Air conditioner refrigeration equipment intelligent energy-saving optimization method and system based on multi-sensor monitoring

Abstract

The present application relates to air conditioning refrigeration equipment energy saving control technical field, disclose a kind of based on multiple sensing monitoring's air conditioning refrigeration equipment intelligent energy-saving optimization method and system.Simultaneously obtain indoor air, target wall boundary and outdoor environment temperature, sensible heat refrigeration capacity, fan speed, deflector angle and local air velocity;Build double-node thermal resistance capacity model, the convective heat transfer resistance that changes with air flow organization is as known time-varying parameter, only when the normalized regression data meets rank excitation and the identification result passes through convergence and physical boundary check, update wall parameter.Detect the risk of temperature rebound caused by wall hysteresis heat release, generate compressor frequency and fan speed joint control sequence, use unit performance mapping and gradually updated convective heat transfer resistance to deduce temperature trajectory, select the minimum comprehensive cost from the sequence that meets temperature, heat release, rebound and equipment safety constraints to execute.The present application can reduce the invalid energy consumption caused by temperature rebound and repeated frequency rise and fall.

Description

Technical Field

[0001] This invention relates to the field of energy-saving control technology for air conditioning and refrigeration equipment, and in particular to an intelligent energy-saving optimization method and system for air conditioning and refrigeration equipment based on multi-sensor monitoring. Background Technology

[0002] Variable frequency air conditioning systems typically adjust the compressor frequency based on the deviation between the indoor air temperature and the target set temperature, and adjust the indoor fan speed according to the load. When the indoor air temperature approaches the target set temperature, the conventional controller will reduce the compressor frequency or enter intermittent operation to avoid continuous overcooling and reduce unit power consumption.

[0003] The building envelope of a room has heat storage and delayed heat release characteristics. During the cooling process, the indoor air temperature can drop relatively quickly, while the walls, affected by solar radiation, high outdoor temperatures, and previous heat accumulation, release their stored heat relatively slowly. When the controller reduces cooling output only based on the indoor air temperature, the warmer walls will still release heat into the indoor air, causing the room temperature to rise again in a short time. To suppress this rise, the unit needs to increase the compressor frequency, which can easily lead to frequent frequency increases and decreases or start-ups and shutdowns, reducing temperature stability and increasing ineffective energy consumption.

[0004] Existing building thermal models can describe the heat transfer process between air and the building envelope using thermal resistance-capacitance networks. However, variations in wall materials, thicknesses, moisture content, and outdoor boundary conditions make it difficult for fixed-parameter models to consistently reflect the thermal inertia of actual rooms. If thermal parameters are directly identified using operational data, the regression data lacks effective excitation when temperature and cooling changes are insufficient, leading to parameter drift. Transient disturbances such as door and window opening, occupant activity, and indoor equipment heating may also be misattributed to wall parameters, further reducing the physical reliability of the identification results.

[0005] The indoor fan speed and the angle of the air deflector alter the coverage of the supply air jet on the wall and the near-wall air velocity, thus changing the intensity of convective heat transfer between the wall and the indoor air. Simplifying this convective heat transfer to a fixed thermal resistance distorts the estimation of heat release power corresponding to the same wall temperature difference under different airflow configurations. Furthermore, if predictive control directly equates compressor frequency to cooling capacity without considering the combined effects of fan speed, return air conditions, and outdoor operating conditions on sensible cooling capacity and unit power consumption, it can easily lead to discrepancies between the predicted temperature trajectory and the actual operating process.

[0006] Therefore, there is a need for an energy-saving optimization scheme that can identify the thermal inertia of the wall using multi-sensor monitoring results, dynamically update the convective thermal resistance of the wall according to the airflow organization, and predict the rebound caused by the delayed heat release of the wall when the room temperature is close to the set value. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent energy-saving optimization method and system for air conditioning and refrigeration equipment based on multi-sensor monitoring. The convective thermal resistance that changes with airflow organization is treated as a known time-varying parameter and separated from the inherent thermal parameters of the wall. Based on the prediction of the delayed heat release of the target wall, the compressor and indoor fan are jointly optimized to improve the reliability of wall parameter identification and reduce room temperature rebound and ineffective energy consumption.

[0008] To achieve the above objectives, the present invention provides an intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring, comprising:

[0009] Simultaneously acquire the indoor air temperature sequence, the target wall boundary temperature sequence within the air jet coverage area, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, as well as the indoor fan actual speed sequence, the air guide plate actual deflection angle sequence, and the local air velocity sequence at the location adjacent to the target wall.

[0010] A two-node thermal resistance-capacitance model is constructed, which includes the equivalent heat capacity of the wall, the outdoor heat transfer thermal resistance, and the variable parameter convective thermal resistance that varies with airflow organization.

[0011] Based on the actual speed sequence of the indoor fan, the actual deflection angle sequence of the air guide plate, and the local air velocity sequence, the actual variable parameter convective thermal resistance is determined at each sampling time, and the actual variable parameter convective thermal resistance is used as a known time-varying parameter to participate in the wall parameter identification.

[0012] When the regression data matrix consisting of observations and known quantities from multiple consecutive sampling periods satisfies the rank excitation condition, the equivalent heat capacity of the wall and the outdoor heat transfer resistance are identified and updated online based on the indoor air temperature sequence, the target wall boundary temperature sequence, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, and the actual variable parameter convective thermal resistance; when the rank excitation condition is not met, the parameter update for this round is stopped and the most recently valid parameter is used.

[0013] When the rebound trigger condition based on the hysteresis heat release estimation of the target wall is met, multiple sets of joint control sequences containing candidate compressor frequencies and candidate indoor fan speeds are generated.

[0014] Based on the unit performance mapping model, the joint control sequence of each group is converted step by step into a predicted sensible heat cooling capacity sequence, a compressor estimated power consumption sequence, and a fan estimated power consumption sequence.

[0015] For each group of joint control sequences, the dual-node thermal resistance-capacity model is used as the state prediction equation, the corresponding predicted sensible heat cooling capacity sequence is used as the cooling capacity input, and the variable parameter convective thermal resistance is updated step by step using the corresponding indoor fan candidate speed to deduce and predict the temperature trajectory.

[0016] From the joint control sequences that satisfy temperature, target wall heat release, room temperature rebound, and equipment safety constraints, the optimal joint control sequence with the minimum comprehensive cost function is selected and executed.

[0017] Optionally, the two-node thermal resistance-capacitance model includes:

[0018] Establish the differential equation for the thermal balance of the air node, where the product of the equivalent heat capacity of the air and the rate of change of the indoor air temperature is equal to the sum of the dynamic convective heat transfer from the target wall to the indoor air and the unmodeled transient thermal disturbance in the room, minus the sensible heat cooling capacity.

[0019] Establish a differential equation for the thermal balance of the wall nodes, wherein the product of the equivalent heat capacity of the wall and the rate of change of the boundary temperature of the target wall is equal to the outdoor heat transfer from the outdoor environment to the target wall minus the dynamic convection heat transfer.

[0020] The dynamic convective heat transfer is obtained by dividing the difference between the target wall boundary temperature and the indoor air temperature by the variable parameter convective thermal resistance; the outdoor heat transfer is obtained by dividing the difference between the outdoor ambient temperature and the target wall boundary temperature by the outdoor heat transfer thermal resistance.

[0021] Optionally, the synchronous acquisition of the sensible heat cooling capacity sequence includes:

[0022] Real-time acquisition of return air dry-bulb temperature and supply air dry-bulb temperature of air conditioning and refrigeration equipment;

[0023] The air volume flow rate is determined based on the actual speed of the indoor fan and the pre-tested duct resistance curve, and then converted into real-time air mass flow rate by combining the air density corresponding to the return air state.

[0024] The sensible heat cooling capacity at the corresponding sampling time is obtained by multiplying the real-time air mass flow rate, the air constant pressure specific heat capacity, and the absolute value of the difference between the return air dry-bulb temperature and the supply air dry-bulb temperature. When the supply air dry-bulb temperature is not lower than the return air dry-bulb temperature, or when the air conditioning refrigeration equipment is in an unstable cooling state, the corresponding data frame is marked as not participating in wall parameter identification.

[0025] Optionally, the online identification and updating of the wall's equivalent heat capacity and outdoor thermal resistance includes:

[0026] The differential equation of thermal equilibrium of the wall nodes in the two-node thermal resistance-capacitance model is discretized into an autoregressive exogenous input difference equation.

[0027] Observations and known quantities from multiple consecutive sampling periods are extracted to construct a regression data matrix, and the scale of each column of the regression data matrix is ​​normalized.

[0028] When the normalized regression data matrix has full column rank and the minimum singular value is not less than the preset excitation threshold, it is determined that the rank excitation condition is met, and rolling fitting is performed using the recursive least squares method with adaptive forgetting factor and sample robust weight.

[0029] When the parameter estimation covariance matrix satisfies the preset convergence condition, and the wall equivalent heat capacity and outdoor heat transfer resistance obtained by inverse analysis are both positive and within the preset wall attribute boundary, the two are registered as new most recent effective parameters.

[0030] When the rank excitation condition, the preset convergence condition, or the preset wall attribute boundary is not met, the identification results of this round are discarded and the most recent valid parameters are used.

[0031] Optionally, the recursive least squares method further includes a disturbance-resistant weighting step based on the innovative residual of air temperature:

[0032] Using the most recent effective parameters and the actual variable convective thermal resistance at the corresponding sampling time, the one-step forward prediction value of indoor air temperature is calculated;

[0033] Subtract the previous forward prediction value from the measured indoor air temperature at the next sampling time to obtain the innovative residual of air temperature.

[0034] When the absolute value of the innovative residual of air temperature is not greater than the fault tolerance threshold, the recursive weight coefficient of the corresponding data sample is set to 1; when the absolute value is greater than the fault tolerance threshold, the recursive weight coefficient is set to the ratio of the fault tolerance threshold to the absolute value.

[0035] When the absolute value of the innovative residual of the air temperature continuously exceeds the preset stop threshold, the parameter update of the current data window is suspended to isolate transient thermal disturbances introduced by opening doors and windows, personnel activities, or indoor equipment heating.

[0036] Optionally, the step-by-step prediction update of the variable parameter convective thermal resistance includes:

[0037] The ratio of the latest measured local air velocity to the aerodynamic correlation mapping function is calculated under the latest actual indoor fan speed and the latest actual deflection angle of the air guide plate. The resulting ratio is then limited between a preset lower limit and a preset upper limit to obtain the airflow organization correction coefficient.

[0038] Within the prediction window, the predicted mapped air velocity, determined by the candidate speed of the indoor fan and the deflection angle of the corresponding air guide plate, is calculated in each prediction step using the aerodynamic correlation mapping function, and the predicted mapped air velocity is corrected using the airflow organization correction coefficient.

[0039] The boundary layer Reynolds number is calculated based on the corrected local air velocity, the characteristic length of the target wall, and the air kinematic viscosity.

[0040] Substituting the boundary layer Reynolds number into the Nusselt number correlation corresponding to the geometry and flow state of the target wall, the dynamic convective heat transfer coefficient is obtained.

[0041] The reciprocal of the product of the dynamic convective heat transfer coefficient and the effective heat transfer area of ​​the target wall is taken as the variable parameter convective thermal resistance of the corresponding prediction step and substituted into the state prediction equation.

[0042] Optionally, the rebound trigger condition includes:

[0043] The absolute temperature difference between the indoor air temperature and the target set temperature shall not exceed the set temperature difference threshold.

[0044] The target wall boundary temperature is higher than the indoor air temperature, and the transient heat release power of the target wall, obtained by dividing the temperature difference between the two by the current actual variable parameter convective thermal resistance, is greater than the set heat release threshold.

[0045] The baseline predicted temperature trajectory, derived by maintaining the current compressor frequency, indoor fan speed, and air guide deflection angle, satisfies at least one of the following: the highest predicted temperature of the baseline predicted temperature trajectory is higher than the sum of the target set temperature and the preset rebound margin; the room temperature rebound rate of a consecutive preset number of prediction steps is greater than the set rebound rate threshold.

[0046] Optionally, the step of selecting the optimal joint control sequence and issuing it for execution includes:

[0047] Extract the lowest and highest predicted temperatures from the predicted temperature trajectories of each group;

[0048] The joint control sequence that is lower than the preset room temperature lower limit or higher than the preset room temperature upper limit is excluded;

[0049] The joint control sequence that excludes any prediction step within the prediction window where the transient heat release power of the target wall exceeds the set upper limit of heat release power or the room temperature rebound rate exceeds the set upper limit of rebound rate is excluded.

[0050] Eliminate joint control sequences that violate the minimum operating frequency limit of the compressor or the lower limit of the safe evaporation pressure determined by the pre-calibrated safe operating envelope of the unit;

[0051] Substitute the remaining joint control sequence into the comprehensive cost function, which is a weighted sum of the temperature tracking error penalty term, the compressor energy consumption penalty term, the fan energy consumption penalty term, the target wall transient heat release power penalty term, and the room temperature rebound rate penalty term.

[0052] The joint control sequence with the smallest output value of the comprehensive cost function is selected, and only the first control step of the joint control sequence is issued. At the next sampling time, prediction and optimization are re-executed based on the newly acquired sensor data.

[0053] Optionally, the establishment and invocation of the unit performance mapping model includes:

[0054] Acquire valid historical operating data under different combinations of compressor operating frequency, indoor fan operating speed and environmental conditions, wherein the environmental conditions include at least return air dry bulb temperature, return air relative humidity and outdoor ambient temperature;

[0055] The sensible heat cooling capacity is calculated based on the air mass flow rate and the dry-bulb temperature difference of the supply and return air corresponding to each group of valid historical operating data, and the corresponding compressor power consumption and fan power consumption are obtained.

[0056] Establish a unit performance mapping model with compressor operating frequency, indoor fan operating speed and the aforementioned environmental conditions as inputs, and sensible heat cooling capacity, compressor power consumption and fan power consumption as outputs;

[0057] Within the prediction window, the candidate compressor frequency, candidate indoor fan speed, and predicted environmental conditions corresponding to each prediction step are input into the unit performance mapping model to obtain the corresponding predicted sensible heat cooling capacity, estimated compressor power consumption, and estimated fan power consumption. For input combinations that exceed the coverage of valid historical operating data, no corresponding joint control sequence is generated.

[0058] This invention also provides an intelligent energy-saving optimization system for air conditioning and refrigeration equipment based on multi-sensor monitoring, comprising:

[0059] The acquisition module is configured to simultaneously acquire the indoor air temperature sequence, the target wall boundary temperature sequence within the coverage area of ​​the air jet, the outdoor ambient temperature sequence and the sensible heat cooling capacity sequence, as well as the indoor fan actual speed sequence, the air guide plate actual deflection angle sequence and the local air velocity sequence at the location adjacent to the target wall.

[0060] The model building module is configured to build a two-node thermal resistance-capacity model that includes the equivalent heat capacity of the wall, the outdoor heat transfer thermal resistance, and the variable parameter convective thermal resistance that varies with airflow organization. The actual variable parameter convective thermal resistance is determined at each sampling time according to the actual speed sequence of the indoor fan, the actual deflection angle sequence of the air guide plate, and the local air velocity sequence. The actual variable parameter convective thermal resistance is provided to the identification module as a known time-varying parameter.

[0061] The identification module is configured to, when the regression data matrix satisfies the rank excitation condition, identify and update the equivalent heat capacity of the wall and the outdoor heat transfer resistance online based on the indoor air temperature sequence, the target wall boundary temperature sequence, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, and the actual variable parameter convective thermal resistance; when the rank excitation condition is not satisfied, the most recent effective parameter is used.

[0062] The sequence generation module is configured to generate multiple sets of joint control sequences, including candidate compressor frequencies and candidate indoor fan speeds, when the rebound trigger condition based on the target wall's delayed heat release estimation is met.

[0063] The mapping and conversion module is configured to convert the joint control sequence of each group into a predicted sensible heat cooling capacity sequence, a compressor estimated power consumption sequence, and a fan estimated power consumption sequence step by step based on the unit performance mapping model.

[0064] The deduction module is configured to use the two-node thermal resistance-capacitance model as the state prediction equation, the corresponding predicted sensible heat cooling capacity sequence as the cooling capacity input, and update the variable parameter convective thermal resistance step by step using the corresponding indoor fan candidate speed to deduce and predict the temperature trajectory for each group of joint control sequences.

[0065] The optimization execution module is configured to select the optimal joint control sequence with the minimum comprehensive cost function from the joint control sequences that satisfy temperature, target wall heat release, room temperature rebound, and equipment safety constraints, and then issue it for execution.

[0066] Compared with the prior art, the present invention has at least the following beneficial effects:

[0067] Simultaneously, indoor air temperature and target wall boundary temperature are observed, and a two-node thermal resistance-capacitance model is used to characterize the dynamic relationship between outdoor heat transfer, wall heat storage, wall heat release and sensible heat cooling capacity, enabling the controller to identify the risk of delayed heat release of the target wall before the room temperature has risen significantly.

[0068] The indoor fan speed, the deflection angle of the air guide plate, and the local air velocity near the target wall are used together to determine the variable parameter convective thermal resistance. This thermal resistance is updated step by step during the candidate control sequence derivation process, so that the influence of fan adjustment on the wall heat release rate and room temperature evolution is incorporated into the predictive control process.

[0069] Parameter updates are only accepted when the scale-normalized regression data satisfies the rank excitation condition, the parameter estimation covariance matrix converges, and the physical parameters are within the boundary of the wall properties. The recursive weight of transient thermal disturbance samples is reduced by using the innovative residual of air temperature, thereby suppressing parameter drift caused by weak excitation and non-wall thermal disturbance.

[0070] By using the unit performance mapping model to convert the candidate compressor frequency and the candidate indoor fan speed into sensible heat cooling capacity and corresponding power consumption, and then taking into account temperature, heat release, rebound, equipment energy consumption and cooling safety constraints to select a joint control sequence, it is beneficial to reduce the increase in energy consumption caused by repeated frequency increases and decreases based solely on instantaneous room temperature. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of the multi-sensor monitoring arrangement and the control connection relationship of the air conditioning refrigeration equipment in an embodiment of the present invention;

[0072] Figure 2 This is a flowchart illustrating the overall process of the intelligent energy-saving optimization method for air conditioning and refrigeration equipment based on multi-sensor monitoring in this embodiment of the invention.

[0073] Figure 3 This is a schematic diagram of the dual-node thermal resistance-capacitance model and online parameter identification process in an embodiment of the present invention;

[0074] Figure 4 This is a flowchart illustrating the airflow organization correction and variable parameter convective thermal resistance determination in an embodiment of the present invention.

[0075] Figure 5 This is a flowchart of the rebound triggering, candidate sequence deduction and constraint optimization in an embodiment of the present invention;

[0076] Figure 6 This is a schematic diagram of the module structure of the intelligent energy-saving optimization system for air conditioning and refrigeration equipment based on multi-sensor monitoring in an embodiment of the present invention. Detailed Implementation

[0077] The embodiments of the present invention will be further described below with reference to the accompanying drawings. These embodiments are used to illustrate the technical principles and implementation process of the present invention and are not intended to limit the scope of protection of the present invention.

[0078] like Figure 1 As shown, the air conditioning refrigeration equipment in this embodiment includes a compressor, an indoor heat exchanger, an indoor fan, an air guide plate, an outdoor heat exchanger, and a controller. The controller is connected to an indoor air temperature sensor, a target wall boundary temperature sensor, an outdoor ambient temperature sensor, a supply air temperature sensor, a return air temperature sensor, a return air relative humidity sensor, a local air velocity sensor, an indoor fan speed detection terminal, an air guide plate angle feedback terminal, and an electrical parameter detection terminal, and can output control commands to the compressor drive terminal, the indoor fan drive terminal, and the air guide plate actuator terminal.

[0079] Indoor air temperature sensors are positioned to avoid direct airflow from air vents, localized radiation from exterior walls, and other significant heat sources. The target wall is the wall or wall region to be monitored within the coverage area of ​​the air jet. The target wall boundary temperature refers to the temperature of its indoor surface boundary. When multiple candidate wall regions exist, the effective heat transfer area of ​​each candidate region is calculated within the most recent stable cooling window, multiplied by its positive temperature difference relative to the indoor air temperature. The region with the largest time average of this product is prioritized; alternatively, multiple regions with a cumulative contribution reaching a preset proportion can be selected, and multiple wall nodes can be established. The target wall boundary temperature can be obtained by an attached temperature sensor array or by an infrared temperature measurement unit with emissivity correction. When setting multiple temperature measurement points, first, direct sunlight patches, points near localized heat sources, and abnormal points exceeding the sensor's reliable range are eliminated. Then, the area weight is determined based on the wall area represented by each temperature measurement point, and the weighted temperature is used as the target wall boundary temperature. The effective heat transfer area of ​​the target wall is the sum of the areas of all wall regions participating in the weighting.

[0080] Local air velocity sensors are positioned at representative locations on the outer edge of the target wall boundary layer, with their measurement direction aligned with the main airflow direction along the wall. If multiple local air velocity sensors are used, the velocity is weighted according to the area of ​​the wall region corresponding to each sensor. The actual deflection angle of the air guide vane is obtained from the vane position feedback value; for devices without an independent angle encoder, it can be calculated based on the cumulative steps of the stepper motor and the zero-position calibration result. The actual rotational speed of the indoor fan is obtained from the Hall feedback signal, the back electromotive force estimation result, or the fan driver feedback value.

[0081] Each sensor value is written into a data frame under a unified clock. The data frame includes at least the sampling time, sensor identifier, measured value, and valid status. The controller performs time alignment of the data frames according to a preset sampling period. When the sampling frequency of a sensor value is higher than the control period, the mean or median value within that control period is used; when the sampling frequency is lower than the control period, the most recent valid value is used within the allowed holding time. Data exceeding the allowed holding time, exceeding the sensor's physical limit, or whose related measurements do not meet the device's operating state constraints are marked as invalid and are not included in the current round of parameter identification and joint optimization. The sampling period can be set according to the room's thermal inertia and the device's communication cycle, for example, from thirty seconds to two minutes, but its specific value does not constitute a limitation of the invention.

[0082] like Figure 2 As shown, the method in this embodiment includes steps S101 to S108.

[0083] Step S101: Simultaneously acquire the indoor air temperature sequence, the target wall boundary temperature sequence, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, the indoor fan actual speed sequence, the air guide plate actual deflection angle sequence, and the local air velocity sequence.

[0084] The current sensible heat cooling capacity is not directly derived from the compressor's nameplate cooling capacity, but is determined based on real-time measurements on the supply and return air sides. Let the current sensible heat cooling capacity be... Real-time air mass flow rate is The specific heat capacity of air at constant pressure is The return air dry bulb temperature is The temperature of the air supply dry bulb is ,but:

[0085] ;

[0086] in, Indicates the sampling time; The unit is watt; The unit is kilograms per second; The unit is joules per kilogram in Kelvin; and The unit is Celsius, and the temperature difference between the two is the same as the temperature difference expressed in Kelvin.

[0087] Real-time air mass flow rate is determined by the following formula:

[0088] ;

[0089] in, This indicates the air density determined based on the return air dry-bulb temperature and return air relative humidity. Indicates air volume flow rate; Indicates the actual rotational speed of the indoor fan; This represents the speed-volume flow rate mapping function determined jointly by the fan performance curve and the pre-tested duct resistance curve. For indoor units that can directly feedback airflow, a calibrated feedback airflow can be used instead. The calculation results.

[0090] Air density is measured in kilograms per cubic meter (kg / m³), air volumetric flow rate is measured in cubic meters per second (m / s), and the actual rotational speed of the indoor fan is measured in revolutions per minute (rpm) or the equivalent rotational speed reported by the equipment. The output of the rotational speed-volumetric flow rate mapping function is uniformly converted to m / s. Return air relative humidity is obtained from a return air relative humidity sensor. Air density can be calculated based on the return air dry-bulb temperature, return air relative humidity, and the measured or calibrated atmospheric pressure at the installation location, or it can be obtained from a table of calibrated humid air properties. Air specific heat capacity at constant pressure is determined from a table of humid air properties under the same return air conditions, or, within the allowable humidity range, a constant value calibrated by the model can be used.

[0091] The above calculations only include the sensible heat portion corresponding to the dry-bulb temperature drop of the air. The latent heat of condensation generated by refrigeration and dehumidification is not included. This ensures that the cooling capacity input to the two-node thermal resistance-capacitance model remains consistent with the sensible heat energy corresponding to changes in indoor air temperature. If the supply air temperature is not lower than the return air temperature, or if the equipment is in an unstable cooling state such as defrosting, heating, or shutdown to blow off residual cold, then the use of this data frame for updating model parameters will be paused, but it can be retained as a record of the equipment status.

[0092] Step S102: Construct a two-node thermal resistance-capacitance model. Figure 3 In this diagram, air nodes represent the equivalent thermal state of indoor air and lightweight objects that rapidly exchange heat with it, while wall nodes represent the equivalent thermal state of the target wall. The outdoor environment is connected to the wall nodes via outdoor heat transfer resistance, and the wall nodes are connected to the air nodes via variable-parameter convective thermal resistance. Sensible heat cooling capacity serves as the input cooling capacity flowing out of the air nodes.

[0093] Let the equivalent heat capacity of air be The equivalent heat capacity of the wall is The indoor air temperature is The target wall boundary temperature is The outdoor ambient temperature is Outdoor thermal resistance is The variable parameter of convective thermal resistance, which varies with airflow organization, is: The unmodeled transient thermal disturbance momentum in the indoor environment is The continuous-time thermal equilibrium relationships for air nodes and wall nodes are as follows:

[0094] ;

[0095] ;

[0096] in, and The unit is joules per Kelvin; and The unit is Kelvin per watt; and The unit is watts. (Air equivalent heat capacity) The effective air volume, air density, and specific heat capacity of air at constant pressure in the room can be determined and corrected through a short-time stable cooling test; the equivalent heat capacity of the wall. Thermal resistance to outdoor heat transfer For parameters to be identified online. Dynamic convection heat transfer. Heat transfer between the outside and the outside They are respectively:

[0097] ;

[0098] In the formula, t represents continuous time in seconds; all temperature quantities can be expressed in degrees Celsius or Kelvin, and the same temperature scale should be maintained when participating in temperature difference calculations; the units for dynamic convection heat transfer and outdoor heat transfer are both watts.

[0099] when A positive value indicates that the target wall releases heat to the indoor air; a negative value indicates that the indoor air transfers heat to the target wall. The model uses a consistent heat flow direction convention to avoid changing the sign definition during the cooling and rebound phases.

[0100] The equivalent heat capacity of air is initially calculated based on the effective air volume of the room, the air density under return air conditions, and the specific heat capacity of air at constant pressure. This initial value is then corrected using short-term stable cooling data. Samples are selected where the equipment is in stable cooling mode, doors and windows are closed, the absolute value of the air temperature change rate is higher than the minimum identifiable threshold, and the air temperature innovation residual does not exceed the fault tolerance threshold. In these samples, the unmodeled transient thermal disturbance is approximated as zero, and the equivalent heat capacity of air for each sample is calculated using the following formula:

[0101] ;

[0102] Where i represents the valid calibration sample number, and the unit of the air equivalent heat capacity sample value is joules per Kelvin; the dynamic convective heat transfer at the sample time is calculated from the target wall boundary temperature, indoor air temperature, and actual variable parameter convective thermal resistance. Indicates the first The sampling time corresponding to each valid correction sample This indicates the sampling period between adjacent sampling times. Indicates by the first The equivalent heat capacity of air calculated from a valid calibration sample. Indicates the first The dynamic convective heat transfer at each effective calibration sample time. Indicates the first Sensible heat cooling at each effective correction sample time. and They represent the first The and the first The indoor air temperature corresponding to each valid calibration sample This indicates that the median of the valid samples within the set is taken. After removing samples with excessively small absolute values ​​in the denominator, non-positive results, or those exceeding the air heat capacity boundary set based on room volume, the median of the remaining results is used as the correction value.

[0103] ;

[0104] When there are insufficient effective calibration samples, the equivalent heat capacity of air obtained based on the effective air volume of the room and corrected by existing model tests is used, and the parameter is not updated using a single abnormal sample.

[0105] Step S103: Determine the actual variable parameter convective thermal resistance at each historical sampling time based on the actual indoor fan speed sequence, the actual deflection angle sequence of the air guide vane, and the local air velocity sequence. For example... Figure 4 As shown, the pneumatic correlation mapping function is first established during the equipment installation or commissioning phase. This function uses the indoor fan speed. and the deflection angle of the air guide plate The input is the mapped air velocity at the location represented by the target wall, and the output is the local air velocity. A two-dimensional lookup table can be built by collecting local air velocities under permissible fan speed settings and guide vane angle combinations. During runtime, bilinear interpolation is used to obtain the mapped air velocity at non-calibrated points; alternatively, a piecewise polynomial with monotonic constraints can be used to fit the mapping.

[0106] At the latest sampling time, the airflow organization correction factor is determined based on the measured values:

[0107] ;

[0108] in, This represents the airflow organization correction factor; This represents the measured value of local airflow velocity; This indicates the actual deflection angle of the air guide plate; This represents the minimum mapped flow rate to prevent the denominator from approaching zero. and These represent the lower and upper limits of the correction factor, respectively. This indicates that the input is restricted to a given upper and lower limit. By setting these limits, sudden changes in predicted thermal resistance caused by transient noise from the local airflow sensor can be avoided.

[0109] The minimum mapped velocity is taken as the larger of the effective measurement lower limit of the local air velocity sensor and the minimum non-zero output value in the aerodynamic correlation mapping table. The lower and upper limits of the correction coefficient are determined based on the lower and upper quantile values ​​of the "ratio of measured local air velocity to mapped air velocity" in the effective samples during the commissioning phase, with sensor error margins allowed outwards. If the ratio exceeds the stated upper and lower limits during operation, the corresponding limit value is adopted, and the aerodynamic correlation mapping function is not updated using this abnormal ratio.

[0110] For historical sampling times, the actual boundary layer Reynolds number is calculated using the measured local air velocity at the corresponding time. For the prediction step... First, according to the candidate indoor fan speed Deflection angle of the corresponding air guide plate Calculate the corrected local air velocity:

[0111] ;

[0112] in, Indicates the current sampling time For prediction step The calculated local airflow velocity. The deflection angle of the air guide plate can be maintained at the current effective angle, or it can be determined according to a preset speed-angle linkage rule. The linkage rule is used to ensure that the air jet continuously covers the target wall, without treating the air guide plate angle as an independent optimization variable.

[0113] As an optional embodiment, the speed-angle linkage rule is specifically manifested as a stepped mapping relationship: when the indoor fan speed... At rpm, set the deflection angle of the air guide vane. ;when At that time, set ;when At that time, set .

[0114] Furthermore, the aerodynamic correlation mapping function A second-order polynomial fitting method based on piecewise rotational speed is used. Taking a fixed guide vane angle as an example, its functional form is: Within the effective air supply range (e.g.) ), according to each Define a fitting interval, and use the least squares method to obtain the polynomial coefficients within each interval. Furthermore, a constraint condition for the continuity of the first derivative is applied at the nodes of adjacent intervals, thereby ensuring the smoothness of the mapped velocity output during the optimization process. , and These represent the constant term coefficient, linear term coefficient, and quadratic term coefficient of the second-order polynomial within the current speed segment, respectively.

[0115] The actual boundary layer Reynolds number, Nusselt number, dynamic convective heat transfer coefficient, and variable parameter convective thermal resistance are determined sequentially according to the following formula:

[0116] ;

[0117] ;

[0118] ;

[0119] in, This represents the characteristic length of the target wall along the direction of the main airflow; Indicates the kinetic viscosity of air; Represents the Prandtl number for air; , and The empirical coefficient of the Nusselt number corresponding to the geometry and flow state of the wall; Indicates the thermal conductivity of air; This represents the minimum permissible convective heat transfer coefficient under natural convection and weak air supply conditions; Indicates the effective heat exchange area of ​​the target wall; Indicates the prediction step The dynamic convective heat transfer coefficient; Indicates the prediction step The variable parameter convective thermal resistance is calculated using the same formula, but with the measured local air velocity at that historical moment used instead. .in, Indicates the current sampling time For prediction step The calculated boundary layer Reynolds number, Indicates the Nusselt number for the corresponding prediction step; vertical bar notation. Indicates the current sampling time For the Each prediction step is used to make a prediction. This indicates taking the maximum value of each item within the parentheses.

[0120] The Reynolds number and Nusselt number of the boundary layer mentioned above are dimensionless; the unit of the characteristic length of the target wall is meters, the unit of air kinematic viscosity is square meters per second, the unit of air thermal conductivity is watts per meter Kelvin, the unit of dynamic convective heat transfer coefficient and its minimum value is watts per square meter Kelvin, the unit of effective heat transfer area of ​​the target wall is square meters, and the unit of the obtained variable parameter convective thermal resistance is Kelvin per watt; the empirical coefficients of the Nusselt number are all dimensionless.

[0121] The Nusselt number correlation is selected based on the target wall geometry, main airflow direction, and Reynolds number applicability range. During installation and commissioning, temporary heat flux measurement points can be set up in representative areas of the target wall. Nusselt number samples are obtained based on the measured surface temperature difference, local flow velocity, and heat flux density. Under the conditions of a fixed Prandtl number exponent and positive coefficients with continuous segmented connections, correlation coefficients are fitted for the low Reynolds number and high Reynolds number segments respectively. After commissioning, the temporary heat flux measurement points are removed. During operation, only local airflow velocity, wall temperature, and air properties are used to calculate the variable-parameter convective thermal resistance. The actual variable-parameter convective thermal resistance obtained in this way is used as a known time-varying parameter in the identification of wall thermal parameters, without being simultaneously estimated without constraints along with the wall's equivalent heat capacity and outdoor heat transfer thermal resistance.

[0122] Within a Reynolds number segment, with the Prandtl number exponent fixed and consistent with the selected correlation, the Nusselt number empirical coefficients are fitted from the effective debug samples according to the following formula:

[0123] ;

[0124] Where i represents the valid test sample number, and the summation range is all valid test samples within the corresponding Reynolds number segment; C and m are the positive parameters to be fitted, and the Prandtl number exponent is kept as a fixed value consistent with the selected correlation. The Nusselt number of each test sample is calculated from the measured heat flux density, target wall temperature difference, air thermal conductivity, and characteristic length. Only samples with positive Nusselt, Reynolds, and Prandtl numbers and an absolute value of the wall temperature difference higher than the measurement resolution are retained to ensure that the definition of logarithmic operation holds. The fitting result is only called within the corresponding Reynolds number range; when crossing segments, the output of adjacent correlations is continuously interpolated in the preset transition zone. Specifically, the Prandtl number exponent consistent with the selected correlation... Determined according to the classical heat transfer criterion equations. When the boundary layer Reynolds number is in the low Reynolds number range (such as under natural convection or very light wind conditions), When using a structure similar to the Churchill-Chu relation, the Prandtl exponent is set. When the boundary layer Reynolds number is in the high Reynolds number range (forced convection state), When using a cooling configuration similar to the Dittus-Boelter correlation, the Prandtl number is set. The clear definition of this index makes the empirical coefficient... and The physical dimensionality reduction fit possesses unique determinism. Among them, and represent the Nusselt number correlation coefficient and Reynolds number exponent obtained from the fitting, respectively; , and They represent the first Nusselt number, Reynolds number, and Prandtl number of a valid debug sample; Indicates in and The parameter values ​​that minimize the sum of squared residuals under the constraints. It represents the natural logarithm.

[0125] Step S104: Identify the equivalent heat capacity of the wall and the outdoor thermal resistance online when the rank excitation condition is met. Perform forward differencing on the wall nodal equations with the sampling period, and define the one-step increment of the wall boundary temperature, the regression vector, and the vector of parameters to be estimated as follows:

[0126] ;

[0127] ;

[0128] Therefore, the autoregressive exogenous input difference relationship of the wall nodes can be expressed as:

[0129] ;

[0130] In this model, the incremental step represents the change in wall boundary temperature between adjacent sampling times, measured in Kelvin. The regression vector is a two-dimensional vector, with its first component being the product of the sampling period and the difference between the outdoor ambient temperature and the wall boundary temperature, and its second component being the negative of the convective energy transferred from the wall node to the air node within the sampling period. The two components of the parameter vector to be estimated are the reciprocal of the product of the wall's equivalent heat capacity and the outdoor thermal resistance, and the reciprocal of the wall's equivalent heat capacity, respectively. The error term is the temperature increment error, which includes measurement noise and discretization error. The sampling period is measured in seconds. Since the two regression components and the two parameter components have different dimensions, rank excitation is only performed after normalization to the column scale. The air node difference equation simultaneously uses sensible heat cooling capacity, actual variable parameter convective thermal resistance, indoor air temperature, and target wall boundary temperature to calculate the forward predicted air temperature value. The sensible heat cooling capacity is used for dual-node state evolution and transient thermal disturbance identification. Indicates the first The wall boundary temperature increments one step at each sampling time. Indicates the first The regression vector at each sampling time point, Represents the vector of parameters to be estimated. Indicates the first Temperature increment error at each sampling time Indicates the sampling period; , , and They represent the first The outdoor ambient temperature, target wall boundary temperature, indoor air temperature, and actual variable parameter convective thermal resistance at each sampling time were measured. Indicates the first The target wall boundary temperature at each sampling time, superscript Represents the transpose of a vector or matrix.

[0131] In length Within the scrolling data window, the regression vectors are arranged sequentially to obtain the regression data matrix:

[0132] ;

[0133] When normalizing the column scale of the regression data matrix, first calculate the root mean square scale of each column within the current scrolling window, and then set a lower bound for the positive scale that is consistent with the dimensions of the corresponding column:

[0134] ;

[0135] Then, the scaling matrix and the normalized regression data matrix are constructed:

[0136] ;

[0137] The lower limit of each column's scale is taken as not less than the minimum effective scale formed by regression construction of the corresponding measurement chain resolution. The diagonal elements of the scale matrix are two columns of scale, and each column of the normalized regression data matrix is ​​dimensionless. A preset excitation threshold is applied to the minimum singular value of the normalized regression data matrix and is determined based on a stable identification experiment that yields acceptable parameter variance under a fixed rolling window length. This indicates the number of samples contained in the scrolling data window. Indicates the time up to the sampling point The regression data matrix is ​​formed by transposing and arranging the regression vectors within the window in rows. Indicates the first The first sample regression vector of the th sample... One portion, Indicates the column number of the regression data matrix; Indicates the first The normalized scales listed in the current window, Indicates the first The lower limit of the positive scale of the column; This represents a diagonal scale matrix consisting of two columns of scales. This represents the normalized regression data matrix. This indicates the construction of a diagonal matrix.

[0138] To avoid distortion in minimum singular value determination caused by differences in the dimensions and numerical scales of the columns in the regression matrix, each column is first normalized according to the root mean square scale within the rolling window. When the normalized regression data matrix has two columns of full rank and its minimum singular value is not less than the preset dimensionless excitation threshold, it is determined that the rank excitation condition is met; otherwise, the parameter update for this round is not initiated.

[0139] After satisfying the rank excitation condition, a recursive least squares method with an adaptive forgetting factor and sample robust weights is employed. Let the residual of the one-step wall temperature prediction be... The parameter update gain, parameter estimate, and parameter estimate covariance matrix are updated according to the following formulas:

[0140] ;

[0141] ;

[0142] in, Indicates the recursive gain; This represents the parameter estimation covariance matrix; Represents the identity matrix; Represents the adaptive forgetting factor; This represents the recursive weighting coefficient of the data samples. The forgetting factor varies between a preset lower limit and an upper limit. When the continuous residuals after robust processing reflect slow changes in thermal parameters, the forgetting factor is appropriately reduced to improve tracking speed; when the parameters are stable, the forgetting factor is made to approach the upper limit to reduce estimation fluctuations. Indicates the sampling time The updated parameter estimation vector, and Let these represent the parameter estimation vector and parameter estimation covariance matrix at the previous sampling time, respectively. Indicates the residual of one-step wall temperature prediction; superscript Represents the transpose of a vector or matrix.

[0143] During recursive least squares initialization, the initial parameter vector is calculated based on the initial wall equivalent heat capacity and outdoor heat transfer resistance registered during the installation phase. The parameter estimation covariance matrix is ​​a second-order symmetric positive definite matrix; the identity matrix is ​​a second-order identity matrix, and the recursive weight coefficients are in the range of greater than zero and not greater than one.

[0144] The adaptive forgetting factor can be determined based on the most recent residual magnitude:

[0145] ;

[0146] The residual statistical window length is a positive integer; the residual statistic is the median of the absolute values ​​of the predicted wall temperature residuals within the window, in Kelvin; the residual baseline threshold is a positive temperature threshold determined by stable operating samples; the lower and upper limits of the forgetting factor satisfy the condition that zero is less than the lower limit and the lower limit is not greater than the upper limit, and the upper limit is not greater than one; the adjustment coefficient is a non-negative dimensionless quantity. To accelerate the algorithm's tracking and approximation of the drift of real physical parameters while maintaining noise resistance, the specific value range of the adjustment coefficient is defined as 0.05 to 0.20. This value range is derived based on the expected tracking time constant when thermal parameters undergo a step change: in actual operation, when the residual statistic reaches twice the residual baseline threshold, the above adjustment coefficient can ensure that the forgetting factor smoothly decreases to between 0.90 and 0.95, thereby suppressing single-point impulse noise while giving sufficient weight to newly sampled data. When the residual statistic does not exceed the residual baseline threshold, the forgetting factor takes the upper limit; when the residual statistic continues to increase, the forgetting factor decreases within a preset range to improve the tracking speed of slow thermal parameter changes. Indicates the length of the residual statistics window. Indicates the time up to the sampling point The residual amplitude statistic, Indicates the first in the window One-step wall temperature prediction residual for a sample Indicates the residual baseline threshold. and These represent the lower and upper limits of the forgetting factor, respectively. This represents the forgetting factor adjustment coefficient. This means that the calculation result is restricted to a given lower and upper limit.

[0147] Unmodeled transient thermal disturbances in the indoor environment are identified through innovative residual analysis at air nodes. Using the most recently effective parameters and known actual variable convective thermal resistance, a one-step forward prediction of the air temperature is calculated without introducing compensation for unknown thermal disturbances.

[0148] ;

[0149] in, Indicates at the sampling time The forward prediction of the indoor air temperature at the next sampling time based on the current known state is denoted as follows: Indicates the time of use Information prediction time Define the innovative residual of air temperature as... .in, Indicates the next sampling time (i.e., the time of sampling). The measured indoor air temperature at each sampling time. The fault tolerance threshold is determined based on the median absolute deviation of the innovative residuals from the stable operating samples. Set the recursive weight coefficients according to the following formula:

[0150] ;

[0151] The stable residual window consists of samples where the device is in stable cooling mode and the sensors are effective, with coefficients set as positive dimensionless coefficients based on the allowable false alarm probability. To prevent the fault tolerance threshold from approaching zero when the stable sample dispersion is too small, the calculated result is used as the fault tolerance threshold when it is lower than the positive lower limit determined by the temperature sensor resolution and the one-step prediction numerical error. The termination threshold is a preset multiple greater than the fault tolerance threshold; when the absolute value of the innovative residual reaches a preset number of times and exceeds the termination threshold, the parameter update of the current data window is paused. This represents the set of sample indices corresponding to the stable residual window. and Each represents a sample index within the set. and These represent the innovative residuals of air temperature for the corresponding samples. Represents the positive dimensionless amplification factor, a constant. This represents the scaling factor for converting the median absolute deviation to the estimate of the normal distribution standard deviation. This indicates taking the median.

[0152] Specifically, the preset excitation threshold is determined as follows: In a stability identification test where the equipment is unloaded and a step control signal is applied, the singular value distribution of the normalized regression data matrix is ​​statistically analyzed, and the lower boundary of the smallest singular value within the 95% confidence interval is taken as the preset excitation threshold, typically ranging from 0.01 to 0.05. The positive dimensionless coefficient in the fault tolerance threshold... Based on the Raida criterion (3σ criterion) of normal distribution, when the allowable false alarm probability is set to 0.27%, take... When the allowable false alarm probability is set to 4.5%, take... The preset termination threshold is specifically set as the fault tolerance threshold. The weight of the sample is 3 to 5 times that of the target sample, and the number of consecutive times the preset value is typically 3 to 5 times. This setting ensures that the system only reduces the weight of the corresponding sample when encountering short-term random thermal disturbances such as people walking around, and can decisively stop the identification when the residuals are continuously stepped due to the doors and windows being open for a long time, thus preventing the wall parameters from being seriously contaminated.

[0153] ;

[0154] in, This indicates the innovative residual of air temperature; This represents the fault tolerance threshold. It is triggered when doors and windows are opened, people enter in large numbers, or indoor equipment suddenly overheats. As the absolute value of increases, the corresponding sample This decreases accordingly, thereby reducing the sample's impact on wall parameter updates. If the absolute value of the innovative residual continuously exceeds a higher termination threshold, parameter updates for the current data window are directly paused. Indicates the first The recursive weighting coefficients for each data sample.

[0155] Before convergence determination, the rows and columns of the parameter estimation covariance matrix are normalized according to the larger of the absolute value of each parameter estimation component and the corresponding lower bound of the positive scale, resulting in a dimensionless normalized covariance matrix. When the trace of the normalized covariance matrix is ​​lower than a preset dimensionless covariance threshold, and the relative change of two parameter estimation components between adjacent updates is lower than a preset convergence threshold for multiple consecutive periods, the parameter estimation is determined to have converged. Specifically, the preset dimensionless covariance threshold is set based on the system environmental noise tolerance after the regression data matrix is ​​scaled and normalized, typically ranging from 0.001 to 0.01; the preset convergence threshold is set to 1% to 3%, and the number of consecutive periods is set to 5 to 10 periods based on the control cycle length of the air conditioning equipment. Using this numerical combination ensures that, under steady-state conditions, the residual fluctuation rate of the thermal parameter estimates is limited to a physically acceptable control dead zone, avoiding the acceptance of parameters before convergence due to an excessively large threshold, or the algorithm getting stuck in infinite waiting due to an excessively small threshold. Then, the equivalent heat capacity of the wall and the outdoor thermal resistance are analyzed in reverse from the parameter estimates:

[0156] ;

[0157] in, and These represent the first and second estimated components of the parameter vector, respectively; This represents the estimated equivalent heat capacity of the wall. This represents the estimated outdoor thermal resistance. The controller pre-sets wall property boundaries based on the target wall's material type, thickness, area, and allowable heat transfer coefficient range. Only... and Only when all parameters are within the corresponding boundaries are they registered as the new most recently valid parameters; otherwise, the results of this round are discarded and the previous most recently valid parameters are continued to be used.

[0158] Before reverse engineering, it is checked that the estimated components of both parameters are positive and far from zero. The wall's equivalent heat capacity boundary is calculated from the allowable range of the target wall material's volume, density, and specific heat capacity; the outdoor heat transfer resistance boundary is calculated from the allowable range of the target wall's effective heat transfer area and heat transfer coefficient. The specific mathematical conversion relationship is: the lower limit of the wall's equivalent heat capacity. upper limit Lower limit of outdoor thermal resistance upper limit .in, For the target wall material volume, This refers to the effective heat exchange area of ​​the target wall. Based on this, the permissible range of physical parameters for different building structures is established upon standard building thermal fundamental parameters. For example, for solid brick or concrete exterior walls of typical civil buildings, density... The allowable range is selected as Specific heat capacity The allowable range is selected as heat transfer coefficient The allowable range is selected as The controller incorporates the aforementioned baseline correspondence table. Lightweight partition walls or glass curtain walls respectively utilize their corresponding low heat capacity and high heat transfer coefficient ranges, ensuring that the parameters obtained from reverse analysis have accurate physical significance. Boundaries can be selected based on the wall type during equipment installation and are only allowed to be adjusted when there is an engineering basis. and These represent the lower and upper limits of the wall's equivalent heat capacity, respectively. and These represent the lower and upper limits of outdoor thermal resistance, respectively. and These represent the lower and upper limits of the target wall material density, respectively. and These represent the lower and upper limits of the specific heat capacity of the target wall material, respectively. and These represent the lower and upper limits of the target wall heat transfer coefficient, respectively.

[0159] When the controller is first put into operation, initial values ​​for the wall's equivalent heat capacity and outdoor thermal resistance are determined based on the target wall material type, thickness, effective heat exchange area, density, specific heat capacity, and allowable range of heat transfer coefficient. These initial values ​​are registered as the most recently valid parameters after installation, commissioning, and verification that the controller can generate stable temperature predictions. If online identification results that have passed rank excitation, convergence, and wall property boundary checks have not yet been obtained, the aforementioned most recently valid parameters are used. After the first valid identification result is registered, this result replaces the initial values.

[0160] Step S105: Determine whether the rebound triggering condition based on the target wall's delayed heat release estimation is met. For example... Figure 5 As shown, the controller first calculates the absolute temperature difference between the indoor air temperature and the target set temperature. Only when this absolute temperature difference is not greater than the set temperature difference threshold will it enter the target wall heat release check, to avoid prematurely limiting the cooling capacity while the room is still in a high-load cooling phase.

[0161] Let the target set temperature be The preset rebound margin is Set the heat release threshold as The transient heat release power of the current target wall is:

[0162] ;

[0163] The controller keeps the current compressor frequency, current indoor fan speed, and corresponding air guide vane deflection angle constant within the prediction window, and uses a two-node thermal resistance-capacitance model to deduce the baseline predicted temperature trajectory. For any adjacent prediction step, the room temperature rebound rate is calculated according to the following formula:

[0164] ;

[0165] in, This indicates the current transient heat release power of the target wall. Indicates the current time For prediction step The predicted value of indoor air temperature Indicates the current time For prediction step Predicted indoor air temperature; This represents the rate of room temperature change between adjacent prediction steps. Joint optimization is triggered when the following combination of conditions is met: the absolute temperature difference between the indoor air temperature and the target set temperature is not greater than the set temperature difference threshold; the target wall boundary temperature is higher than the indoor air temperature. Greater than Furthermore, the highest temperature of the baseline predicted temperature trajectory is higher than... Or, a preset number of prediction steps. All are greater than the set rebound rate threshold. Among them, Indicates the prediction step number. Indicates the time interval between adjacent prediction steps; , and These represent the current target wall boundary temperature, the current indoor air temperature, and the current actual variable parameter convective thermal resistance, respectively.

[0166] The unit of transient heat release power of the target wall is watts, and the unit of room temperature rebound rate is Kelvin per second or degrees Celsius per second; the set heat release threshold and the set rebound rate threshold are respectively set in the same unit as the corresponding physical quantity.

[0167] The set temperature difference threshold is no greater than the room temperature regulation bandwidth allowed by the air conditioning controller; the preset rebound margin is determined by the allowable comfort deviation above the target set temperature; the set rebound rate threshold is determined by the allowable increase in room temperature per unit time. The set heat release threshold is determined based on the upper bound of the estimated error of the target wall heat release power and the upper quantile of the heat release power of historical samples without rebound, and the larger of the two values ​​is taken. The above thresholds are registered after sensor calibration and short-term trial operation in the same room, and will not be automatically relaxed based on a single abnormal event during operation.

[0168] In specific engineering applications, the set temperature difference threshold is typically set to a value of [value missing]. This is to ensure that the room temperature enters the steady-state control region; the typical value of the preset rebound margin is... to The set rebound rate threshold value is... Furthermore, the upper bound of the heat release power estimation error can be set by normalization using the equipment's maximum cooling capacity, typically taken as the equipment's rated sensible heat cooling capacity. For example, a unit with a rated sensible heat cooling capacity of The device has a set heat release threshold. It can be specifically quantified as A fixed value between.

[0169] Step S106: Generate multiple sets of joint control sequences, and use the unit performance mapping model to obtain the predicted sensible heat cooling capacity and estimated power consumption step by step. The controller selects candidate values ​​from the allowed compressor frequency set and indoor fan speed set respectively, and constructs a joint control sequence with the number of prediction steps based on the compressor frequency change rate, minimum hold time, fan speed change rate, and noise limit construction length. To reduce the computational load, a segmented hold method for the control quantity can be adopted, allowing the control quantity to change gradually in the first few prediction steps, while keeping the control quantity unchanged for several steps in subsequent prediction steps; each prediction step still uses a uniform prediction time step.

[0170] The unit performance mapping model is represented as follows:

[0171] ;

[0172] in, This represents the unit performance mapping model; Indicates the candidate frequency of the compressor; Indicates the candidate speed of the indoor fan; This indicates the predicted return air dry-bulb temperature; This indicates the predicted relative humidity of the return air; This indicates the predicted outdoor ambient temperature; This indicates the predicted sensible heat cooling capacity; and These represent the estimated power consumption of the compressor and the estimated power consumption of the fan, respectively.

[0173] The unit for the candidate compressor frequency is Hertz, and the unit for the candidate indoor fan speed is revolutions per minute or the equivalent speed unit fed back by the equipment. The units for the predicted return air dry-bulb temperature and the predicted outdoor ambient temperature are degrees Celsius. The predicted return air relative humidity uses a dimensionless value or percentage from zero to one and is consistent at the model input. The units for the predicted sensible heat cooling capacity, the estimated power consumption of the compressor, and the estimated power consumption of the fan are all watts.

[0174] The unit performance mapping model can be implemented using multidimensional lookup tables and partitioned interpolation. Modeling data comes from valid historical operating data of the equipment under different combinations of compressor operating frequencies, indoor fan operating speeds, and environmental conditions. The calibrated sensible heat cooling capacity in each data set is still calculated using air mass flow rate and the dry-bulb temperature difference between the supply and return air. Compressor power consumption and fan power consumption are determined by feedback values ​​from electrical parameter detection terminals and fan drive. Candidate sequences are not generated for extrapolation regions lacking valid samples to avoid using performance mapping results unsupported by operating data. The performance mapping model can be incrementally updated when the equipment is operating stably and the sensors are valid, but the updated mapping values ​​must satisfy the physical boundaries of sensible heat cooling capacity variation with compressor frequency, the non-negative power consumption boundary, and the unit's rated capacity boundary.

[0175] Within the short prediction window, the predicted return air dry-bulb temperature is the sum of the predicted indoor air temperature and the median deviation of the return air dry-bulb temperature from the indoor air temperature in recent valid samples; the predicted return air relative humidity remains the current valid return air relative humidity when an independent humidity prediction model is lacking, and the predicted value of the humidity state model is used when a humidity state model is configured; the predicted outdoor ambient temperature uses the current valid value or the external meteorological forecast value. This ensures that the environmental condition inputs to the unit performance mapping model have a clear source at each prediction step.

[0176] Step S107: The dual-node thermal resistance-capacitance model is used as the state prediction equation to deduce the predicted temperature trajectory corresponding to each joint control sequence. The most recently effective equivalent heat capacity of the wall and the outdoor thermal resistance are used as fixed parameters, and the current measured indoor air temperature and the target wall boundary temperature are used as the initial prediction state. In each prediction step, the variable parameter convective thermal resistance is first calculated based on the candidate indoor fan speed and the corresponding air guide deflection angle, and then the predicted sensible heat cooling capacity output by the unit performance mapping model is used as the air node cooling capacity input.

[0177] When using the forward discrete form, the state recursion relationship for each prediction step is as follows:

[0178] ;

[0179] ;

[0180] in, This represents the predicted temperature of the target wall boundary. This indicates the predicted outdoor ambient temperature. and This indicates the most recently valid wall parameters currently in use; This indicates the amount of unmodeled thermal disturbance in the prediction step. The finite amplitude compensation value. Within a short prediction window, the most recent outdoor ambient temperature can be kept constant, or external meteorological forecasts can be used. For unmodeled thermal disturbances at air nodes, compensation is performed by converting the mean of recent innovative residuals after outlier removal into thermal power; if a reliable compensation value is lacking, then... The value is set to zero to avoid extrapolating a one-time disturbance across the entire prediction window. Among these, This indicates the predicted indoor air temperature. Indicates the prediction step Variable parameter convective thermal resistance, Indicates the prediction step Sensible heat cooling capacity. This represents the equivalent heat capacity of air. Indicates the prediction step size; subscript Indicates the current sampling time The prediction result for the state of the next prediction step.

[0181] In the formula, the predicted indoor air temperature, the predicted target wall boundary temperature, and the predicted outdoor ambient temperature all use the same temperature scale; the units for the predicted sensible heat cooling capacity and the compensation value for unmodeled thermal disturbances are watts, and the unit for the prediction step size is seconds. To ensure the numerical stability of the forward discrete step, the prediction step size is no greater than a preset proportion of the minimum thermal time constant calculated from the current heat capacity and thermal resistance; if this is not satisfied, a control cycle is subdivided into multiple integral substeps, and the control quantity remains unchanged within each substep.

[0182] The unmodeled thermal disturbance compensation value is obtained by converting the recent residual of air temperature innovation that was not judged as abnormal:

[0183] ;

[0184] The disturbance estimation window includes only samples where the sensor is active and has not triggered the termination threshold. The upper limit of the disturbance power is a positive value set according to the allowable internal thermal gain range of the room, in watts. Within the prediction window, the unmodeled thermal disturbance compensation value decays progressively according to a discrete exponential decay model. The relationship between the compensation values ​​for each prediction step is as follows: ,in A preset attenuation coefficient is used, with a value ranging from 0.85 to 0.95. Using this exponential attenuation relationship and coefficient value, the natural heat dissipation dynamics of typical transient thermal disturbances such as short-term opening and closing of doors and windows or movement of people in an indoor space can be accurately reflected. When the effective sample size is insufficient, the compensation value is set to zero. Indicates the sampling time Unmodeled thermal disturbance power estimate This represents the set of sample indices for the perturbation estimation window. This indicates the sample index within the window. This represents the innovative residual of air temperature for the corresponding sample. This indicates the upper limit of the disturbance power. This indicates that the calculation results will be restricted to... Within the range. In the attenuation relationship, Indicates by The initial value of the current disturbance compensation has been determined. Indicates the current sampling time Next The perturbation compensation value for each prediction step. This indicates the preset attenuation coefficient. This indicates the prediction step number.

[0185] Step S108 involves constraint screening and comprehensive cost calculation for the joint control sequences. First, the lowest and highest predicted temperatures are extracted from each predicted temperature trajectory. If the lowest predicted temperature is below the preset lower limit of room temperature, or the highest predicted temperature is above the preset upper limit of room temperature, the sequence is discarded. Then, the target wall transient heat release power and room temperature rebound rate for each prediction step are calculated, and sequences with any index exceeding the corresponding hard constraint upper limit are discarded. The controller also checks the compressor minimum operating frequency and the evaporation pressure safety lower limit. The evaporation pressure safety lower limit is determined using a pre-calibrated unit safety operating envelope, indexed by compressor frequency, indoor fan speed, return air status, and outdoor ambient temperature, recording allowed combinations verified through unit testing. When the equipment is equipped with an evaporation pressure sensor, further verification is performed using the current evaporation pressure and the pressure change margin corresponding to the candidate combinations. Any prediction step exceeding the verified safety operating envelope or failing to meet the evaporation pressure safety lower limit is excluded from cost comparison. The unit safety operating envelope has a clear logical decision boundary: for example, when the predicted return air dry-bulb temperature is below... And the predicted outdoor ambient temperature is lower than At that time, the lower limit of the compressor's minimum operating frequency was forcibly increased to... Simultaneously, by consulting tables, it is ensured that the predicted evaporation pressure corresponding to any control combination is not lower than... Otherwise, the control sequence is directly rejected.

[0186] For the remaining compliant sequences, a dimensionless comprehensive cost function is adopted:

[0187] ;

[0188] ;

[0189] ;

[0190] ;

[0191] In this embodiment, the comprehensive cost value is dimensionless, the number of prediction steps is a positive integer, and the duration of each prediction step is a positive value. When a uniform prediction step size is used, the duration of each prediction step is equal to the sampling period. Temperature tracking, equipment energy consumption, transient heat release power of the target wall, and room temperature rebound rate correspond to four types of penalty terms. The transient heat release power of the target wall is determined by dividing the difference between the wall boundary temperature and the indoor air temperature at the corresponding prediction step by the variable parameter convective thermal resistance and taking a non-negative value. The temperature reference value is positive and takes the allowable room temperature deviation. The compressor reference energy and fan reference energy are positive and take the rated or calibrated energy consumption of the corresponding actuator within a reference control cycle. The heat release power reference value and the rebound rate reference value are both positive soft reference values, not exceeding the corresponding hard constraint upper limit, and preferably taking a set heat release threshold and a set rebound rate threshold; thus, feasible sequences between the soft reference value and the hard constraint upper limit are still penalized. All weights are non-negative dimensionless coefficients. After the above normalization, each penalty item has a comparable magnitude, and weights are then set based on comfort priority and equipment energy-saving goals. The specific weight configuration is switched in a matrix manner according to the user-selected operating mode. For example, when the equipment is in comfort-priority mode, temperature tracking and rebound penalty items dominate, and the weight combination can be set as follows: When the device is in energy-saving priority mode, energy consumption penalty items take precedence, and the weight combination switches to: This explicit weighting matrix allows the cost function to output deterministic numerical results for comparison. Represents the combined cost of a set of joint control sequences. Indicates the number of prediction steps; , , and These represent the prediction steps. Temperature tracking error penalty, equipment energy consumption penalty, target wall transient heat release power penalty, and room temperature rebound rate penalty; , , , and These represent the weights corresponding to temperature tracking, compressor energy consumption, fan energy consumption, target wall heat release, and room temperature rebound, respectively. and These represent the predicted indoor air temperature and the target set temperature, respectively. This indicates a reference value for temperature deviation. and These represent the estimated power consumption of the compressor and the estimated power consumption of the fan, respectively. Indicates the first The duration of each prediction step and These represent the compressor reference energy and the fan reference energy, respectively. This indicates the predicted transient heat release power of the target wall. This indicates a reference value for heat release power. Indicates the room temperature rebound rate. This indicates a reference value for the rebound rate. This indicates the larger of zero and the corresponding excess amount.

[0192] The joint control sequence with the lowest overall cost is selected as the optimal joint control sequence. Only the compressor frequency and indoor fan speed of the first control step of this sequence are issued. At the next sampling time, the sequence is re-identified, predicted, and optimized based on the newly acquired sensor data to form rolling time-domain control. If all candidate sequences are eliminated due to constraints, a conservative control strategy that meets the equipment safety limits is prioritized, such as maintaining the current fan speed and limiting the compressor frequency within a safe frequency range that does not trigger the lower limit of the evaporation pressure.

[0193] The following are examples illustrating the calculation process. The equipment capacity, room size, and threshold values ​​in these examples are exemplary values ​​and do not constitute a limitation on the scope of protection of this invention. The target room uses a variable frequency air conditioning unit with a control cycle of sixty seconds and a target set temperature of twenty-six degrees Celsius. The exterior wall area covered by the air jet and having a large positive temperature difference relative to the indoor air is defined as the target wall. The effective heat exchange area of ​​the target wall is twenty-four square meters, and its characteristic length is two point eight meters.

[0194] Within a data window satisfying the rank excitation condition, the controller obtains the actual variable-parameter convective thermal resistance at each sampling time based on the local air velocity and the empirical formula for the Nusselt number. The minimum singular value of the regression data matrix is ​​higher than the preset excitation threshold. After covariance convergence checks and wall property boundary checks, the recursive least squares result yields the equivalent heat capacity of the wall. Outdoor thermal resistance is Both are then registered as the most recently valid parameters. This value is only used to demonstrate how the parameters are incorporated into subsequent calculations.

[0195] At a subsequent sampling time, the indoor air temperature was 25.9 degrees Celsius, the target wall boundary temperature was 28.2 degrees Celsius, and the current actual variable parameter convective thermal resistance was... The absolute temperature difference between room temperature and the target set temperature is less than the set temperature difference threshold, and the transient heat release power of the target wall is approximately 271 watts, exceeding the set heat release threshold. Maintaining the highest predicted temperature derived from the current operating state is higher than the sum of the target set temperature and the preset rebound margin, thus triggering joint optimization.

[0196] As an example, the set temperature difference threshold can be set to 0.3 degrees Celsius, the set heat release threshold to 220 watts, and the preset rebound margin to 0.4 degrees Celsius. If the highest predicted temperature obtained from the current operating state is higher than 26.4 degrees Celsius, or if the room temperature rebound rate exceeds the corresponding threshold for a consecutive preset number of prediction steps, then joint optimization is triggered. The above values ​​are only used to illustrate the triggering logic; the actual values ​​are determined according to the aforementioned calibration method.

[0197] The controller uses 24 Hz, 28 Hz, and 32 Hz as the currently allowed candidate compressor frequencies, and 450 rpm, 600 rpm, and 750 rpm as the currently allowed candidate indoor fan speeds, generating a joint control sequence while satisfying the rate of change and minimum hold-up time. For each sequence, the unit performance mapping model outputs the sensible heat cooling capacity, compressor power consumption, and fan power consumption step by step, while the aerodynamic correlation mapping function outputs the predicted mapped air velocity. After airflow organization correction, the variable parameter convective thermal resistance is updated step by step. After screening based on room temperature, heat release power, rebound rate, minimum operating frequency, and evaporation pressure constraints, the controller compares the comprehensive cost of the remaining sequences, selects the sequence with the lowest cost, and executes its first control step. Thus, the influence of fan speed on near-wall heat transfer and the influence of compressor frequency on sensible heat cooling capacity are jointly evaluated in the same prediction process.

[0198] Parameter identification and control optimization employ different validity thresholds. Even if the current regression data matrix does not meet the rank excitation condition, the controller can still use the most recently valid parameters for rebound prediction and joint optimization; however, it will not use data lacking excitation to update wall parameters. When the target wall boundary temperature sensor, local air velocity sensor, or supply and return air temperature sensor is determined to be faulty, parameter updates relying on the corresponding measurements cease, and the system switches to the original temperature control strategy or validated conservative parameters, without using incomplete data to generate new wall parameters.

[0199] For rooms with multiple main exterior walls or large glass curtain walls, wall nodes can be established for each target wall, and the dynamic convective heat transfer from each wall node to the air node can be summed. Alternatively, equivalent target wall boundary temperature, equivalent wall heat capacity, and equivalent thermal resistance can be derived based on the effective heat transfer area and historical heat release contribution of each wall region. Regardless of whether a single-wall or multi-wall equivalent form is used, the temperature, thermal resistance, and area involved in parameter identification and prediction all use the same target wall definition to avoid mismatch between the temperature measurement area and the heat transfer area.

[0200] Figure 6The system structure of the present invention is shown. The acquisition module is used to simultaneously acquire the indoor air temperature sequence, the target wall boundary temperature sequence, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, the actual indoor fan speed sequence, the actual deflection angle sequence of the air guide plate, and the local air velocity sequence, and to complete time alignment and validity marking.

[0201] The model building module is used to construct a two-node thermal resistance-capacitance model. It determines the actual variable-parameter convective thermal resistance at each historical sampling time based on the actual indoor fan speed, the actual deflection angle of the air guide vane, and the local air velocity, and uses this as known time-varying model parameters. The model building module also saves the effective heat transfer area, characteristic length, air properties, and empirical Nusselt number coefficients of the target wall.

[0202] The identification module constructs a regression data matrix and checks the rank excitation condition. When the rank excitation condition is met, it performs recursive least squares fitting with an adaptive forgetting factor and robust weights, completing covariance convergence checks, inverse analysis of physical parameters, and boundary checks of wall properties. If any check fails, the current update is stopped and the most recently valid parameters are read. The identification module also innovatively identifies unmodeled transient thermal disturbances in the room based on air temperature residuals and adjusts the sample recursive weights.

[0203] The sequence generation module generates multiple sets of joint control sequences based on the allowable operating sets, rate of change limits, and hold-up time limits of the compressor and indoor fan when the rebound trigger condition is met. The mapping and conversion module calls the unit performance mapping model to convert each joint control sequence into predicted sensible heat cooling capacity, estimated compressor power consumption, and estimated fan power consumption step by step.

[0204] The deduction module uses a two-node thermal resistance-capacitance model as the state prediction equation. It dynamically updates the variable-parameter convective thermal resistance based on the candidate indoor fan speed and corresponding guide vane deflection angle at each prediction step, and deduces the indoor air temperature and target wall boundary temperature. The optimization execution module performs checks on room temperature, target wall heat release power, room temperature rebound rate, compressor minimum frequency, and evaporation pressure safety lower limit. It calculates the comprehensive cost function of the compliant sequence, selects the joint control sequence with the lowest cost value, and sends the first control step to the equipment control terminal.

[0205] The aforementioned modules can be deployed in the local controller of the air conditioning and refrigeration equipment, the building energy management controller, or a control platform composed of both. Data acquisition and security protection are preferentially performed by the local controller, while parameter identification and candidate sequence solving can be performed by a processor with corresponding computing capabilities. Regardless of whether each module is implemented in software, hardware, or a combination of both, the aforementioned data and control relationships are maintained, with the original equipment security protection as the highest priority.

Claims

1. A method for intelligent energy-saving optimization of air conditioning refrigeration equipment based on multi-sensor monitoring, characterized in that, include: Simultaneously acquire the indoor air temperature sequence, the target wall boundary temperature sequence within the air jet coverage area, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, as well as the indoor fan actual speed sequence, the air guide plate actual deflection angle sequence, and the local air velocity sequence at the location adjacent to the target wall. A two-node thermal resistance-capacitance model is constructed, which includes the equivalent heat capacity of the wall, the outdoor heat transfer thermal resistance, and the variable parameter convective thermal resistance that varies with airflow organization. Based on the actual speed sequence of the indoor fan, the actual deflection angle sequence of the air guide plate, and the local air velocity sequence, the actual variable parameter convective thermal resistance is determined at each sampling time, and the actual variable parameter convective thermal resistance is used as a known time-varying parameter to participate in the wall parameter identification. When the regression data matrix consisting of observations and known quantities from multiple consecutive sampling periods satisfies the rank excitation condition, the equivalent heat capacity of the wall and the outdoor heat transfer resistance are identified and updated online based on the indoor air temperature sequence, the target wall boundary temperature sequence, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, and the actual variable parameter convective thermal resistance; when the rank excitation condition is not met, the parameter update for this round is stopped and the most recently valid parameter is used. When the rebound trigger condition based on the hysteresis heat release estimation of the target wall is met, multiple sets of joint control sequences containing candidate compressor frequencies and candidate indoor fan speeds are generated. Based on the unit performance mapping model, the joint control sequence of each group is converted step by step into a predicted sensible heat cooling capacity sequence, a compressor estimated power consumption sequence, and a fan estimated power consumption sequence. For each group of joint control sequences, the dual-node thermal resistance-capacity model is used as the state prediction equation, the corresponding predicted sensible heat cooling capacity sequence is used as the cooling capacity input, and the variable parameter convective thermal resistance is updated step by step using the corresponding indoor fan candidate speed to deduce and predict the temperature trajectory. From the joint control sequences that satisfy temperature, target wall heat release, room temperature rebound, and equipment safety constraints, the optimal joint control sequence with the minimum comprehensive cost function is selected and executed.

2. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The dual-node thermal resistance-capacitance model includes: Establish the differential equation for the thermal balance of the air node, where the product of the equivalent heat capacity of the air and the rate of change of the indoor air temperature is equal to the sum of the dynamic convective heat transfer from the target wall to the indoor air and the unmodeled transient thermal disturbance in the room, minus the sensible heat cooling capacity. Establish a differential equation for the thermal balance of the wall nodes, wherein the product of the equivalent heat capacity of the wall and the rate of change of the boundary temperature of the target wall is equal to the outdoor heat transfer from the outdoor environment to the target wall minus the dynamic convection heat transfer. The dynamic convective heat transfer is obtained by dividing the difference between the target wall boundary temperature and the indoor air temperature by the variable parameter convective thermal resistance; the outdoor heat transfer is obtained by dividing the difference between the outdoor ambient temperature and the target wall boundary temperature by the outdoor heat transfer thermal resistance.

3. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The synchronous acquisition of the sensible heat cooling capacity sequence includes: Real-time acquisition of return air dry-bulb temperature and supply air dry-bulb temperature of air conditioning and refrigeration equipment; The air volume flow rate is determined based on the actual speed of the indoor fan and the pre-tested duct resistance curve, and then converted into real-time air mass flow rate by combining the air density corresponding to the return air state. The sensible heat cooling capacity at the corresponding sampling time is obtained by multiplying the real-time air mass flow rate, the air constant pressure specific heat capacity, and the absolute value of the difference between the return air dry-bulb temperature and the supply air dry-bulb temperature. When the supply air dry-bulb temperature is not lower than the return air dry-bulb temperature, or when the air conditioning refrigeration equipment is in an unstable cooling state, the corresponding data frame is marked as not participating in wall parameter identification.

4. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The online identification and updating of the wall's equivalent heat capacity and outdoor thermal resistance includes: The differential equation of thermal equilibrium of the wall nodes in the two-node thermal resistance-capacitance model is discretized into an autoregressive exogenous input difference equation. Observations and known quantities from multiple consecutive sampling periods are extracted to construct a regression data matrix, and the scale of each column of the regression data matrix is ​​normalized. When the normalized regression data matrix has full column rank and the minimum singular value is not less than the preset excitation threshold, it is determined that the rank excitation condition is met, and rolling fitting is performed using the recursive least squares method with adaptive forgetting factor and sample robust weight. When the parameter estimation covariance matrix satisfies the preset convergence condition, and the wall equivalent heat capacity and outdoor heat transfer resistance obtained by inverse analysis are both positive and within the preset wall attribute boundary, the two are registered as new most recent effective parameters. When the rank excitation condition, the preset convergence condition, or the preset wall attribute boundary is not met, the identification results of this round are discarded and the most recent valid parameters are used.

5. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 4, characterized in that, The recursive least squares method also includes a disturbance-resistant weighting step based on the innovative residual of air temperature: Using the most recent effective parameters and the actual variable convective thermal resistance at the corresponding sampling time, the one-step forward prediction value of indoor air temperature is calculated; Subtract the previous forward prediction value from the measured indoor air temperature at the next sampling time to obtain the innovative residual of air temperature. When the absolute value of the innovative residual of air temperature is not greater than the fault tolerance threshold, the recursive weight coefficient of the corresponding data sample is set to 1; when the absolute value is greater than the fault tolerance threshold, the recursive weight coefficient is set to the ratio of the fault tolerance threshold to the absolute value. When the absolute value of the innovative residual of the air temperature continuously exceeds the preset stop threshold, the parameter update of the current data window is suspended to isolate transient thermal disturbances introduced by opening doors and windows, personnel activities, or indoor equipment heating.

6. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The step-by-step prediction update of the variable parameter convective thermal resistance includes: The ratio of the latest measured local air velocity to the aerodynamic correlation mapping function is calculated under the latest actual indoor fan speed and the latest actual deflection angle of the air guide plate. The resulting ratio is then limited between a preset lower limit and a preset upper limit to obtain the airflow organization correction coefficient. Within the prediction window, the predicted mapped air velocity, determined by the candidate speed of the indoor fan and the deflection angle of the corresponding air guide plate, is calculated in each prediction step using the aerodynamic correlation mapping function, and the predicted mapped air velocity is corrected using the airflow organization correction coefficient. The boundary layer Reynolds number is calculated based on the corrected local air velocity, the characteristic length of the target wall, and the air kinematic viscosity. Substituting the boundary layer Reynolds number into the Nusselt number correlation corresponding to the geometry and flow state of the target wall, the dynamic convective heat transfer coefficient is obtained. The reciprocal of the product of the dynamic convective heat transfer coefficient and the effective heat transfer area of ​​the target wall is taken as the variable parameter convective thermal resistance of the corresponding prediction step and substituted into the state prediction equation.

7. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The rebound triggering conditions include: The absolute temperature difference between the indoor air temperature and the target set temperature shall not exceed the set temperature difference threshold. The target wall boundary temperature is higher than the indoor air temperature, and the transient heat release power of the target wall, obtained by dividing the temperature difference between the two by the current actual variable parameter convective thermal resistance, is greater than the set heat release threshold. The baseline predicted temperature trajectory, derived by maintaining the current compressor frequency, indoor fan speed, and air guide deflection angle, satisfies at least one of the following: the highest predicted temperature of the baseline predicted temperature trajectory is higher than the sum of the target set temperature and the preset rebound margin; the room temperature rebound rate of a consecutive preset number of prediction steps is greater than the set rebound rate threshold.

8. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The step of selecting the optimal joint control sequence and issuing it for execution includes: Extract the lowest and highest predicted temperatures from the predicted temperature trajectories of each group; The joint control sequence that is lower than the preset room temperature lower limit or higher than the preset room temperature upper limit is excluded; The joint control sequence that excludes any prediction step within the prediction window where the transient heat release power of the target wall exceeds the set upper limit of heat release power or the room temperature rebound rate exceeds the set upper limit of rebound rate is excluded. Eliminate joint control sequences that violate the minimum operating frequency limit of the compressor or the lower limit of the safe evaporation pressure determined by the pre-calibrated safe operating envelope of the unit; Substitute the remaining joint control sequence into the comprehensive cost function, which is a weighted sum of the temperature tracking error penalty term, the compressor energy consumption penalty term, the fan energy consumption penalty term, the target wall transient heat release power penalty term, and the room temperature rebound rate penalty term. The joint control sequence with the smallest output value of the comprehensive cost function is selected, and only the first control step of the joint control sequence is issued. At the next sampling time, prediction and optimization are re-executed based on the newly acquired sensor data.

9. The intelligent energy-saving optimization method for air conditioning refrigeration equipment based on multi-sensor monitoring according to claim 1, characterized in that, The establishment and invocation of the unit performance mapping model includes: Acquire valid historical operating data under different combinations of compressor operating frequency, indoor fan operating speed and environmental conditions, wherein the environmental conditions include at least return air dry bulb temperature, return air relative humidity and outdoor ambient temperature; The sensible heat cooling capacity is calculated based on the air mass flow rate and the dry-bulb temperature difference of the supply and return air corresponding to each group of valid historical operating data, and the corresponding compressor power consumption and fan power consumption are obtained. Establish a unit performance mapping model with compressor operating frequency, indoor fan operating speed and the aforementioned environmental conditions as inputs, and sensible heat cooling capacity, compressor power consumption and fan power consumption as outputs; Within the prediction window, the candidate compressor frequency, candidate indoor fan speed, and predicted environmental conditions corresponding to each prediction step are input into the unit performance mapping model to obtain the corresponding predicted sensible heat cooling capacity, estimated compressor power consumption, and estimated fan power consumption. For input combinations that exceed the coverage of valid historical operating data, no corresponding joint control sequence is generated.

10. An intelligent energy-saving optimization system for air conditioning refrigeration equipment based on multi-sensor monitoring, characterized in that, include: The acquisition module is configured to simultaneously acquire the indoor air temperature sequence, the target wall boundary temperature sequence within the coverage area of ​​the air jet, the outdoor ambient temperature sequence and the sensible heat cooling capacity sequence, as well as the indoor fan actual speed sequence, the air guide plate actual deflection angle sequence and the local air velocity sequence at the location adjacent to the target wall. The model building module is configured to build a two-node thermal resistance-capacity model that includes the equivalent heat capacity of the wall, the outdoor heat transfer thermal resistance, and the variable parameter convective thermal resistance that varies with airflow organization. The actual variable parameter convective thermal resistance is determined at each sampling time according to the actual speed sequence of the indoor fan, the actual deflection angle sequence of the air guide plate, and the local air velocity sequence. The actual variable parameter convective thermal resistance is provided to the identification module as a known time-varying parameter. The identification module is configured to, when the regression data matrix satisfies the rank excitation condition, identify and update the equivalent heat capacity of the wall and the outdoor heat transfer resistance online based on the indoor air temperature sequence, the target wall boundary temperature sequence, the outdoor ambient temperature sequence, the sensible heat cooling capacity sequence, and the actual variable parameter convective thermal resistance; when the rank excitation condition is not satisfied, the most recent effective parameter is used. The sequence generation module is configured to generate multiple sets of joint control sequences, including candidate compressor frequencies and candidate indoor fan speeds, when the rebound trigger condition based on the target wall's delayed heat release estimation is met. The mapping and conversion module is configured to convert the joint control sequence of each group into a predicted sensible heat cooling capacity sequence, a compressor estimated power consumption sequence, and a fan estimated power consumption sequence step by step based on the unit performance mapping model. The deduction module is configured to use the two-node thermal resistance-capacitance model as the state prediction equation, the corresponding predicted sensible heat cooling capacity sequence as the cooling capacity input, and update the variable parameter convective thermal resistance step by step using the corresponding indoor fan candidate speed to deduce and predict the temperature trajectory for each group of joint control sequences. The optimization execution module is configured to select the optimal joint control sequence with the minimum comprehensive cost function from the joint control sequences that satisfy temperature, target wall heat release, room temperature rebound, and equipment safety constraints, and then issue it for execution.

Images