An internet of things building energy consumption dynamic monitoring and intelligent control system

Abstract

The application discloses a kind of building energy consumption dynamic monitoring and intelligent control system of Internet of Things, belong to monitoring and control technical field, including Internet of Things perception module, storage and analysis module, strategy generation module, heterogeneous execution module and update module, based on building energy consumption monitoring demand equipment list generates initial energy consumption control instruction set, and collects building dynamic data;Storage and analysis module utilizes energy consumption baseline model and comfort constraint model, output calibrated energy consumption trend and comfort constraint condition, generate correction coefficient matrix;Strategy generation module establishes double objective function and solves Pareto optimal solution, obtains energy consumption intelligent control scheme;Heterogeneous execution module resolves scheme into device operation instruction and controls, records deviation value and feedback;Update module updates correction coefficient matrix by full-link iteration mechanism, calculates energy consumption and comfort deviation rate, the application realizes building energy consumption dynamic monitoring and intelligent control, improves energy-saving efficiency and comfort.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of monitoring and regulation, and particularly relates to a building energy consumption dynamic monitoring and intelligent regulation system of Internet of Things. BACKGROUND

[0002] With the acceleration of urbanization and the continuous expansion of building scale, the proportion of building energy consumption in total social energy consumption is increasing year by year. The traditional building energy consumption management mode mostly relies on manual inspection and experience regulation, and has problems such as non-real-time energy consumption monitoring, low regulation accuracy, and unsatisfactory energy saving effect. Although some building energy consumption monitoring systems have appeared at present, these systems often have defects such as incomplete data collection, weak data analysis capability, and single regulation strategy, and cannot realize dynamic monitoring and intelligent optimization regulation of building energy consumption, so it is difficult to maximize energy consumption while ensuring the comfort of people in the building. Therefore, a system that can comprehensively consider building energy consumption data, environmental parameters and personnel status, realize dynamic monitoring and intelligent regulation is urgently needed. SUMMARY

[0003] In view of the deficiencies of the prior art, the application provides a building energy consumption dynamic monitoring and intelligent regulation system of Internet of Things, which comprises an Internet of Things sensing module, a storage and analysis module, a strategy generation module, a heterogeneous execution module and an updating module, generates an initial energy consumption regulation instruction set based on a building energy consumption monitoring demand equipment list, and collects building dynamic data; the storage and analysis module uses an energy consumption baseline model and a comfort constraint model to output calibrated energy consumption trend and comfort constraint conditions, and generates a correction coefficient matrix; the strategy generation module establishes a double-objective function and solves a Pareto optimal solution to obtain an energy consumption intelligent regulation scheme; the heterogeneous execution module decomposes the scheme into device operation instructions and regulates, records deviation values and feeds back; the updating module updates the correction coefficient matrix through a full-link iteration mechanism, calculates energy consumption and comfort deviation rates, and the application realizes building energy consumption dynamic monitoring and intelligent regulation, improves energy saving efficiency and comfort.

[0004] To achieve the above object, the application provides the following technical scheme.

[0005] A building energy consumption dynamic monitoring and intelligent regulation system of Internet of Things, comprising an Internet of Things sensing module, a storage and analysis module, a strategy generation module, a heterogeneous execution module and an updating module.

[0006] Based on a building energy consumption monitoring demand equipment list, an initial energy consumption regulation instruction set is generated, and meanwhile, a plurality of types of sensor nodes deployed by the Internet of Things sensing module are used to collect building dynamic data; the building dynamic data comprises building basic energy consumption data, environmental parameters and personnel status data.

[0007] Based on the building dynamic data, the calibrated energy consumption trend and comfort constraint condition are output through the built-in energy consumption baseline model and comfort constraint model of the storage and analysis module; the energy consumption baseline model combines the historical execution data of the initial energy consumption regulation instruction set to generate a correction coefficient matrix;

[0008] Based on the energy consumption trend prediction result and comfort constraint condition, a double-objective function is established through the optimization engine of the strategy generation module, and an energy consumption intelligent regulation scheme is generated by solving the Pareto optimal solution;

[0009] Based on the generated energy consumption intelligent regulation scheme, the energy consumption intelligent regulation scheme is disassembled into device operation instructions and regulated through the hierarchical control network constructed by the heterogeneous execution module, while recording the instruction execution deviation value and feeding back the execution result;

[0010] Based on the feedback execution result and instruction execution deviation value, the correction coefficient matrix is updated and the energy consumption and comfort deviation rate is calculated through the full-link iteration mechanism of the update module.

[0011] Specifically, the generation process of the initial energy consumption regulation instruction set includes:

[0012] Based on the device type and corresponding functional partition attribute recorded in the building energy consumption monitoring demand device list, a device-partition mapping table containing the correspondence between devices and functional partitions is constructed through a hash mapping algorithm; the device-partition mapping table includes device ID, partition ID, device type label and cross-partition association weight;

[0013] According to the rated power, historical running reference value and preset energy consumption threshold of each device in the device-partition mapping table, initial running parameter instructions are generated for each device;

[0014] Based on the initial running parameter instructions and the cross-partition association weight in the device-partition mapping table, a directed graph topology is used to construct an instruction dependency relationship tree to generate an initial energy consumption regulation instruction set containing device ID, execution time window and operation parameter.

[0015] Specifically, the initial running parameter is the ratio of the historical running reference value and the rated power of the device, multiplied by the rated power of the device and the threshold margin coefficient;

[0016] The threshold margin coefficient is the difference between 1 and the energy consumption ratio; the energy consumption ratio is the ratio of the current partition cumulative energy consumption to the preset energy consumption threshold.

[0017] Specifically, the Internet of Things sensing module includes a time slot scheduling unit and an edge preprocessing unit;

[0018] According to the building energy consumption monitoring demand device list, the time slot scheduling unit activates the monitoring node corresponding to the building energy consumption monitoring demand device list, generates a dynamic time slot allocation table based on the RSSI value and collision detection result in the monitoring node response frame, and adopts an adaptive time slot allocation strategy; the dynamic time slot allocation table contains a node ID, a time slot number, a transmission period, and a channel frequency;

[0019] The edge preprocessing unit pre-processes the collected building dynamic data according to the dynamic time slot allocation table by adopting an outlier detection algorithm, and generates a target energy consumption data space; the target energy consumption data space includes real-time power, running state code, environmental temperature and humidity, personnel density, and data confidence markers of each device; the target energy consumption data space is packaged in JSON format and attached with a check code.

[0020] Specifically, the generation process of the dynamic time slot allocation table includes:

[0021] Based on the building energy consumption monitoring demand device list, the time slot scheduling unit sends an activation signal to each monitoring node, receives a response state frame returned by the node, and generates a node initial response state table;

[0022] Based on the node initial response state table, a collision prediction algorithm based on Markov chain is adopted to analyze the collision probability distribution of node data transmission, and the time slot length and allocation ratio are dynamically adjusted in combination with the priority weight of the node energy consumption data;

[0023] According to the adjusted time slot length and allocation ratio, binary expansion and reallocation are performed on the collision time slot, a non-conflict dynamic time slot allocation table is generated through the mapping relationship between the node ID hash value and the time slot number;

[0024] The time slot utilization rate of the dynamic time slot allocation table is monitored in real time, and when the monitored time slot utilization rate exceeds the preset maximum utilization rate or is lower than the preset minimum utilization rate, a time slot adaptive adjustment mechanism is triggered and the dynamic time slot allocation table is updated.

[0025] Specifically, the acquisition process of the target energy consumption data space includes:

[0026] The edge preprocessing unit receives the original building dynamic data transmitted by each monitoring node according to the dynamic time slot allocation table, detects outliers by adopting an isolation forest algorithm, and obtains abnormal building dynamic data and normal building dynamic data;

[0027] The abnormal building dynamic data is repaired by adopting cubic spline interpolation, and the normal building dynamic data is compressed and denoised by wavelet transform, and the processed building dynamic data is integrated;

[0028] Based on the device-zone mapping table, the processed building dynamic data of the same type of device in the same area is spatio-temporally aligned, and a data confidence marker is added to generate a target energy consumption data space;

[0029] The data confidence marker is based on a comprehensive evaluation of data transmission signal-to-noise ratio and sampling frequency

[0030] Specifically, the storage and analysis module includes a data preprocessing unit and a model operation unit.

[0031] The data preprocessing unit takes the target energy consumption data space as input, preprocesses the building dynamic data, and generates standardized building dynamic data.

[0032] The model operation unit loads the energy consumption baseline model, inputs the standardized building dynamic data and the historical execution data of the initial energy consumption regulation instruction set into the energy consumption baseline model, calculates the energy consumption benchmark value through the sliding window algorithm, loads the comfort constraint model, maps the environmental parameters and personnel state data in the standardized building dynamic data to quantitative comfort indicators, and generates comfort constraint conditions.

[0033] A correction coefficient matrix is generated based on the benchmark value deviation rate of the energy consumption benchmark value and the actual energy consumption. The correction coefficient matrix includes time decay coefficient, device type coefficient and partition weight coefficient.

[0034] The model operation unit combines the correction coefficient matrix to calibrate the energy consumption trend, and outputs the calibrated energy consumption trend and comfort constraint conditions.

[0035] Specifically, the strategy generation module includes a scheme evaluation unit. The scheme evaluation unit uses NSGA-II algorithm to solve Pareto optimal solution, generates N groups of candidate regulation schemes, selects the optimal regulation scheme through entropy weight coefficient method, and outputs the energy consumption intelligent regulation scheme. The energy consumption intelligent regulation scheme includes device operation sequence, execution priority, expected energy consumption and comfort value.

[0036] Specifically, the heterogeneous execution module includes a hierarchical control network unit and an instruction execution unit.

[0037] The hierarchical control network unit is based on the device-zone mapping table to construct a three-level architecture of central controller-zone gateway-device controller. The central controller is used to receive the energy consumption intelligent regulation scheme and distribute it to the corresponding zone gateway. The zone gateway is used for instruction analysis and conflict detection. The device controller is bound to the device.

[0038] The instruction execution unit takes the energy consumption intelligent regulation scheme as input, and decomposes the scheme into device operation instructions according to the hierarchical control network topology, and then sends the device operation instructions to the device controller after protocol conversion, and synchronously records the device operation instruction sending time, receiving state and execution feedback.

[0039] Specifically, the instruction decomposition process of the instruction execution unit includes:

[0040] Obtain the energy consumption intelligent regulation scheme, the device-zone mapping table and the hierarchical control network topology.

[0041] Extract the device operation sequence in the energy consumption intelligent regulation scheme, then extract the target device ID of each instruction in the operation sequence, then query the device zone ID according to the device-zone mapping table, and then group the device operation sequence according to the zone ID to generate the zone-grouped operation sequence; the zone-grouped operation sequence is in the format of zone-instruction list key-value pair.

[0042] Traverse the zone-grouped operation sequence, read the instruction ID, and query the predecessor and successor instructions of the corresponding instruction based on the instruction dependency relationship tree, then add constraint marks to each predecessor and successor instruction to obtain the zone operation sequence with constraint marks.

[0043] Extract the communication protocol type corresponding to each device from the hierarchical control network topology, then load the protocol mapping table, convert the abstract operation parameters into device protocol instructions, and add the device controller physical address to the converted device protocol instructions to generate the zone instruction set with protocol format.

[0044] Add a check code and a timeout retransmission parameter to each device protocol instruction in the zone instruction set with protocol format to obtain the device operation instruction set divided according to zones.

[0045] Compared with the prior art, the present application has the following advantages:

[0046] The present application provides a building energy consumption dynamic monitoring and intelligent regulation system of Internet of Things, and optimizes and improves the architecture, operation steps and process, and the system has the advantages of simple process, low investment and operation cost, and low production cost.

[0047] The application provides a building energy consumption dynamic monitoring and intelligent control system of Internet of Things, which realizes comprehensive collection of building dynamic data through an Internet of Things sensing module, combines an energy consumption baseline model and a comfort constraint model of a storage and analysis module, and can accurately output calibrated energy consumption trends and comfort constraint conditions. A double-objective function optimization mechanism of a strategy generation module generates an optimal energy consumption control scheme under the premise of ensuring personnel comfort, effectively realizes fine management of building energy consumption, and improves energy saving efficiency; meanwhile, the system constructs a full-link closed-loop mechanism of generation-execution-feedback-update, a heterogeneous execution module ensures accurate landing of control instructions, and an update module continuously improves model prediction accuracy and control effect by iteratively optimizing a correction coefficient matrix. This combination of dynamic monitoring and intelligent control not only reduces building operation energy consumption costs, but also adaptively adjusts according to actual operating states. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 FIG. 1 is a system architecture diagram of a building energy consumption dynamic monitoring and intelligent control system of Internet of Things according to the application;

[0049] Figure 2 FIG. 2 is a principle flowchart of a building energy consumption dynamic monitoring and intelligent control system of Internet of Things according to the application. DETAILED DESCRIPTION

[0050] Embodiment 1

[0051] Please refer to Figure 1 and Figure 2 The application provides an embodiment: a building energy consumption dynamic monitoring and intelligent control system of Internet of Things, which comprises an Internet of Things sensing module, a storage and analysis module, a strategy generation module, a heterogeneous execution module and an update module.

[0052] Based on a building energy consumption monitoring demand equipment list, an initial energy consumption control instruction set is generated, and meanwhile, a plurality of types of sensor nodes deployed by the Internet of Things sensing module are used to collect building dynamic data; the building dynamic data comprises building basic energy consumption data, environmental parameters and personnel state data.

[0053] Further, the building basic energy consumption data comprises device-level data, partition-level data and building-level data, wherein the device-level data comprises real-time active power, reactive power, voltage, current, cumulative power consumption and running time; the partition-level data comprises functional partition total power consumption, peak power, average power factor and energy consumption composition ratio; and the building-level data comprises total energy consumption, partition energy consumption ratio, time-of-use energy consumption and historical same period comparison value.

[0054] Further, the environmental parameters include physical environmental parameters, spatial environmental parameters and meteorological parameters; the physical environmental parameters include temperature, relative humidity, light intensity, CO2 concentration; the spatial environmental parameters include partition area, layer height, orientation, door and window state; the meteorological parameters include outdoor temperature, humidity, rainfall, light intensity, which are obtained through a meteorological interface.

[0055] Further, the personnel state data includes: 1) personnel density: real-time number of people in the partition, personnel density; 2) activity state: static / dynamic proportion, average stay time; 3) behavior characteristics: historical peak period of entry / exit, personnel flow coefficient of each partition.

[0056] Based on the building dynamic data, the calibrated energy consumption trend and the comfort constraint condition are output through the energy consumption baseline model and the comfort constraint model built in the storage and analysis module; the energy consumption baseline model combines the historical execution data of the initial energy consumption regulation and control instruction set to generate a correction coefficient matrix;

[0057] Based on the energy consumption trend prediction result and the comfort constraint condition, a double-objective function is established through the optimization engine of the strategy generation module, and an energy consumption intelligent regulation and control scheme is generated by solving the Pareto optimal solution;

[0058] Further, the double-objective function solving process of the optimization engine includes:

[0059] (1) initialize the population, and each individual represents a combination of device operation parameters;

[0060] (2) calculate the objective function value of the individual, i.e. energy consumption and comfort;

[0061] (3) perform non-dominated sorting to determine the Pareto level of the individual;

[0062] Calculate the crowding distance to maintain population diversity, wherein the calculation formula of the crowding distance is the prior art content in the field, and is not the inventive scheme of the present application, which will not be repeated here;

[0063] (4) generate the next generation population through selection, crossover and mutation operations;

[0064] (5) repeat the calculation, sorting, distance calculation and population generation operations until the iteration number reaches the preset value or converges, extract the Pareto optimal solution set from the final population, and output multiple candidate schemes.

[0065] Based on the generated energy consumption intelligent regulation and control scheme, the hierarchical control network constructed by the heterogeneous execution module decomposes the energy consumption intelligent regulation and control scheme into device operation instructions and performs regulation and control, while recording the instruction execution deviation value and feeding back the execution result;

[0066] Based on the feedback of the execution result and the instruction execution deviation value, the updating module updates the correction coefficient matrix and calculates the energy consumption and comfort deviation rate through the full-link iteration mechanism.

[0067] Further, the recording process of the instruction execution deviation value includes:

[0068] (1) The device controller collects the actual operation parameters such as actual power and temperature in real time after executing the instruction;

[0069] (2) Calculate the deviation value; the deviation value is the difference between the actual and instruction parameter values, and the percentage obtained by comparing the instruction parameter value;

[0070] (3) Mark according to the deviation level: mark as excellent if the deviation value is less than or equal to h%, mark as good if the deviation value is greater than h% and less than or equal to s%, and mark as poor if the deviation value is greater than s%, and h < s;

[0071] (4) Generate an execution result feedback package; the execution result feedback package includes instruction ID, execution time, actual parameter, deviation value, deviation level, and device status code.

[0072] Further, the updating module includes a deviation analysis unit and a model iteration unit:

[0073] (1) The deviation analysis unit takes the execution result feedback package as input and calculates the energy consumption deviation rate; the energy consumption deviation rate is the difference between the actual energy consumption and the expected energy consumption, and the percentage obtained by comparing the expected energy consumption;

[0074] Calculate the comfort deviation rate; the comfort deviation rate is the difference between the actual comfort and the expected comfort, and the percentage obtained by comparing the expected comfort;

[0075] (2) The model iteration unit updates the correction coefficient matrix according to the deviation rate level based on the full-link iteration mechanism:

[0076] When the deviation rate is greater than 15%, adjust the type coefficient of the corresponding device by ±0.1;

[0077] When the deviation rate is greater than 10% for 3 consecutive times, trigger the energy consumption baseline model to retrain;

[0078] (3) Output the updated correction coefficient matrix and deviation analysis report.

[0079] Further, the full-link iteration mechanism includes:

[0080] Periodic iteration: perform full iteration based on all execution results of the day every 24 hours;

[0081] Trigger iteration: when the single device instruction execution deviation is greater than 10% for 3 consecutive times, immediately perform incremental iteration.

[0082] Further, the iteration process in the full-link iteration mechanism is:

[0083] (1) The deviation analysis unit outputs the energy consumption deviation rate and the comfort deviation rate;

[0084] (2) The model iteration unit locates the deviation source, including device failure, model error, and instruction conflict;

[0085] (3) For model error, adjust the corresponding field of the correction coefficient matrix;

[0086] (4) Generate an iteration log, including the coefficient values before and after adjustment, and the deviation rate change;

[0087] (5) Synchronize the updated correction coefficient matrix to the storage and analysis module.

[0088] Further, the deviation source positioning method includes constructing a deviation diagnosis rule based on a decision tree model; if the single device deviation rate is greater than 15% and the same type device is normal, it is determined as device failure; if the deviation rate of multiple devices in the same partition is greater than 10% and is reproduced across time periods, it is determined as model error; if the instruction execution time window overlaps and the deviation is negatively correlated, it is determined as instruction conflict; different types of deviations trigger corresponding processing mechanisms.

[0089] The strategy generation module includes a scheme evaluation unit; the scheme evaluation unit uses the NSGA-II algorithm to solve the Pareto optimal solution, generates N groups of candidate control schemes, selects the optimal control scheme through the entropy weight coefficient method, and outputs the energy consumption intelligent control scheme; the energy consumption intelligent control scheme includes device operation sequence, execution priority, expected energy consumption, and comfort value.

[0090] Embodiment 2

[0091] The generation process of the initial energy consumption control instruction set in this embodiment includes:

[0092] Based on the device type and corresponding functional partition attribute recorded in the device list of building energy consumption monitoring requirements, a device-partition mapping table containing the correspondence between devices and functional partitions is constructed through a hash mapping algorithm; the device-partition mapping table includes device ID, partition ID, device type label, and cross-partition association weight, wherein the hash mapping algorithm is a prior art in the art and is not the inventive scheme of the present application, and will not be described here.

[0093] According to the rated power, historical operation reference value, and preset energy consumption threshold of each device in the device-partition mapping table, an initial operation parameter instruction is generated for each device;

[0094] Further, according to the rated power, historical operation reference value and preset energy consumption threshold of each device in the device-zone mapping table, an initial operation parameter instruction is generated for each device, including: first, extracting and verifying the core parameters such as device rated power, historical operation reference value and zone energy consumption threshold from the device-zone mapping table; second, calculating the zone energy consumption balance coefficient to reflect the remaining energy consumption quota of the zone; then determining the load coefficient interval according to the device type and historical load characteristics to calculate the initial power parameter; then correcting the initial parameter by the zone energy consumption balance coefficient to ensure compliance with the overall energy consumption constraint of the zone; finally, combining the time period characteristics of historical operation, adding time window constraints to the parameter to form the initial operation parameter instruction.

[0095] Further, the zone energy consumption balance coefficient is equal to the difference between the preset energy consumption threshold and the daily zone cumulative energy consumption, and the quotient obtained by dividing the result by the preset energy consumption threshold.

[0096] Further, the initial power parameter is equal to the product of the rated power and the median value in the load coefficient interval.

[0097] Further, the initial power parameter is corrected by multiplying the initial power parameter by the zone energy consumption balance coefficient to obtain the corrected initial power parameter.

[0098] Based on the initial operation parameter instruction and the cross-zone correlation weight in the device-zone mapping table, a directed graph topology is used to construct an instruction dependency tree to generate an initial energy consumption control instruction set containing device ID, execution time window and operation parameter. Specifically, it includes: first, extracting and standardizing the cross-zone correlation weight from the device-zone mapping table to identify the predecessor-successor dependency relationship between devices; second, constructing a directed graph with devices as nodes and dependency relationships as directed edges, and dividing the execution hierarchy by topological sorting to eliminate circular dependencies; then, based on the hierarchical relationship and correlation weight, the execution time window of each device is calculated to ensure the time coordination of dependent devices; finally, the instruction set is integrated with device ID, time window and operation parameter to form the final initial energy consumption control instruction set.

[0099] Further, identifying the predecessor-successor dependency relationship between devices includes: traversing all initial operation parameter instructions of devices, and identifying the direct dependency relationship between devices combined with the standardized correlation weight. For example, if the correlation weight between the water chiller and the circulating water pump is B, which is greater than or equal to the preset weight threshold C, it is determined that the water chiller is the predecessor device of the circulating water pump, i.e. the circulating water pump can only run after the water chiller starts; if the correlation weight between the lighting system and the human body sensor is D, and D is greater than or equal to C, it is determined that the human body sensor is the trigger device of the lighting system, i.e. the lighting start needs to be based on the sensor detecting personnel; for device pairs with correlation weight less than the preset weight threshold C, mark them as having no direct dependency.

[0100] Further, the construction process of the directed graph includes: taking the device ID as the node and the identified dependency relationship as the directed edge to construct the directed graph, wherein the predecessor device points to the successor device, such as the water chiller-circulating water pump, and the weight of the edge is the standardized correlation weight; for the cross-partition device, such as the central air conditioner host and the end air disc of each partition, the partition identifier is added in the directed edge, such as host-air disc, to clearly indicate the spatial relationship of the cross-partition dependency.

[0101] Further, the execution hierarchy is divided by topological sorting to eliminate the circular dependency, including: adopting Kahn algorithm to perform topological sorting on the directed graph to eliminate the circular dependency, and after sorting, dividing into multiple levels according to the device execution order: level 1 is the root device without predecessor dependency, such as the total power switch; level 2 is the device dependent on level 1, such as the central air conditioner host; level E is the device dependent on level E-1, such as the end air disc; the devices in each level have no direct dependency and can execute instructions in parallel.

[0102] Further, the execution time window of each device is calculated based on the level relationship and the correlation weight, including: based on the level relationship after topological sorting, the initial running parameter instruction of each device is allocated with an execution time window; the time window of the level 1 device starts at the initial time of system scheduling, and the time length is determined according to the preset running time length in the initial running parameter instruction; the time window of the level 2 device starts at the time window end time of all its predecessor devices plus the buffer time, wherein the buffer time is equal to the product of the correlation weight and the preset 60 seconds, and the higher the weight, the shorter the buffer time; and the same applies to the subsequent devices, so as to ensure that the successor device executes the instruction after all the predecessor devices are normally started, for example, the time window of the water chiller in level 2 is 00:05-08:05, and the correlation weight of the circulating water pump in level 3 is 0.8, so the time window starts at 00:05+0.8x60 seconds=00:05:48, and the time length is determined according to the parameter itself.

[0103] The initial running parameter is the ratio of the historical running benchmark value and the rated power of the device, multiplied by the rated power of the device and the threshold margin coefficient;

[0104] The threshold margin coefficient is the difference between 1 and the energy consumption ratio; the energy consumption ratio is the ratio of the current partition cumulative energy consumption to the preset energy consumption threshold.

[0105] The Internet of Things sensing module includes a time slot scheduling unit and an edge preprocessing unit.

[0106] The Internet of Things sensing module includes a time slot scheduling unit and an edge preprocessing unit.According to the building energy consumption monitoring demand device list, the monitoring node corresponding to the building energy consumption monitoring demand device list is activated by the time slot scheduling unit, and based on the RSSI value and the collision detection result in the monitoring node response frame, an adaptive time slot allocation strategy is used to generate a dynamic time slot allocation table; the dynamic time slot allocation table contains node ID, time slot number, transmission period and channel frequency, specifically including: first, extracting the monitoring node information from the device list and establishing a mapping table, awakening the node through the activation signal and collecting the response data; second, analyzing the RSSI value to divide the communication quality level, identifying the collision node through the test time slot; then, combining the node priority to dynamically calculate the time slot length, allocating the time slot according to the priority and generating an initial allocation table; then, checking and eliminating the time slot conflict to form the final dynamic time slot allocation table; finally, setting the update trigger condition, monitoring in real time and regenerating the time slot table if necessary to ensure efficient and collision-free data transmission.

[0107] Further, the RSSI (Received Signal Strength Indication) value, i.e. the received signal strength indication, is a quantitative indicator for measuring the signal strength received by the receiving end in wireless communication, usually in decibel milliwatts dBm, the numerical range is generally between -30dBm and -110dBm, the larger the value, the closer to 0, indicating that the received signal strength is stronger and the communication quality is better; the smaller the value, the weaker the signal, which may be caused by problems such as too far distance, obstacles blocking or interference, and the communication stability may decrease.

[0108] Further, the process of analyzing the RSSI value to divide the communication quality level includes: quantitatively analyzing the RSSI value in the received response frame, and dividing the node communication quality into three levels: excellent, medium and poor, wherein when the RSSI value is greater than or equal to the preset upper limit of the decibel threshold, it is rated as excellent, when the RSSI value is less than the preset upper limit of the decibel threshold and greater than or equal to the preset lower limit of the decibel threshold, it is rated as medium, and when the RSSI value is less than the preset lower limit of the decibel threshold, it is rated as poor; for the nodes that do not return the response frame, mark them as not activated, and resend the activation signal after K minutes, and at most retry 3 times.

[0109] Further, the process of identifying the collision node through the test time slot includes: the time slot scheduling unit presets a temporary time slot pool in the initial activation stage, and lets all activated nodes send test data frames in random time slots; by detecting the overlap degree of the data frames, recording the collision time slot number and the involved node ID, and generating a collision node list; for the nodes that have collided in the test for 3 times in a row, mark them as high collision nodes.

[0110] The edge preprocessing unit adopts an outlier detection algorithm according to the dynamic time slot allocation table to preprocess the collected building dynamic data, and generates a target energy consumption data space; the target energy consumption data space includes real-time power, running state code, environmental temperature and humidity, personnel density and data confidence mark of each device; the target energy consumption data space is packaged in JSON format and attached with a check code.

[0111] The generation process of the dynamic time slot allocation table includes:

[0112] Based on the building energy consumption monitoring demand equipment list, the time slot scheduling unit sends an activation signal to each monitoring node, receives the response state frame returned by the node, generates a node initial response state table, and specifically includes two parts of extracting monitoring node information from the equipment list and establishing a mapping table and awakening the node through the activation signal and collecting response data;

[0113] Further, extracting monitoring node information from the equipment list and establishing a mapping table includes: extracting the associated node ID, device type, preset sampling frequency and communication protocol type of all monitored devices from the building energy consumption monitoring demand equipment list, and establishing a node-device mapping table; each node ID is format checked and stored in groups according to the device type, which is convenient for subsequent targeted scheduling.

[0114] Further, awakening the node through the activation signal and collecting the response data includes: the time slot scheduling unit groups the device types, and sends a timestamped activation signal to all monitoring nodes in the node-device mapping table, wherein the activation signal contains the node ID, the awakening instruction, the synchronization clock information and the response timeout threshold; at the same time, start the listening mechanism to record the response frame returned by the node, including the node ID, the current battery voltage, the signal strength, whether it is in an idle state and the historical collision record.

[0115] Based on the node initial response state table, a collision prediction algorithm based on Markov chain is used to analyze the collision probability distribution of node data transmission, and the priority weight of node energy consumption data is combined to dynamically adjust the time slot length and allocation ratio, specifically including: first, analyzing the node initial response state table, quantifying the transmission state, response delay and signal quality; second, constructing a Markov chain state transition matrix to predict the collision probability distribution of future time slots; then, combining the collision probability to calibrate the priority weight of the node; then, dynamically adjusting the time slot length according to the average collision probability and signal quality, optimizing the time slot allocation ratio according to the priority; finally, checking the validity of the adjustment result, and through iterative optimization to ensure that the time slot resource allocation not only meets the needs of high-priority nodes, but also reduces the overall collision probability.

[0116] Further, the node initial response state table is parsed, and the transmission state, response delay and signal quality are quantified, including: structurally parsing the data in the node initial response state table, extracting the historical transmission state, response delay, RSSI value and battery power of each node; quantifying the transmission state into a discrete state value; dividing into fast, medium and slow three levels according to the response delay, respectively corresponding to the weights 0.9, 0.6 and 0.3; and combining the RSSI value to determine the signal quality coefficient.

[0117] Further, the process of constructing the Markov chain state transition matrix includes: taking the state sequence of the node in the past 100 transmissions as a sample to construct a three-order state transition matrix, wherein the elements in the three-order state transition matrix are representing the probability of the next state being j when the current state is i, and For example, if any node is converted from a successful state to a collision in 100 transmissions, i.e. the number of times of converting 0 to 1 is 15, then For a new node with a sample size less than 100, the average transition matrix of the same device type is used as the initial value.

[0118] Further, the process of predicting the collision probability distribution of the future time slot includes: based on the constructed Markov chain state transition matrix, using the Markov chain prediction model to calculate the collision probability of the next 10 time slots of the current time slot, the prediction process is: taking the current state of the node as the initial state, iteratively calculating the state probability distribution of each time slot in the next 10 time slots of each current time slot by matrix multiplication, extracting the probability value of the collision state to form a collision probability sequence, and for a high collision node, increasing the prediction probability by 10% as a correction value.

[0119] Further, the time slot length is dynamically adjusted and set based on the node communication quality and collision frequency.

[0120] Further, the time slot priority allocation: combining the importance weight of the device type and the communication quality level, calculating the comprehensive priority of each node: the comprehensive priority is equal to the product of the device type weight and the communication quality weight; sorting according to the comprehensive priority from high to low, preferentially allocating time slot resources to high priority nodes to ensure the timeliness of critical data transmission.

[0121] According to the adjusted time slot length and allocation ratio, binary expansion and reallocation are performed on the collision time slot, and through the mapping relationship between the node ID hash value and the time slot number, a dynamic time slot allocation table without collision is generated;

[0122] Further, according to the adjusted time slot length and allocation ratio, binary expansion and reallocation are performed on the collision time slot, and through the mapping relationship between the node ID hash value and the time slot number, a dynamic time slot allocation table without collision is generated, including:

[0123] (1) Traverse the adjusted initial time slot allocation table, extract all time slots marked as collisions, that is, time slots with 2 or more nodes allocated in the same time slot, record their time slot number, the list of node IDs involved and the original length of the time slot. At the same time, mark each collision time slot as a state to be expanded, and count the number of high priority nodes in it as the basis for determining the expansion priority.

[0124] (2) Perform binary expansion for collision time slots, that is, double the time slot length each time until all nodes in the time slot can be redistributed to sub-time slots without conflict. Assume that the expansion limit is G times the original length. If there is still a conflict after reaching the limit, the time slot is marked as a high-conflict time slot.

[0125] (3) The expanded time slots are divided according to the binary splitting principle. There are n sub-time slots, where n represents the number of expansions. The length of each sub-time slot is equal to the ratio of the total length after expansion to the number of sub-time slots, and the sub-time slot length must be greater than or equal to the minimum transmission time of the node.

[0126] (4) For each node within the collision time slot, extract its 16-bit unique ID, calculate its hash value using the SHA-1 hash algorithm, and take the last 8 bits of the hash value as the node's hash feature value; based on the number of sub-time slots, set a mapping rule: pair the hash feature values ​​with... The remainder obtained by taking the modulo is the sub-slot number that should be allocated to the node. The SHA-1 hash algorithm is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0127] (5) After allocating sub-time slots to all nodes in the collision time slot according to the mapping rules, check whether each sub-time slot is allocated only 1 node. If there are still sub-time slots allocated multiple nodes, then perform binary expansion again on the sub-time slot, that is, add one expansion on the basis of the original expansion, redivide the sub-time slots and allocate them according to the above mapping rules until all sub-time slots are free of conflict or reach the expansion limit.

[0128] (6) For high-conflict time slots that still have conflicts even after reaching the expansion limit, sort the nodes in them according to the calibrated priority. The node with the highest priority is retained in the first sub-time slot of the original time slot, and the remaining nodes are migrated to the adjacent free time slots in turn. If the adjacent time slots are occupied, a new time slot is temporarily created for the nodes of the high-conflict time slots that still have conflicts even after reaching the expansion limit, and inserted at the end of the time slot table. At the same time, it is marked as a temporary time slot.

[0129] (7) Integrate all non-collision time slots, non-conflicting sub-time slots after expansion, and temporary time slots in ascending order of time slot number to form a dynamic time slot allocation table. The dynamic time slot allocation table includes time slot number, sub-time slot number, corresponding node ID, time slot length, start timestamp, and transmission priority. At the same time, construct a node-time slot index table. The key in the node-time slot index table is the node ID, and the value is the corresponding time slot number and sub-time slot number.

[0130] The system monitors the time slot utilization rate of the dynamic time slot allocation table in real time. When the monitored time slot utilization rate exceeds the preset maximum utilization rate or falls below the preset minimum utilization rate, the system triggers the time slot adaptive adjustment mechanism and updates the dynamic time slot allocation table.

[0131] The process of acquiring the target energy consumption data space includes:

[0132] The edge preprocessing unit receives the original building dynamic data transmitted by each monitoring node according to the time slot based on the dynamic time slot allocation table, and detects outliers through the isolated forest algorithm to obtain abnormal building dynamic data and normal building dynamic data. The isolated forest algorithm is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0133] Abnormal building dynamic data is repaired using cubic spline interpolation, and normal building dynamic data is compressed and denoised using wavelet transform. The two are then integrated to form the processed building dynamic data. Cubic spline interpolation and wavelet transform are existing technologies in this field and are not the inventive solutions of this application, so they will not be described in detail here.

[0134] Based on the device-partition mapping table, the processed building dynamic data of the same type of device in the same area are spatiotemporally aligned and data confidence markers are added to generate a target energy consumption data space.

[0135] The data confidence level is determined based on a comprehensive evaluation of the data transmission signal-to-noise ratio and the sampling frequency.

[0136] The storage and analysis module includes a data preprocessing unit and a model computation unit;

[0137] The data preprocessing unit takes the target energy consumption data space as input, preprocesses the building dynamic data, and generates standardized building dynamic data.

[0138] The model computing unit loads the energy consumption baseline model, inputs the standardized building dynamic data and the historical execution data of the initial energy consumption control instruction set into the energy consumption baseline model, calculates the energy consumption benchmark value through the sliding window algorithm, and loads the comfort constraint model at the same time, maps the environmental parameters and personnel status data in the standardized building dynamic data into quantitative comfort indicators, and generates comfort constraint conditions.

[0139] Further, the training process of the energy consumption baseline model comprises: inputting the standardized building dynamic data of the past A days, the historical execution data of the initial energy consumption regulation instruction set; performing feature engineering to extract time features, environment features, and personnel features; training the model using a random forest regression algorithm, taking the energy consumption benchmark value as the label, and optimizing the hyperparameters through 5-fold cross-validation; when the model determination coefficient is greater than or equal to 0.85, it is determined to be qualified for training, otherwise, the feature dimension is increased for retraining, and the trained model is stored in PMML format.

[0140] It should be understood that the determination coefficient ranges from 0 to 1, and the closer the value is to 1, the better the model fits the data, and the higher the proportion of the model that can explain the dependent variable, i.e., the energy consumption benchmark value; the closer the value is to 0, the worse the model fitting effect, and the model cannot effectively explain the change rule of the data.

[0141] In the training process of the energy consumption baseline model, when the determination coefficient is greater than or equal to 0.85, it means that the model can explain more than 85% of the energy consumption benchmark value change and is determined to be qualified for training; if the threshold is not reached, the model needs to be optimized and retrained by increasing the feature dimension, etc., to improve the fitting ability and prediction accuracy of the model to the energy consumption data. The trained model is stored in PMML format for easy loading and application in the system.

[0142] Further, the quantification rules of the comfort constraint model include the quantification standards of temperature comfort, humidity comfort, and personnel density coefficient. In the present application, the temperature comfort is 1.0 when the temperature is in [24, 26]℃, 0.8 when in [22, 24)℃ or (26, 28]℃, 0.5 when in [20, 22)℃ or (28, 30]℃, and 0.2 when less than 20℃ or greater than 30℃, wherein ℃ represents Celsius temperature; the humidity comfort is 1.0 when the humidity is in [40, 60]%, 0.7 when in [30, 40)% or (60, 70)%, 0.4 when in [20, 30)% or (70, 80)%, and 0.1 when less than 20% or greater than 80%; the personnel density coefficient is 1.0 when the personnel density is less than or equal to 0.1 person per square meter, 0.8 when greater than 0.1 and less than or equal to 0.3 person per square meter, 0.6 when greater than 0.3 and less than or equal to 0.5 person per square meter, and 0.4 when greater than 0.5 person per square meter.

[0143] A correction coefficient matrix is generated based on the benchmark value deviation rate of the energy consumption benchmark value and the actual energy consumption; the correction coefficient matrix includes a time decay coefficient, a device type coefficient, and a partition weight coefficient.

[0144] Further, the correction coefficient matrix is generated based on the benchmark value deviation rate of the energy consumption benchmark value and the actual energy consumption, comprising:

[0145] (1) Extract the energy consumption benchmark value and actual energy consumption value in a statistical period from the storage and analysis module, align by device ID and time granularity, calculate the benchmark value deviation rate for each hour segment of each device; the deviation rate is equal to the difference between the actual energy consumption and the energy consumption benchmark value, and the quotient obtained by dividing the energy consumption benchmark value, if the energy consumption benchmark value is 0, it is marked as no benchmark data, and does not participate in subsequent calculation, in the present application, the statistical period is set to 24 hours;

[0146] (2) Take the current time as the reference point to obtain the historical deviation rate data, give different weights according to the time distance, set the time attenuation factor, in the present application, the time attenuation factor is set to 0.95, that is, daily attenuation of 5%;

[0147] (3) For each device, the weighted average value of the historical deviation rate is calculated according to the time attenuation factor, and the average value is mapped as the time attenuation coefficient;

[0148] (4) According to the device type, such as air conditioner, lighting, elevator, office equipment, etc., the device is partitioned, and the average benchmark value deviation rate of each group of devices after partitioning is calculated;

[0149] (5) According to the energy consumption importance and deviation characteristics of the functional partition, the partition weight coefficient is calculated, including: first, calculate the proportion of the total energy consumption of each partition in the total energy consumption of the building, then calculate the average deviation rate of all devices in the partition, wherein the ratio of the energy consumption ratio and the average energy consumption ratio is calculated to obtain the ratio result, and the sum of 1 and the partition deviation rate percentage is obtained to obtain the sum result, then the ratio result is multiplied by the sum result to obtain the partition weight coefficient, wherein the average energy consumption ratio is the arithmetic mean of all partition energy consumption ratios;

[0150] (6) A three-dimensional correction coefficient matrix is constructed with device ID as row index and time, device type, partition as column index; the value of each element in the three-dimensional correction coefficient matrix is the product of the time attenuation coefficient, the device type coefficient and the partition weight coefficient of the corresponding device;

[0151] (7) Set the correction coefficient range to [0.5-2.0], and perform truncation processing on the coefficients exceeding the range, that is, set the coefficients lower than 0.5 to 0.5, and set the coefficients higher than 2.0 to 2.0;

[0152] (8) Calculate the deviation of each coefficient from the historical same period coefficient, if the deviation exceeds 20%, trigger manual review to confirm whether it is caused by abnormal data such as device failure and sensor abnormality; if it is abnormal data, eliminate it and recalculate the coefficient; if it is a normal trend change, keep the adjusted coefficient;

[0153] (9) Add a timestamp and a version number to each generated correction coefficient matrix, store the historical versions for traceability, and verify the optimization effect on energy consumption trend prediction in the test environment after generating a new three-dimensional correction coefficient matrix. If the optimization effect is greater than or equal to 5%, the new three-dimensional correction coefficient matrix is officially put into effect in the next system update; otherwise, adjust the calculation parameters of each coefficient again.

[0154] The model operation unit calibrates the energy consumption trend in combination with the correction coefficient matrix, and outputs the calibrated energy consumption trend and comfort constraint condition, specifically including:

[0155] (A1) The model operation unit calls the energy consumption baseline model, inputs the standardized building dynamic data, generates an initial energy consumption trend prediction result, extracts the corresponding timestamp, device ID, and partition information, matches the index in the correction coefficient matrix, and ensures that each prediction value can find the corresponding correction coefficient;

[0156] (A2) Find the corresponding three-dimensional correction coefficient, i.e., the product of the time decay coefficient, the device type coefficient, and the partition weight coefficient, in the correction coefficient matrix using the device ID in the initial energy consumption trend as the key. For devices that do not match the coefficient, such as newly added devices, use the average correction coefficient of the type and partition to which they belong as a temporary value and mark them as to be calibrated. For matched coefficients, check their validity;

[0157] (A3) Calibrate each hour's predicted energy consumption value of each device using the corresponding correction coefficient, i.e., the calibrated device energy consumption equals the product of the initial predicted energy consumption and the correction coefficient;

[0158] (A4) Accumulate the calibrated energy consumption values of all devices in the same partition by hour to obtain the partition-level calibrated energy consumption trend. Then, accumulate the energy consumption values of all partitions to obtain the building-level calibrated energy consumption trend. During the aggregation process, mark the energy consumption values of the to-be-calibrated devices in the partition and calculate their proportion in the total energy consumption of the partition: if the proportion is less than or equal to the preset energy consumption proportion threshold, directly include it in the aggregation; if the proportion is greater than the preset energy consumption proportion threshold, use the historical average energy consumption deviation of the partition in the same period to fine-tune the total trend again;

[0159] (A5) Extract environmental parameters and personnel state data from the building dynamic data, input the comfort constraint model, and quantize the parameters into comfort indicators according to the quantization rules of the comfort constraint model to obtain a comprehensive comfort index;

[0160] (A6) Draw a constraint threshold based on the comprehensive comfort index: when the comprehensive comfort index is greater than or equal to the preset upper limit of the index threshold 0.8, the constraint condition is to maintain the current comfort, and the energy consumption fluctuation is allowed ± 10%; when the comprehensive comfort index is greater than or equal to the lower limit of the index threshold and less than the upper limit of the index threshold, the constraint condition is to improve the comfort to be greater than or equal to 0.8, and the energy consumption increase is less than or equal to 15%; when the comprehensive comfort index is less than the lower limit of the index threshold, the constraint condition is to preferentially improve the comfort to be greater than or equal to 0.7, and the energy consumption increase is less than or equal to 20%, in the application, the index threshold is set to [0.6, 0.8];

[0161] (A7) Check whether the calibrated energy consumption trend meets the comfort constraint condition: if the building-level total energy consumption increase is within the constraint allowed range, and each subzone comfort index meets the standard, it is determined that the verification is passed; if the energy consumption increase of any subzone is out of limit, repeat (A3)-(A4) until the conflict is eliminated; if the conflict cannot be eliminated, such as device performance limitation, a constraint exemption application is generated;

[0162] (A8) The calibrated device-level, subzone-level and building-level energy consumption trends are arranged in time sequence, the source of the correction coefficient of each data point and the calibration basis are marked, and the comfort constraint conditions are classified by subzone, the threshold range and adjustment priority of each index are clear, and finally the complete results including trend charts, data tables and constraint explanations are output, and are synchronized to the strategy generation module as the input of the optimization engine.

[0163] The heterogeneous execution module includes a hierarchical control network unit and an instruction execution unit;

[0164] The hierarchical control network unit is based on a device-subzone mapping table to build a three-level architecture of a central controller-subzone gateway-device controller; the central controller is configured to receive an energy consumption intelligent regulation and control scheme and distribute it to a corresponding subzone gateway; the subzone gateway is configured to perform instruction analysis and conflict detection; and the device controller is bound to a device;

[0165] The instruction execution unit takes the energy consumption intelligent regulation and control scheme as input, decomposes the scheme into device operation instructions according to the hierarchical control network topology, and sends the device operation instructions to the device controller after protocol conversion, and synchronously records the device operation instruction sending time, receiving state and execution feedback; the device operation instruction includes an operation code, a parameter value and an execution time limit.

[0166] Further, the communication mechanism of the hierarchical control network comprises: the central controller and the partition gateway use Ethernet communication, and the transmission protocol is MQTT; the partition gateway and the device controller use hybrid communication: the power device uses 485 bus, and the environmental device uses LoRa; all communication frames contain a frame header, a length field, a data segment, a check code and a frame tail; a heartbeat detection mechanism is established: the central controller sends a heartbeat frame to the partition gateway every M seconds, and if there is no response after timeout, it is marked as offline and a standby gateway is enabled.

[0167] The instruction disassembly process of the instruction execution unit comprises:

[0168] An energy consumption intelligent control scheme, a device-partition mapping table and a hierarchical control network topology are obtained.

[0169] The device operation sequence in the energy consumption intelligent control scheme is extracted, the target device ID of each instruction in the operation sequence is extracted, then the device-partition mapping table is queried to obtain the partition ID to which the device belongs, and the device operation sequence is grouped according to the partition ID to generate a partition-grouped operation sequence; the partition-grouped operation sequence is in a partition-instruction list key-value pair format.

[0170] The partition-grouped operation sequence is traversed, the instruction ID is read, and the predecessor and successor instructions of the corresponding instruction are queried based on the instruction dependency tree, then a constraint mark is added to each predecessor and successor instruction to obtain a partition operation sequence with a constraint mark, which specifically comprises:

[0171] (1) information is extracted from the partition-grouped operation sequence, including the partition ID, the instruction ID, the device ID, the operation parameter and the planned execution time window;

[0172] (2) an index table is established according to the partition ID, all instruction IDs of the same partition are sorted according to the start time of the time window to form a time sequence operation queue within the partition;

[0173] (3) a pre-generated instruction dependency tree is loaded; the instruction dependency tree takes the instruction ID as a node, the parent node as a predecessor instruction and the child node as a successor instruction;

[0174] (4) a bidirectional index table is constructed: in the forward index, the key is the instruction ID, and the value is a list of all successor instruction IDs; in the reverse index, the key is the instruction ID, and the value is a list of all predecessor instruction IDs;

[0175] (5) each instruction in the partition operation sequence is traversed, the reverse index table is queried with the instruction ID as the key to obtain all predecessor instruction IDs, i.e. the instructions that must be completed before the current instruction is executed, and the forward index table is queried to obtain all successor instruction IDs, i.e. the instructions that must be started after the current instruction is executed; if the query result is empty, such as an initial instruction or a termination instruction, it is marked as having no predecessor or successor.

[0176] (6) According to the dependency relationship type, three types of constraint labels are defined, including: predecessor constraint label, indicating that the current instruction must be executed after the execution of the predecessor instruction ID is completed, and the execution result of the predecessor instruction ID needs to be successful; successor constraint label, indicating that the current instruction needs to send a ready signal to the successor instruction ID after the execution of the current instruction is completed, and the successor instruction ID can be started; timing constraint label, that is, if there is an overlap between the time window of the predecessor or successor instruction and the current instruction, a preset minimum interval constraint is automatically added, and in the present application, the minimum interval is set to 60 seconds;

[0177] (7) A corresponding predecessor or successor constraint label is added to each instruction, and after adding, the label conflict is checked: if any instruction simultaneously exists and , and X and Y have a dependency conflict, such as X needs to be executed after Y, then it is marked as a constraint conflict, and the conflict instruction ID pair is recorded, wherein X and Y represent two predecessor instruction IDs, and P represents a predecessor, which is used to clearly mark the predecessor instruction dependency relationship of the current instruction;

[0178] (8) For the instructions marked as constraint conflicts, the execution priority of the predecessor instruction is determined by calling the topological sorting result of the dependency relationship tree;

[0179] (9) The instructions after adding the constraint labels are re-integrated according to the original partition and time window order to form a partition operation sequence with constraints, and each instruction record includes: partition ID, instruction ID, device ID, operation parameter, planned time window, predecessor constraint label list, successor constraint label list, and conflict state;

[0180] (10) Check whether all instructions have added constraint labels, count the number of conflict instructions to be manually processed, and if the conflict quantity accounts for less than or equal to a preset account threshold, it is determined that the sequence is valid; otherwise, the constraint label rule is re-adjusted, and in the present application, the preset account threshold is set to 5%;

[0181] (11) Generate a verification log to record the total number of instructions of each partition, the constraint label addition rate, and the conflict resolution rate.

[0182] The communication protocol type corresponding to each device is extracted from the hierarchical control network topology, and then the protocol mapping table is loaded to convert the abstract operation parameter into a device protocol instruction, and a device controller physical address is added to the converted device protocol instruction to generate a partition instruction set with a protocol format;

[0183] A check code and a timeout retransmission parameter are added to each device protocol instruction in the partition instruction set with a protocol format to obtain a device operation instruction set divided by partition.

[0184] The embodiments of the present application are described above with reference to the accompanying drawings, but the present application is not limited to the above-described specific embodiments, and the above-described specific embodiments are merely illustrative, but not restrictive, and a person of ordinary skill in the art can make changes, modifications, replacements and variations to the above-described embodiments without departing from the purpose of the present application and the protected scope, and these are all within the protection of the present application.

Claims

1. An Internet of Things (IoT) system for dynamic monitoring and intelligent control of building energy consumption, characterized in that, include: The IoT sensing module, storage and analysis module, strategy generation module, heterogeneous execution module, and update module; Based on the list of equipment required for building energy consumption monitoring, an initial set of energy consumption control instructions is generated. At the same time, through the various types of sensor nodes deployed by the IoT sensing module, dynamic building data is collected. The dynamic building data includes basic building energy consumption data, environmental parameters, and personnel status data. Based on building dynamic data, the energy consumption baseline model and comfort constraint model built into the storage and analysis module are used to output the calibrated energy consumption trend and comfort constraint conditions; the energy consumption baseline model is combined with the historical execution data of the initial energy consumption control instruction set to generate a correction coefficient matrix; Based on the energy consumption trend prediction results and comfort constraints, a dual objective function is established through the optimization engine of the strategy generation module, and an intelligent energy consumption control scheme is generated by solving the Pareto optimal solution. Based on the generated intelligent energy consumption control scheme, the hierarchical control network constructed by the heterogeneous execution module decomposes the intelligent energy consumption control scheme into equipment operation commands and controls them, while recording the command execution deviation value and feeding back the execution result. Based on the feedback execution results and instruction execution deviation values, the correction coefficient matrix is ​​updated and the energy consumption and comfort deviation rate are calculated through the full-link iterative mechanism of the update module. The storage and analysis module includes a data preprocessing unit and a model computation unit; The data preprocessing unit takes the target energy consumption data space as input, preprocesses the building dynamic data, and generates standardized building dynamic data. The model computing unit loads the energy consumption baseline model, inputs the standardized building dynamic data and the historical execution data of the initial energy consumption control instruction set into the energy consumption baseline model, calculates the energy consumption benchmark value through the sliding window algorithm, and loads the comfort constraint model at the same time, maps the environmental parameters and personnel status data in the standardized building dynamic data into quantitative comfort indicators, and generates comfort constraint conditions. A correction coefficient matrix is ​​generated based on the deviation rate between the energy consumption benchmark value and the actual energy consumption benchmark value; the correction coefficient matrix includes a time decay coefficient, an equipment type coefficient, and a partition weight coefficient. The model operation unit combines the correction coefficient matrix to calibrate the energy consumption trend and outputs the calibrated energy consumption trend and comfort constraints. The process of generating the correction coefficient matrix includes aligning the energy consumption benchmark value with the actual energy consumption value according to the device ID and time granularity, calculating the benchmark value deviation rate, assigning time weight to the historical deviation rate and mapping it to the time decay coefficient, calculating the average benchmark value deviation rate according to the device type to obtain the device type coefficient, calculating the partition weight coefficient according to the energy consumption importance and deviation characteristics of the functional partition, constructing a three-dimensional correction coefficient matrix with the device ID as the row index, time, device type, and partition as the column index, and generating the final correction coefficient matrix after range truncation, anomaly verification, and effectiveness verification of the coefficients. The strategy generation module includes a scheme evaluation unit; the scheme evaluation unit uses the NSGA-II algorithm to solve for the Pareto optimal solution, generates N sets of candidate control schemes, selects the optimal control scheme through the entropy weight coefficient method, and outputs the energy consumption intelligent control scheme; the energy consumption intelligent control scheme includes equipment operation sequence, execution priority, expected energy consumption and comfort value. The update module includes a deviation analysis unit and a model iteration unit. The deviation analysis unit takes the execution result feedback package from the heterogeneous execution module as input and calculates the energy consumption deviation rate and comfort deviation rate. The energy consumption deviation rate is the percentage of the difference between actual and expected energy consumption to the expected energy consumption, and the comfort deviation rate is the percentage of the difference between actual and expected comfort to the expected comfort. The model iteration unit updates the correction coefficient matrix according to the deviation rate level based on the full-link iteration mechanism. When the deviation rate is greater than 15%, the type coefficient of the corresponding device is adjusted by ±0.

1. When the deviation rate is greater than 10% for three consecutive times, the energy consumption baseline model is retrained, and the updated correction coefficient matrix and deviation analysis report are output. The full-link iteration mechanism includes a full cycle iteration performed every 24 hours and an incremental iteration triggered when the deviation of a single device is greater than 10% for three consecutive instruction executions. During the iteration process, the deviation diagnosis rules constructed by the decision tree model are used to locate the deviation source and adjust the corresponding fields of the correction coefficient matrix.

2. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 1, characterized in that, The process of generating the initial energy consumption control instruction set includes: Based on the equipment types and corresponding functional partition attributes recorded in the list of equipment required for building energy consumption monitoring, an equipment-partition mapping table containing the correspondence between equipment and functional partitions is constructed using a hash mapping algorithm. The equipment-partition mapping table includes equipment ID, partition ID, equipment type label, and cross-partition association weight. Based on the rated power, historical operating baseline value and preset energy consumption threshold of each device in the device-partition mapping table, generate initial operating parameter instructions for each device. Based on the initial operating parameter instructions and the cross-partition association weights in the device-partition mapping table, a directed graph topology sort is used to construct an instruction dependency tree, generating an initial energy consumption control instruction set containing device ID, execution time window, and operating parameters.

3. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 2, characterized in that, The initial operating parameters are the ratio of historical operating baseline values ​​to the rated power of the equipment, multiplied by the rated power of the equipment and the threshold margin coefficient. The threshold margin coefficient is the difference between 1 and the energy consumption ratio; the energy consumption ratio is the ratio of the current partition's cumulative energy consumption to the preset energy consumption threshold.

4. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 3, characterized in that, The IoT sensing module includes a time slot scheduling unit and an edge preprocessing unit; Based on the list of building energy consumption monitoring equipment requirements, the monitoring nodes corresponding to the list of building energy consumption monitoring equipment requirements are activated through the time slot scheduling unit. Based on the RSSI value and collision detection result in the response frame of the monitoring node, an adaptive time slot allocation strategy is adopted to generate a dynamic time slot allocation table. The dynamic time slot allocation table includes node ID, time slot number, transmission period and channel frequency. The edge preprocessing unit preprocesses the collected building dynamic data using an outlier detection algorithm based on the dynamic time slot allocation table to generate a target energy consumption data space. The target energy consumption data space includes the real-time power, operating status code, ambient temperature and humidity, personnel density, and data confidence level of each device. The target energy consumption data space is encapsulated in JSON format and an additional verification code is attached.

5. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 4, characterized in that, The generation process of the dynamic time slot allocation table includes: Based on the list of equipment required for building energy consumption monitoring, the time slot scheduling unit sends activation signals to each monitoring node, receives response status frames returned by the nodes, and generates an initial response status table for the nodes. Based on the node initial response state table, a collision prediction algorithm based on Markov chains is used to analyze the collision probability distribution of node data transmission. Combined with the priority weight of node energy consumption data, the time slot length and allocation ratio are dynamically adjusted. Based on the adjusted time slot length and allocation ratio, binary expansion and reallocation are performed on the collision time slots. A conflict-free dynamic time slot allocation table is generated through the mapping relationship between node ID hash value and time slot number. The system monitors the time slot utilization rate of the dynamic time slot allocation table in real time. When the monitored time slot utilization rate exceeds the preset maximum utilization rate or falls below the preset minimum utilization rate, the system triggers the time slot adaptive adjustment mechanism and updates the dynamic time slot allocation table.

6. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 5, characterized in that, The process of acquiring the target energy consumption data space includes: The edge preprocessing unit receives the original building dynamic data transmitted by each monitoring node according to the time slot based on the dynamic time slot allocation table, and detects outliers through the isolated forest algorithm to obtain abnormal building dynamic data and normal building dynamic data. Abnormal building dynamic data is repaired using cubic spline interpolation, while normal building dynamic data is compressed and denoised using wavelet transform. The two are then integrated to form the processed building dynamic data. Based on the device-partition mapping table, the processed building dynamic data of the same type of device in the same area are spatiotemporally aligned and data confidence markers are added to generate a target energy consumption data space. The data confidence level is determined based on a comprehensive evaluation of the data transmission signal-to-noise ratio and the sampling frequency.

7. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 6, characterized in that, The heterogeneous execution module includes a hierarchical control network unit and an instruction execution unit; The hierarchical control network unit is based on the device-partition mapping table and constructs a three-level architecture of central controller-partition gateway-device controller; The central controller is used to receive intelligent energy consumption control schemes and distribute them to the corresponding zone gateways; The partition gateway is used for instruction parsing and conflict detection; the device controller is bound to the device; The instruction execution unit takes the intelligent energy consumption control scheme as input, decomposes the scheme into equipment operation instructions according to the hierarchical control network topology, and sends them to the equipment controller after protocol conversion. It also records the sending time, receiving status and execution feedback of the equipment operation instructions. The equipment operation instructions include operation codes, parameter values ​​and execution time limits.

8. The IoT-based building energy consumption dynamic monitoring and intelligent control system as described in claim 7, characterized in that, The instruction decomposition process of the instruction execution unit includes: Acquire intelligent energy consumption control solutions, device-zone mapping tables, and hierarchical control network topologies; Extract the device operation sequence from the intelligent energy consumption control scheme, then extract the target device ID of each instruction in the operation sequence, then query the partition ID to which the device belongs based on the device-partition mapping table, and group the device operation sequence according to the partition ID to generate the partition-grouped operation sequence; the partition-grouped operation sequence is in partition-instruction list key-value pair format; Traverse the operation sequence after partitioning and grouping, read the instruction ID, and query the predecessor and successor instructions of the corresponding instruction based on the instruction dependency tree. Then add constraint tags to each predecessor and successor instruction to obtain the partition operation sequence with constraint tags. The communication protocol type corresponding to each device is extracted from the hierarchical control network topology. Then, the protocol mapping table is loaded, the abstract operation parameters are converted into device protocol instructions, and the device controller physical address is added to the converted device protocol instructions to generate a partition instruction set with protocol format. Add checksums and timeout retransmission parameters to each device protocol instruction in the partitioned instruction set with protocol format to obtain the device operation instruction set divided by partition.

Images