the number of historical conflict failures; a queue congestion component that is positively correlated with the estimated queue length; and a power stress component determined by the system's remaining power budget.

[0082] Specifically, when a device fails to receive the expected ACK confirmation after sending data, a collision is determined to have occurred. The device does not immediately retransmit; instead, it calculates an adaptive backoff time 'b'. The calculation of 'b' incorporates multiple technical components, and an exemplary formula is: b = min(b _max α·2 ^n +β·q+λ·(1-P _remain / B _P Where b is the final calculated retreat time, in milliseconds (ms); b _max The maximum allowable backoff time; n is the number of consecutive conflict failures in this task, α·2 ^n This is the exponential backoff component, reflecting the exponential avoidance of consecutive conflicts; q is the current network queue length estimated by the device or gateway, and β·q is the queue congestion component, which increases the backoff time to alleviate congestion when the network is congested; P _remain B is the current remaining available power budget for the system. _P For the total power budget, λ·(1-P) _remain / B _PThis is the power stress component. This component reflects the system's energy state considerations: when the overall system power budget is tight (P... _remain (Small), the value of this item will increase, actively extending the backoff time, reducing the power consumption caused by immediate retransmission, and making the equipment behave more conservatively; α, β, and λ are the weighting coefficients of their respective components.

[0083] Furthermore, the thermal state factor is reflected in the maximum retreat time b. _max On dynamic adjustment. _max It is not fixed, but rather depends on the thermal stress U of the equipment. _thermal Negative correlation, for example, b _max =b _0 ·exp(-μ·U _thermal ). Among them b _0 The maximum backoff time is the base value, and μ is the adjustment coefficient. For equipment with very high thermal urgency (e.g., urgently needing heating), the maximum allowable backoff time will be shortened so that its critical mission can be completed and retransmitted as soon as possible.

[0084] Furthermore, to prevent unforeseen excessive energy consumption due to backoff and retransmission, before finally determining the adaptive backoff time, the method also includes: estimating the total energy consumption required to perform the backoff process and subsequent retransmission attempts based on the initially calculated backoff time; comparing the estimated total energy consumption with the available energy budget; and, when the estimated total energy consumption exceeds the available energy budget, applying a compression adjustment to the initially calculated backoff time to generate an energy-constrained final backoff time.

[0085] Optionally, after calculating the initial backoff time b according to the formula, the system will further estimate the total energy E required to complete this backoff and subsequent retransmission attempts. _backoff E _backoff The estimate may include: idle listening energy consumption P during the backoff period. _idle ×b, and the transmission energy P for performing retransmission attempts. _retry Subsequently, the system will E _backoff With available energy budget _threshold The budget can be compared using the power token held by the device. _P The quantity is calculated from E. _backoff >Energy _threshold If the estimated energy consumption exceeds the budget, the system will adjust the backoff time accordingly, for example, b. _adjusted =b×(Energy _threshold / E _backoffThis mechanism generates a final, energy-constrained backoff time that ensures the energy budget is not exceeded. This adds a hard safety net to the system's energy consumption control, preventing conflict management actions from undermining higher-level energy scheduling strategies.

[0086] This invention designs a power-aware adaptive backoff mechanism, enabling devices to intelligently adapt their behavior to the system's macroscopic energy state and their own physical task urgency when dealing with communication conflicts. In traditional CSMA mechanisms, the calculation of backoff time is random and blind; however, in this invention, the calculation formula for backoff time incorporates a power tension component λ·(1-P). _remain / B _P It will include the system's remaining power budget P. _remain This translates into a direct impact on the backoff time, causing the backoff behavior of all devices to be systematically prolonged when the total power of the entire building system approaches its limit. These devices become more yielding and conservative, proactively reducing the additional power surge caused by communication retransmissions, thus providing underlying technical support for ensuring grid security and preventing tripping. Furthermore, the maximum backoff time b... _max With thermal urgency U _thermal The negative correlation design compresses the backoff limit when a device's task (such as a water heater urgently needing heat) is critical, allowing it to retry more quickly. This differentiated backoff strategy again solves the problem of communication scheduling blindly responding to physical demands, ensuring the Quality of Service (QoS) of critical services in harsh channel environments.

[0087] Example 7: An optional implementation scheme for generating decorrelational time slots using virtual phase mapping. In scenarios with a large number of devices communicating concurrently, even with priority scheduling and backoff mechanisms, severe, periodic communication storms can still occur if the devices exhibit inherent synchronicity (e.g., all reporting data at the top of the hour). This example aims to eliminate the behavioral correlation between devices.

[0088] In this embodiment, the generation of the communication resource scheduling scheme further includes: constructing a timestamp-driven virtual phase system based on the virtual phase period associated with the thermal time constant of the device; performing hash mapping on the device identifier, virtual phase, and random salt value to generate a decorrelated communication time slot for each device; and setting the virtual phase period in a proportional relationship with the thermal time constant so that the communication scheduling period matches the thermal response time.

[0089] Specifically, the system will establish a virtual phase system for the entire network; the core of this system is the virtual phase period T. _virtual To ensure that communication scheduling matches the response speed of physical devices, T _virtual The setting and typical thermal time constant τ of the equipment _th(this value can be estimated by the method in embodiment eight) is associated, for example, can be set to T _virtual =k _phase ×τ _th , wherein k _phase is a proportional coefficient, such as 0.1. This enables the control period to be much shorter than the thermal response time, avoiding control lag.

[0090] Based on this period, at any time t _current the normalized virtual phase φ can be calculated as: φ=(t _current mod T _virtual ) / T _virtual . The value of the phase φ periodically changes between 0 and 1. When a device needs to communicate, it does not randomly select a time slot anymore, but generates its exclusive access time slot index s through deterministic hash mapping: s=H((floor(K·φ)<<C1)⊕(DevID&C2)⊕(Salt<<C3))modN; wherein, s is the finally generated time slot index; H() is a hash function, such as SHA-256, to ensure uniform distribution of the output; φ is the current normalized virtual phase; K is the number of phase slices, which is used to divide the virtual period into multiple phases; floor(K·φ) is the current phase slice; DevID is the unique identifier of the device; Salt is a random salt value periodically broadcast by the gateway, which is used to increase randomness and resolve potential hash collisions; C1, C2, C3 are bit operation constants; ⊕ is the XOR operation; & is the bitwise AND operation; << is the left shift operation; N is the total number of available time slots in a communication period.

[0091] This mechanism enables the access time slot of each device at any time to be jointly determined by the public time phase, the device's own identity ID and the random salt value. Due to the avalanche effect of the hash function, even if there is only a tiny difference in DevID or phase φ, the finally generated time slot index s will be greatly different. This allows a large number of devices to be uniformly and pseudo-randomly distributed over the entire communication timeline, statistically reducing the probability that they attempt to access at the same time and realizing decorrelation of time slots.

[0092] Optionally, a multi-period nested virtual phase system is constructed to meet scheduling requirements of different time scales. For example, a slow period T matching the thermal response time can be set at the same time _slow (such as 60 seconds) and a fast period T for real-time communication scheduling _fast (such as 6 seconds). The device passes through the two periodic phases φ _slow and φ _fastBy combining these elements and hashing them, a more time-series-based time slot allocation can be obtained. For example, devices in the same slow phase (such as the heating phase) can have their access behavior finely scheduled in different fast phases, enabling more complex cross-layer collaborative control strategies.

[0093] This embodiment introduces a virtual phase mapping decorrelation time slot generation mechanism to solve the problem of periodic and synchronous access conflicts in large-scale IoT device communication. For example, in building energy consumption control scenarios, a large number of devices may generate communication requests at the same time due to firmware logic or external events (such as the hour), forming a communication storm, causing the traditional CSMA mechanism to almost fail due to continuous channel congestion. This invention performs a hash operation on the common time phase, the unique ID of each device, and a random salt value to generate a deterministic but statistically independent access time slot for each device at any time; its input (device ID, timestamp) and output (time slot index) are closely integrated with specific communication scheduling technologies; it evenly and pseudo-randomly distributes access requests that may originally be highly concentrated on the time axis throughout the entire communication cycle, reducing the probability of instantaneous conflicts. This not only improves the access success rate and overall throughput of the network in large-scale device scenarios, but also makes the system's communication behavior more stable and predictable, providing a more stable and reliable communication foundation for upper-layer applications.

[0094] Example 8 provides a preferred implementation of thermal state prediction based on a physical model. Traditional data acquisition methods may only rely on discrete readings of sensors at specific times. However, this example introduces a physical model, which enables the system to not only see the current state but also predict future trends.

[0095] Optionally, the acquisition of thermal state data further includes: updating the thermal resistance and thermal capacity parameters of the device in real time using a recursive parameter estimation method based on a first-order thermal capacity-thermal resistance (RC) physical model; using the model to predict the current temperature, the rate of temperature change, and the remaining thermal buffer energy; and using the remaining thermal buffer energy to constrain subsequent power allocation decisions.

[0096] Specifically, the system abstracts each controlled thermal device (such as an electric water heater) into a first-order thermal capacity-thermal resistance circuit model. The dynamic behavior of this model can be described by differential equations, the solutions of which are in the form: T(t) = T _env +P(t)·R _th ·(1-exp(-t / τ _th )); where T(t) is the temperature of the equipment at time t; T _env The ambient temperature of the device can be obtained from an independent sensor or set to an approximately constant value; P(t) is the input electrical power of the device at time t, which can be accurately obtained from the smart socket; R _thτ is the equivalent thermal resistance of the equipment, characterizing the rate of heat loss between the equipment and the environment. _th The thermal time constant of the device is equal to the thermal resistance and heat capacity C. _th The product (τ) _th =R _th ·C _th ( ), characterizing how quickly a device responds to temperature changes.

[0097] Because of R _th and C _th These two key physical parameters change slowly due to factors such as equipment aging and scale buildup. This embodiment employs an online parameter estimation method, such as Recursive Least Squares (RLS), to continuously and dynamically update the values ​​of these two parameters. The RLS algorithm continuously compares the model's predicted temperature output with the sensor's actual measured temperature and iteratively corrects RLS based on the error between the two. _th and C _th The estimated values ​​ensure that the physical model always remains consistent with the actual physical characteristics of the device.

[0098] With a dynamically calibrated, accurate physical model, or rather, through a dynamically calibrated, accurate physical model, the system can acquire a much richer thermal state vector Θ than a single sensor reading. _th ={T * dT * / dt, E _buf}; where T * The model outputs T as the current temperature predicted by the model, compared to the raw sensor readings which may contain noise or delay. * After smoothing and filtering, it is more stable and reliable; dT * / dt is the rate of temperature change directly calculated by the model. This value is crucial for determining whether the equipment is in a heating, cooling, or heat preservation state, and how fast the temperature change is; E _buf This is the residual heat buffer energy, a key predictive indicator, and its calculation formula is E. _buf =C _th ·(T _max -T * ), where T _max E represents the highest safe temperature that the equipment can withstand. _buf This represents how much energy the device can absorb without exceeding the temperature limit. This value plays a crucial constraining role in subsequent power allocation decisions. For example, in the power allocation stage of Example 6, even if the device holds a large number of power tokens, if its E _buf Even though it's very small, the system will still limit its actual heating power to prevent overheating.

[0099] In this embodiment, the thermal state data acquired by the system is no longer an isolated, static snapshot, but a dynamic, multi-dimensional state description that includes current values, trends, and future safety margins, providing a solid data foundation for achieving accurate, safe, and efficient energy consumption control.

[0100] Example 9 describes an optional implementation process of a cross-layer joint reselection triggering mechanism, illustrating how the system intelligently determines when to initiate a new round of communication link evaluation and selection, which may consume significant resources. In complex building energy management systems, continuously scoring and reselecting all communication links is impractical, resulting in enormous computational and signaling overhead. Therefore, an efficient triggering mechanism is needed to initiate reselection only when system performance deteriorates significantly or when there is significant room for optimization.

[0101] In a further embodiment, the triggering step for determining when to initiate a new round of scheduling decisions includes: constructing a cross-layer triggering function, the input of which combines at least performance indicators obtained from the communication network and device thermal state deviations parsed from thermal state data; and initiating a new round of scheduling decisions for communication resources when the output value of the cross-layer triggering function meets preset triggering conditions.

[0102] Here, "cross-layer" refers to the fact that the input data of the trigger function spans multiple different layers of the communication protocol stack and the state layer of the physical device. Specifically, the system constructs a cross-layer trigger function Φ, whose exemplary calculation formula is: Φ = γ _1 CollisionRate+γ _2 ·(ACK _P95 / ACK _target )+γ _3 ·(ΔT / T _tolerance )-γ _4 ·ΔU _total ;

[0103] Among them, CollisionRate and ACK _P95 (95th percentile acknowledgment delay) is a performance indicator obtained from the communication network layer or MAC layer, which directly reflects the congestion and reliability status of the current communication link; ΔT (deviation between actual temperature and set temperature) is the state deviation parsed from the thermal state data of the physical device layer, which directly reflects the final execution effect of the energy consumption control task; ΔU _total (The potential score difference between the best alternative link and the current link) is the evaluation metric of the application / control layer, representing the potential benefit of performing a reselection; γ _1 γ _2 γ _3 γ _4 These are the weighting coefficients of each factor.

[0104] The system periodically calculates the value of Φ. When the value of Φ exceeds a preset trigger threshold τ... _trigger When this occurs, it indicates that the overall performance of the system has dropped to an unacceptable level, or that there is a highly attractive optimization opportunity, which will trigger a new round of link evaluation and selection process (i.e., return to Execution Example 3 and subsequent steps).

[0105] Alternatively, to make the triggering mechanism more robust and intelligent, the following enhanced design can be introduced: dynamic determination of weights: weight coefficient γ _1 To γ _4 It is not fixed, but can be determined using the Analytic Hierarchy Process (AHP). The system can adjust the AHP decision matrix based on higher-level operating strategies (e.g., whether it is currently in energy-saving priority mode or experience-first mode), dynamically adjusting the relative importance of each triggering factor.

[0106] Adaptive adjustment of threshold: trigger threshold τ _trigger It can also be adaptive. For example, τ _trigger The value of τ can be positively correlated with the recent link switching frequency, i.e., τ _trigger =τ _base ×(1+k _stability ×switch _frequency This design can effectively suppress system jitter or oscillation in unstable states, i.e., excessively frequent switching of links.

[0107] Trigger persistence determination: To avoid false triggering due to instantaneous network fluctuations or data spikes, the system determines the duration of triggering when Φ > τ. _trigger When the condition is met, a reselection will not be triggered immediately. Instead, it will examine whether the condition has lasted for a certain time window (e.g., the median value of Φ has been greater than the threshold in the past 10 calculations) or whether the emergency triggering condition is met at the same time (e.g., the thermal deviation ΔT exceeds the safety red line). Only when these stricter conditions are met will a reselection trigger signal be issued.

[0108] Example 10 provides an optional implementation scheme for adaptive optimization with enhanced strategy feedback and causal inference. This example describes how the system performs review and introspection after executing a complete control cycle to achieve closed-loop and adaptive optimization of the control strategy.

[0109] In this embodiment, after executing the scheduling scheme, the system enters the strategy feedback and parameter adaptive optimization stage. This stage first performs a comprehensive quantitative evaluation of the execution effect of the previous cycle. Specifically, the system reads the actual execution log of the scheduling scheme, the resource consumption records generated by steps such as in Embodiment Four, and the final thermal achievement status reported by the device (e.g., actual water temperature, user satisfaction feedback), and calculates a multi-dimensional set of key performance indicators (KPIs) based on these.

[0110] This set of KPIs should include at least: communication efficiency: such as average access latency, P95 acknowledgment latency, and final packet loss rate; energy efficiency: such as total system energy consumption, peak power, and energy consumption per unit of service required to complete a unit of thermal service (e.g., heating 100L of water from 20°C to 95°C); and thermal comfort: such as the final achievement rate of temperature targets (1-|T). * _final -T _set | / T _set ), the timeliness of service response from when the user issues a command to when the temperature begins to change, etc.; resource utilization: such as the final utilization rate of the communication token and power token generated by Example 4.

[0111] Preferably, to accurately identify the root cause of performance bottlenecks rather than simply adjusting parameters, the system constructs a pearl causal graph containing key variables. Nodes in the graph represent key variables in the system, such as {thermal urgency U, power allocation P, backoff time b, access delay D, final temperature deviation ΔT, total energy consumption E}. Directed edges between nodes represent direct causal relationships between variables. For example, U→P indicates that urgency directly affects power allocation decisions; P→ΔT indicates that power allocation directly affects temperature changes; U→b indicates that urgency also affects backoff strategies; b→D indicates that backoff time directly affects access delay. Based on this causal graph, the system can use do-operators to calculate the causal effects of specific interventions, separating direct from indirect effects between variables. For example, to analyze the impact of backoff time b on total energy consumption E, the system will find that b not only indirectly affects energy consumption by influencing access delay D (b→D→E), but may also have a direct impact because the system design associates high power consumption with long backoff times. By using the do-operator for intervention analysis P(E|do(b)), the confounding effects of other paths can be eliminated, revealing the pure causal contribution of backoff time to energy consumption. Based on this, the system can identify the root cause of performance bottlenecks. For example, if the analysis finds that the access latency D is mainly contributed by the direct effect (DE) of backoff time b, rather than the indirect effect (IE) of other factors such as network load, then the design of the current backoff algorithm itself is the main cause of excessive latency, and subsequent parameter optimization should focus on backoff-related parameters.

[0112] After accurate attribution, the system initiates multi-objective weight optimization and parameter updates. This process adjusts a large number of adjustable parameters within the system (e.g., the four-dimensional scoring weights w in Example 3). _q w _l w _e w _t (The retreat coefficients α, β, λ, etc. in Example 6) are used to minimize the weighted squared error loss function L=Σk between the KPI and the preset target. _i KPI _i -Target _i )². Since there may be constraints between the various KPIs (e.g., excessively pursuing low energy consumption may sacrifice thermal comfort), this optimization is a multi-objective optimization problem. The system can utilize Pareto front analysis to determine the relative importance of each KPI objective (i.e., the weight k). _i The system employs efficient search algorithms, such as Bayesian optimization, to find new optimal parameter combinations. Bayesian optimization constructs a Gaussian process model of the relationship between parameters and performance, and uses a sampling function (such as the expected improvement in EI) to intelligently select the next most promising parameter point for trial, finding or approximating the global optimum with fewer iterations. Based on this, the optimized new parameter set is synchronously updated to the corresponding modules in the system to guide the operation of the next control cycle, forming a complete closed-loop adaptive control process from execution to evaluation to attribution to optimization.

[0113] Example 11 provides an optional implementation of dynamic adaptive sampling. This example describes how the system intelligently adjusts the sensitivity of its sensors, that is, dynamically adjusts the sampling frequency of data such as the thermal state of the device, in order to achieve the best balance between control accuracy, communication overhead and device energy consumption.

[0114] In a specific embodiment, thermal state data of at least one device is acquired, and the sampling frequency is dynamically adjusted. The dynamic adjustment of the sampling frequency depends on the evaluation of one or more system state indicators. The system state indicators include at least: thermal indicators reflecting the dynamic physical state of the device, and communication indicators reflecting the channel quality of the communication network.

[0115] The static sampling frequency setting strategy is inefficient: when the device is in a stable state, high-frequency sampling generates a large amount of redundant data, wasting the device's communication bandwidth and battery power (if applicable); while when the device state changes drastically, low-frequency sampling may miss critical state transition points, leading to control lag. Therefore, this embodiment adopts a dynamic adaptive sampling strategy.

[0116] The preferred implementation is to use the optimal sampling frequency f _s_optimal The problem is thus defined as a multi-objective optimization problem. Its objective function can be expressed as: f_s_optimal =argmin(λ _1 ·U _model +λ _2 ·C _sampling -λ _3 ·IG _required ); where f _s_optimal It is the optimal sampling frequency obtained by optimization; λ _1 , λ _2 , λ _3 The weighting coefficients for the three optimization objectives; U _model This represents model uncertainty. The value is a weighted average of the mean square error (MSE) of predictions from each prediction model within the system (such as the thermal model in Example 8). As the model's prediction error increases, the system's understanding of the actual state of the equipment deteriorates. _model The value of will increase, driving the system to increase the sampling rate to obtain more real data to correct the model; C _sampling This represents the sampling cost, which is the sampling frequency f. _s functions, C _sampling =f _s ×(E _sensor +E _transmit +E _process ), where E _sensor E _transmit E _process These represent the sensing, transmission, and processing energy consumption involved in a single sampling, respectively. This item is used to constrain the total resource consumption of the sampling behavior; IG _required Representing the information gain requirement, this value uses entropy from information theory to assess how much information a new sample is expected to bring to the system, i.e., to what extent it can reduce the system's uncertainty about the true state of the device. When the device state changes gradually, the information gain from the new sample is small; conversely, when the state changes drastically, the information gain is large.

[0117] Solving the optimization problem allows the system to find a balanced sampling frequency that simultaneously minimizes model uncertainty and sampling cost while maximizing information gain. Furthermore, the system can be layered with a rule-based event-driven adjustment mechanism to handle unexpected situations: adjustment based on thermal indicators: when the absolute value of the rate of change of equipment temperature |dT is detected... *When the temperature drops sharply (e.g., a user starts using a lot of hot water, causing a sudden drop in temperature), the system will immediately ignore the current optimization results and forcibly increase the sampling frequency of the thermal sensor to a high level temporarily (e.g., from the usual 0.2Hz to 5Hz) and maintain it for a period of time (e.g., 30 seconds) to accurately capture this dynamic process. Based on communication metrics: When the network collision rate (CollisionRate) exceeds the congestion threshold, the system will proactively reduce the reporting frequency of non-critical state data (e.g., from 1Hz to 0.2Hz) to reduce device activity and avoid further exacerbating network congestion. Based on stable state: When the KPI metrics (e.g., in Example 10) are very stable for multiple consecutive control cycles and the variance is less than the minimum value, the system will enter a maintenance mode, reducing all sampling rates to the lowest baseline level to maximize energy savings.

[0118] The system outputs a new sampling frequency command f _s_new It will be done through a smooth transition (e.g., f) _s_new =0.7×f _s_current +0.3×f _s_optimal This is done by executing commands to avoid sudden changes in sampling frequency from impacting the system. These commands are then fed back to the data acquisition module in step one to complete the dynamic, closed-loop adjustment of the system's sensing capabilities.

[0119] Example 12 provides a specific numerical calculation case of a building energy consumption control method, which instantiates the abstract concepts and algorithm flow in the aforementioned multiple examples, and clearly demonstrates how the various technical features of the present invention work together.

[0120] Scenario: In an office building, a LoRa gateway is managing three 5kW electric water heaters, labeled Device A, Device B, and Device C. At this moment, the system needs to make a scheduling decision for Device A. Initial condition: Target temperature T _set =95℃; Ambient temperature T _ambient =20℃; Current equipment temperature: T * _A =75℃, T * _B =94℃, T * _C =90℃; Equipment temperature change rate: dT * _A / dt=-0.01℃ / s,dT * _A / dt=-0.005℃ / s,dT * _C / dt=-0.008℃ / s; User event: Device A has just received an emergency heating command, weight w _A =1.0, the time difference from the current time Δt _A =10s; No user events in devices B and C; Urgency calculation parameters: α=0.5, β=0.2, γ=0.3; Attenuation constant λ=300s; Candidate communication links: Link 1 (currently bound), Link 2 (alternate); Its communication quality Q, load balancing L, and energy efficiency E scores are known.

[0121] Step S12.1: Calculate the thermal urgency of device A (refer to Example 2);

[0122] Static temperature difference dimension score = α × (T) _set -T * _A )+=0.5×(95-75)=10.0; Dynamic change dimension score=β×|dT * _A / dt|=0.2×|-0.01|=0.002; Immediate demand dimension score=γ×w _A ·exp(-Δt _A / λ)=0.3×1.0×exp(-10 / 300)≈0.3×0.967=0.29; Total thermal stress U of equipment A _thermal_A =10.0 + 0.002 + 0.29 = 10.292; Calculate the U of devices B and C using the same method. _thermal The value will be much lower than that of device A (due to the small temperature difference and no demand events).

[0123] Step S12.2: Evaluate the communication link (refer to Example 3);

[0124] Assume that the (Q, L, E) score of link 1 is (0.9, 0.8, 0.7) and the score of link 2 is (0.8, 0.9, 0.9); assume the thermal synergy weight w _t =0.2, other weights are 0.267; calculate the thermal synergy score T(k, g) of device A on the two links; due to U _thermal_A Very high, its T _urgency The score is close to 1; assuming the gateway where link 1 is located has sufficient power, and the gateway where link 2 is located has limited power, then the T of link 1 _capacity Link 1 has a higher score than Link 2; switching to Link 2 incurs a switching penalty; assuming Link 1's T-score for device A is 0.8 and Link 2's is 0.4; the overall score U... _total :U _total_1 =0.267×(0.9+0.8+0.7)+0.2×0.8=0.64+0.16=0.80; U _total_2=0.267×(0.8+0.9+0.9)+0.2×0.4=0.69+0.08=0.77; Conclusion: Link 1 has a higher score, and the system tends to maintain the current binding.

[0125] Step S12.3: Perform incremental scanning and token generation (refer to Example 4);

[0126] Since only device A requires emergency service, the scanning process only evaluates link 1; assuming that admitting link 1 requires a communication cost ΔC. _RF =10mJ, power cost ΔP=5000W (start-up heating); assuming total budget B _RF =100mJ, B _P =15000W; Check budget: 10mJ < 100mJ and 5000W < 15000W, conditions are met; Check marginal utility: As it is the first choice, the utility is huge, far exceeding the threshold; The final selection set only contains link 1; After scanning, budget margin: ΔB _RF =90mJ, ΔB _P =10000W; Let the token granularity be C. _unit =5mJ, P _step =100W; Generate Token: Token _RF_total =floor(90 / 5) = 18 tokens; _P_total =floor(10000 / 100)=100.

[0127] Since only device A has high urgency, it will be prioritized for allocation of most power tokens, for example, it will receive 50 power tokens (corresponding to a 5000W power authorization).

[0128] Step S12.4, Scheduling and Conflict Handling (refer to Example 6);

[0129] The heating control task of equipment A is due to its extremely high U _thermal_A and the power tokens held, their mixed priority _hybrid The calculation will be very high; the system schedules it in a protected, low-collision downlink time slot and sends a closing command to its smart socket; assuming that at this time, a collision occurs in the routine status reporting task (low priority) of device C, with n=1; the system calculates the backoff time b for device C. At this time, the system power budget is tight (because A is heating), P _remain / B _P The smaller value results in a power tension component λ·(1-P) _remain / B _P The value of ) is relatively large; therefore, the backoff time b of device C will be longer than when the system is idle, actively avoiding the critical business cycle of device A.

[0130] In a specific embodiment, the present invention successfully transforms the urgent physical needs of device A into a series of high-priority scheduling decisions. While meeting its needs, it enables other non-urgent tasks to be avoided in an orderly manner through adaptive conflict management, thereby achieving efficient, stable and intelligent operation of the entire system.

[0131] Example 13 provides an optional implementation of a bidirectional adjustment mechanism for thermal deviation and communication congestion. During the scheduling execution process of this invention, the system not only passively executes the preset scheduling scheme but also performs real-time, dynamic adjustments at the microscopic scale to cope with sudden changes in the state of the physical and communication worlds. This adjustment mechanism is embodied in bidirectional adjustment logic based on thermal deviation and communication congestion states.

[0132] Step S13.1: The system will perform dynamic priority boosting based on thermal deviation. Specifically, while executing the scheduling scheme, the system will frequently monitor the thermal status of each device. When the system detects a thermal deviation ΔT=|T of a certain device... * -T _set |Suddenly increased and exceeded the preset attention threshold ΔT _threshold When the temperature reaches 2°C, the system will determine that the device's control may have malfunctioned or encountered unforeseen strong interference. In this case, the system will temporarily and dynamically increase the scheduling priority of all communication tasks related to the device's thermal control (e.g., emergency reporting, forced power regulation commands). This increase can be based on the Priority calculated in Example Six. _hybrid Based on this, a dynamic gain factor greater than 1 is applied, or the task is directly inserted at the front of the scheduling queue. This ensures that when the physical control effect does not meet expectations, related communication can be carried out with the highest priority, allowing the system to intervene and correct the situation as quickly as possible, thus guaranteeing the robustness of the control.

[0133] Step S13.2: The system will perform dynamic suppression of non-urgent tasks based on communication congestion. Preferably, the system will continuously monitor congestion indicators of the current communication link, such as collision rate and ACK confirmation queue length. When these congestion indicators exceed a preset congestion threshold, the system will determine that the current channel resources are highly strained. To prevent non-critical services from exacerbating network congestion and affecting truly high-priority urgent tasks, the system will selectively and actively suppress a portion of traffic. The targets of suppression are those with thermal urgency U. _thermal Below a certain non-emergency threshold U _lowNon-critical communication tasks initiated by devices (e.g., routine log reporting, periodic heartbeats, etc.) can be suppressed. Suppression methods include temporarily prohibiting the transmission of these tasks or reducing their duty cycle to decrease their transmission frequency. When communication resources are strained, limited resources can be concentrated on ensuring the most critical services, demonstrating the system's intelligent contraction and resource focus capabilities under communication constraints.

[0134] This two-way adjustment mechanism enables the scheduling and execution process to have higher dynamic adaptability. It can respond quickly when there are deviations in physical control and actively avoid congestion in the communication network. In complex and ever-changing environments, it further improves the stability and efficiency of the entire building energy consumption control system.

[0135] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A building energy consumption control method, characterized in that, It is applied to at least one device connected to a communication network, including: Acquire thermal state data for at least one device; Based on thermal state data, a quantitative thermal urgency index is generated to characterize the thermal requirements of the equipment. Based on the quantitative thermal urgency index, communication resources allocated to at least one device are scheduled to generate a communication resource scheduling scheme. Execute communication resource scheduling schemes to control equipment; The scheduling of communication resources allocated to at least one device is based on a quantitative thermal urgency index, including: establishing a multi-dimensional evaluation system for multiple candidate communication links for at least one device; the multi-dimensional evaluation system includes at least a thermal synergy dimension, and its evaluation result is determined by the quantitative thermal urgency index; evaluating each candidate communication link using the multi-dimensional evaluation system to generate its respective comprehensive score; and determining the target communication link from multiple candidate communication links based on the comprehensive score to form a communication resource scheduling scheme. The scheduling of communication resources allocated to at least one device is based on a quantitative thermal urgency index, including: determining a mixed priority for each task in a series of tasks to be scheduled; the determination of the mixed priority incorporates a service level factor characterizing the service level of the task and a thermal state factor determined by the quantitative thermal urgency index; and scheduling the series of tasks to be scheduled based on the mixed priority to generate a communication resource scheduling scheme. The quantitative thermal urgency index includes: a static temperature difference dimension representing the difference between the current temperature and the set temperature; a dynamic change dimension representing the rate of temperature change over time; and a dimension representing the immediate needs triggered by the user's immediate behavior.

2. The method according to claim 1, characterized in that, The multidimensional evaluation system can evaluate each candidate communication link in terms of thermal synergy, which can be achieved by generating thermal synergy scores. The generation of thermal synergy scores includes: Taking into account at least three evaluation factors related to candidate communication links, including: The first evaluation factor characterizes the urgency of the task, and its magnitude is positively correlated with the quantitative thermal urgency index. The second evaluation factor characterizes the power adequacy, and its magnitude is positively correlated with the available power margin of the candidate communication link. The third evaluation factor characterizes the handover cost, and a handover penalty is imposed when a candidate communication link needs to be switched from the current communication link.

3. The method according to claim 1 or 2, characterized in that, The target communication link is determined based on the comprehensive score, including: Initialize the cumulative communication cost and cumulative power cost; Multiple candidate communication links are evaluated progressively based on their overall scores, from highest to lowest. If including the next candidate communication link in the evaluation would cause the cumulative communication cost to exceed the preset communication energy consumption budget, or cause the cumulative power cost to exceed the preset equipment power budget, the evaluation process will be terminated. Candidate communication links that were included in the selection before the end of the evaluation process were identified as target communication links.

4. The method according to claim 3, characterized in that, After the evaluation process is terminated, it also includes: Based on the difference between the communication energy consumption budget and the final cumulative communication cost, a communication token is generated to authorize communication access opportunities; Based on the difference between the device power budget and the final cumulative power cost, a power token is generated to authorize the device's power usage.

5. The method according to claim 1, characterized in that, Also includes: During the execution of the communication resource scheduling scheme, when a communication conflict is detected, an adaptive backoff time is calculated for the scheduling tasks involved in the conflict. Determining the retreat time includes: System energy state considerations include the fact that as the available remaining power budget of the system decreases, the backoff time increases accordingly. The thermal state factor of the task to be scheduled, wherein the higher the quantitative thermal urgency index of the task, the shorter the maximum allowable backoff time.

6. The method according to claim 5, characterized in that, The calculation of adaptive backoff time is based on a combination of at least three technical components: The exponential retreat component is exponentially related to the number of historical conflict failures; Queue congestion components that are positively correlated with the network estimated queue length; The power stress component is determined by the system's remaining power budget.

7. The method according to claim 1, characterized in that, A comprehensive quantification of immediate demand dimensions, including: Based on historical demand data, predictive demand intensity is generated through a predictive model. Calculate the intensity of immediate demand based on real-time user demand event data; By combining the predicted demand intensity with the immediate demand intensity, the final quantitative result of the immediate demand dimension is obtained.

Images