A deep learning-based air-cooling air conditioning control method and system

Abstract

The present invention relates to the field of air-cooled air-conditioning control technology, specifically a method and system for air-cooled air-conditioning control based on deep learning, comprising the following steps: obtaining sensor readings of indoor temperature, outdoor temperature and humidity and Wi-Fi channel status information, determining the number of personnel positions and activity status, calculating the personnel position distribution in combination with a time window, and integrating to obtain a multi-dimensional environment and load state vector. In the present invention, the multi-dimensional environment and load state vector integrates indoor and outdoor temperature, humidity, Wi-Fi channel status information, and personnel position distribution and activity status, dynamically calculates the trend of personnel density changes through a time window, enhances the comprehensiveness and real-time nature of environmental perception, and provides high-dimensional data support for control decisions. A basic comfort deviation index is constructed based on the temperature deviation value and the humidity deviation value, and a multi-objective performance index set is generated in combination with the wind speed value, the compressor start-stop frequency, and the energy consumption power.

Description

Technical Field

[0001] The present invention relates to the technical field of air-cooling air conditioning control, and in particular to an air-cooling air conditioning control method and system based on deep learning. Background Art

[0002] Air-cooled air conditioning control is a sub-branch of the HVAC technology field, focusing on controlling refrigerant circulation, air flow rate, temperature distribution, and humidity levels by adjusting the operating parameters of core components such as compressors, fans, and air guides. In existing technologies, temperature overshoot or undershoot can occur when fluctuations in occupant density or local airflow disturbances are not quantified. The separate calculation of start-stop frequency and energy consumption indicators makes it difficult to balance equipment operating efficiency and user comfort. Frequent starts and stops exacerbate compressor wear and increase energy consumption. Therefore, improvements are needed. Summary of the Invention

[0003] The purpose of the present invention is to solve the shortcomings of the prior art and propose a deep learning-based air-cooling air conditioning control method and system.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a deep learning-based air-cooling air conditioning control method, comprising the following steps:

[0005] Obtain sensor readings of indoor temperature, outdoor temperature, and humidity, as well as Wi-Fi channel status information, to determine the number of people and their activity status. Combined with the time window, the distribution of people's locations is estimated, and the multi-dimensional environment and load state vectors are integrated.

[0006] Based on the multi-dimensional environment and load state vector and the user-set temperature, a temperature deviation value and a humidity deviation value are calculated to obtain a basic comfort deviation value. Based on the basic comfort deviation value, a wind discomfort index is calculated in combination with the current wind speed value, a start-stop frequency index is calculated by correlating the compressor operation record, and a current power value is calculated by correlating the energy consumption data to obtain a set of multi-objective performance indicators.

[0007] Based on the multi-dimensional environment and load state vector and the multi-objective performance indicator set, and in accordance with preset operating rules, potential control action combinations that meet the conditions are screened to obtain a list of candidate control parameters. Based on the list of candidate control parameters, a set of instructions specifying the compressor frequency, the fan speed, and the air deflector angle is determined to generate a collaborative control instruction set.

[0008] Based on the collaborative control instruction set, an adjustment signal of the compressor frequency, fan speed and air guide plate angle is generated and output to the air-cooled air conditioning control unit, and a drive signal of the device to be executed is established. Based on the drive signal of the device to be executed, the compressor, fan and air guide motor are driven to perform actions, the refrigerant circulation status and changes in the working status of the condenser and evaporator are monitored, and the current operating parameters of the air-conditioning equipment are obtained.

[0009] Preferably, the steps of obtaining the multi-dimensional environment and load state vector are:

[0010] The temperature sensor, humidity sensor, and Wi-Fi signal receiver collect real-time values ​​of indoor temperature, outdoor temperature, and humidity, respectively. The active connection times and average signal strength of the device MAC address in the Wi-Fi channel status information are analyzed to generate a basic sensor data set.

[0011] Calculating personnel dynamic indicators based on the basic sensor data set;

[0012] Based on the personnel dynamic indicators, Gaussian kernel density estimation is performed on the equipment connection positions within the continuous time window, and the spatial coordinates of the kernel density peak points are counted to generate the personnel position distribution. Combined with the basic sensor data set, a multidimensional environment and load state vector is constructed.

[0013] Preferably, the steps for obtaining the basic comfort deviation value are:

[0014] Extracting the measured indoor temperature value, the measured outdoor temperature value, and the measured humidity value from the multidimensional environment and load state vector, and reading the user-set temperature value and the upper and lower limits of the user-preset humidity range from the user configuration file, and establishing a five-tuple parameter set including the measured indoor temperature value, the user-set temperature value, the measured humidity value, the lower limit of the humidity range, and the upper limit of the humidity range;

[0015] Based on the indoor temperature measured value and the user-set temperature value in the five-tuple parameter set, the absolute value of the difference between the two is calculated as the original temperature deviation value. At the same time, based on the humidity measured value and the lower limit value and the upper limit value of the humidity range, the absolute value of the measured humidity deviating from the median of the preset humidity range is calculated as the original humidity deviation value;

[0016] The original temperature deviation value is divided by the preset maximum allowable temperature deviation threshold to achieve normalization processing, and the original humidity deviation value is divided by the preset maximum allowable humidity deviation threshold to achieve normalization processing. The normalized temperature deviation value and humidity deviation value are weighted and summed according to a weight ratio of 6:4 to generate a basic comfort deviation value.

[0017] Preferably, the steps of obtaining the multi-objective performance indicator set are:

[0018] Read the current air speed from the air conditioner in real time, obtain the basic comfort deviation value, analyze the number of starts and stops and the total operating time in the last hour from the compressor operation record, collect the current air conditioner voltage and current, and generate the original parameter set;

[0019] Calculating a wind discomfort index based on the original parameter set;

[0020] Based on the number of starts and stops and the total operating time in the last hour, the start-stop frequency index per unit time is calculated. The current power value is calculated based on the voltage and current. The wind discomfort index, start-stop frequency index and current power value are integrated to generate a set of multi-objective performance indicators.

[0021] Preferably, the steps of obtaining the candidate control parameter list are:

[0022] Extracting the measured indoor temperature value, the measured humidity value, and the occupant location distribution density from the multi-dimensional environment and load state vector, extracting the wind discomfort index, the start-stop frequency index, and the current power value from the multi-objective performance index set, and performing one-to-one mapping with the temperature deviation upper limit threshold, the humidity deviation upper limit threshold, the wind speed safety threshold, the start-stop frequency limit, and the power limit in the preset operating rules to generate an initial candidate parameter combination set;

[0023] Based on the initial candidate parameter combination set, all preset operating rule entries are traversed. If a rule requires that the measured temperature deviation value is less than or equal to the temperature deviation upper limit threshold and the start-stop frequency index is less than or equal to the start-stop frequency limit, then the parameter combination that meets all the conditions in the rule is screened. If at least one rule is not met, the parameter combination is eliminated to generate a valid parameter combination subset;

[0024] According to each parameter combination in the valid parameter combination subset, the corresponding compressor frequency preset gear value, fan speed gradient value and air guide plate angle adjustment value are extracted, and they are sorted in ascending order of compressor frequency and descending order of fan speed to generate a candidate control parameter list containing frequency-speed-angle triplets.

[0025] Preferably, the steps of acquiring the collaborative control instruction set are:

[0026] Extracting the compressor frequency setting value, fan speed setting value, and air guide plate angle setting value of each candidate item from the candidate control parameter list, obtaining the frequency stability deviation, speed fluctuation rate, and angle adjustment response time of the corresponding item in historical operation data, and generating a candidate parameter performance feature set;

[0027] Calculating a comprehensive performance score based on the candidate parameter performance feature set;

[0028] All candidate items are sorted in descending order according to their comprehensive performance scores, and the compressor frequency setting value, fan speed setting value, and air guide plate angle setting value corresponding to the highest comprehensive performance score are selected to generate a collaborative control instruction set.

[0029] Preferably, the step of acquiring the drive signal of the device to be executed is:

[0030] Extracting the compressor frequency setting value, the fan speed setting value, and the air deflector angle setting value from the collaborative control instruction set, converting the frequency setting value into a PWM duty cycle control signal of the compressor drive module, mapping the speed setting value into a 0-10V analog voltage output signal of the fan inverter, and converting the air deflector angle setting value into a stepper motor control pulse number to generate an original drive signal set;

[0031] Based on the PWM duty cycle control signal, the 0-10V analog voltage output signal, and the number of stepper motor control pulses in the original drive signal set, respectively checking whether the duty cycle exceeds the maximum frequency limit allowed by the compressor, whether the voltage signal is within the rated operating range of the fan, and whether the number of pulses matches the mechanical limit angle of the air deflector, to generate a compliant drive signal set;

[0032] According to the PWM duty cycle control signal, 0-10V analog voltage output signal and stepper motor control pulse number in the compliant drive signal set, they are encapsulated into a complete message frame including the device address code, function code, data field and CRC check code according to the Modbus-RTU protocol to generate the device drive signal to be executed.

[0033] Preferably, the steps for obtaining the current operating parameters of the air-conditioning equipment are:

[0034] The device drive signal to be executed is sent to the air conditioning control unit via the communication bus, and the compressor PWM duty cycle setting value, the fan analog voltage signal value, and the air guide plate pulse number setting value in the Modbus-RTU protocol message are parsed to respectively drive the compressor inverter to adjust the output frequency, the fan speed control module to adjust the supply voltage, and the air guide stepper motor to execute the target pulse number rotation. The start timestamp and completion status of each execution action are recorded to generate a device drive execution record set;

[0035] Based on the device driver execution record set, real-time pressure and temperature values ​​are collected through the high-pressure side pressure sensor, low-pressure side pressure sensor, evaporator inlet temperature sensor, and condenser outlet temperature sensor of the refrigerant circulation pipeline, and the volume flow rate data of the refrigerant is synchronously read to generate a refrigerant circulation state parameter set;

[0036] Monitor the surface temperature distribution of the condenser heat sink, detect the thickness of frost on the evaporator fins, integrate the refrigerant cycle status parameter set with the condenser and evaporator working status data, and generate the current operating parameters of the air conditioning equipment.

[0037] The present invention provides an air-cooling air-conditioning control system, comprising:

[0038] The environmental data processing module is used to integrate indoor and outdoor temperature and humidity sensor data with Wi-Fi channel status information, and combine it with the location density of people within the time window to generate a multidimensional state vector that reflects environmental dynamics and heat load distribution;

[0039] The control index calculation module is used to calculate the temperature deviation, humidity deviation, and wind speed discomfort index based on the multi-dimensional state vector and the user-set temperature value, and correlate the compressor start and stop frequency with the real-time energy consumption power to construct a multi-objective performance index set that quantifies the balance between comfort and energy efficiency;

[0040] A collaborative optimization module is used to screen control parameter combinations that meet preset rules based on a multi-dimensional state vector and a set of multi-objective performance indicators. It generates a list of candidate control parameters that balance efficiency and stability through priority sorting, and optimizes the collaborative control instructions for the compressor, fan, and air deflector.

[0041] A drive signal generation module is used to convert the compressor frequency, fan speed, and air deflector angle in the candidate parameter list into PWM duty cycle signals, analog voltage signals, and stepper motor pulse instructions to generate a standardized control signal set that can directly drive the air conditioner actuator;

[0042] The equipment execution monitoring module is used to execute the driving signal and monitor the refrigerant circulation pressure, condenser and evaporator temperatures and fin frost thickness parameters in real time, and feedback the equipment operation status data to form a key status monitoring link for closed-loop control.

[0043] Compared with the prior art, the advantages and positive effects of the present invention are:

[0044] In the present invention, the multi-dimensional environment and load state vector integrates indoor and outdoor temperature, humidity, Wi-Fi channel status information, and personnel location distribution and activity status, and dynamically calculates the changing trend of personnel density through a time window, thereby enhancing the comprehensiveness and real-time nature of environmental perception and providing high-dimensional data support for control decisions. A basic comfort deviation index is constructed based on the temperature deviation value and the humidity deviation value, and a multi-objective performance index set is generated by combining the wind speed value, the compressor start-stop frequency, and the energy consumption power. The user comfort demand is quantitatively associated with the equipment operating efficiency to avoid energy efficiency imbalance or comfort degradation caused by single parameter adjustment. The candidate control parameter list is screened by preset operating rules and multi-objective trade-offs, and the coordinated adjustment of the compressor frequency, fan speed, and air guide plate angle is mapped to the equipment drive signal. The real-time monitoring of the refrigerant cycle state and the condenser and evaporator working parameters is fed back to the control instruction generation link to form a closed-loop adjustment mechanism, dynamically correct the execution error, and reduce the temperature control fluctuations caused by environmental mutations or nonlinear characteristics of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] See also Figure 1 The present invention provides a technical solution, a deep learning-based air-cooling air conditioning control method, comprising the following steps:

[0048] Obtain sensor readings of indoor temperature, outdoor temperature, and humidity, as well as Wi-Fi channel status information, to determine the number of people and their activity status. Combined with the time window, the distribution of people's locations is estimated, and the multi-dimensional environment and load state vectors are integrated.

[0049] Based on the multi-dimensional environment and load state vector and the user-set temperature, the temperature deviation value and humidity deviation value are calculated to obtain the basic comfort deviation value. Based on the basic comfort deviation value, the wind discomfort index is calculated in combination with the current wind speed value, the start-stop frequency index is calculated by correlating it with the compressor operation record, and the current power value is calculated by correlating it with the energy consumption data to obtain a set of multi-objective performance indicators;

[0050] Based on the multi-dimensional environment and load state vectors and a set of multi-objective performance indicators, the system compares the preset operating rules to screen potential control action combinations that meet the conditions, obtains a list of candidate control parameters, and then determines a set of instructions that specify the compressor frequency, fan speed, and air guide vane angle based on the candidate control parameter list to generate a collaborative control instruction set.

[0051] Based on the collaborative control instruction set, the control unit of the air-cooled air conditioner generates adjustment signals for the compressor frequency, fan speed and air guide plate angle, establishes the drive signal of the device to be executed, and drives the compressor, fan and air guide motor to execute the action based on the drive signal of the device to be executed. The refrigerant circulation status and the changes in the working status of the condenser and evaporator are monitored to obtain the current operating parameters of the air-conditioning equipment.

[0052] The steps to obtain the multi-dimensional environment and load state vector are:

[0053] The temperature sensor, humidity sensor, and Wi-Fi signal receiver collect real-time values ​​of indoor temperature, outdoor temperature, and humidity, respectively. The active connection times and average signal strength of the device MAC address in the Wi-Fi channel status information are analyzed to generate a basic sensor data set.

[0054] Based on the basic sensor data set, the personnel dynamic index is calculated using the following formula:

[0055] ;

[0056] in, is a personnel dynamic indicator, For the The standard deviation of Wi-Fi signal strength within a time window, is the preset time window length, is the absolute value of the number of MAC address changes in adjacent time windows, The number of active device MAC addresses detected in the Wi-Fi channel;

[0057] Based on the personnel dynamic indicators, Gaussian kernel density estimation is performed on the equipment connection locations within the continuous time window. The spatial coordinates of the kernel density peak points are counted to generate the personnel location distribution. Combined with the basic sensor data set, a multidimensional environment and load state vector is constructed.

[0058] Specifically, by deploying the DHT22 digital temperature and humidity sensor module indoors and the SHT31 high-precision temperature and humidity sensor module outdoors, the real-time temperature readings (Celsius) and relative humidity readings (percentage) of the indoor environment, as well as the real-time temperature readings (Celsius) and relative humidity readings (percentage) of the outdoor environment, are continuously collected at a sampling frequency of once every 5 seconds. At the same time, the ESP32 wireless module configured in monitoring mode is used as a Wi-Fi signal receiver to scan the surrounding 2.4GHz Wi-Fi channels, capture ProbeRequest frames and data frames associated with other access points (APs), parse the captured frames, and extract the channel state information (CSI). In particular, for each detected device MAC address, its original equipment manufacturer (OUI) information, received signal strength indicator (RSSI) value, and The system then calculates the average RSSI value for each independent MAC address over six consecutive time windows (constituting a 30-second observation period), and filters out signals below -85dBm as invalid connections or too far away. The collected indoor temperature, outdoor temperature, and humidity values, along with the identifier of each valid MAC address, the total number of active connections within the observation period, the average RSSI value within the observation period, and the timestamp when the data was collected, are integrated and stored as structured data records, for example, forming a data structure containing the fields {timestamp, indoor temperature, outdoor temperature, humidity, MAC address list [{MAC, number of active connections, average RSSI}, ...]}, to generate a basic sensor dataset.

[0059] formula: , the formula is useful in that it incorporates the fluctuation of Wi-Fi signal strength ( ) and changes in the number of devices in the environment ( ) is used to quantify the activity level of indoor personnel. Compared with using only the number of devices or average signal strength, this formula can better capture subtle environmental dynamic changes caused by personnel movement or changes in device usage status, thereby more accurately reflecting the actual indoor load disturbance.

[0060] The steps to obtain the parameter are as follows: The first detected The standard deviation of the Wi-Fi signal strength of an active device (uniquely identified by its MAC address) quantifies the degree of fluctuation in the signal strength of the device. Its value is calculated by processing the sequence of RSSI values ​​recorded in the basic sensor data set. Specifically, for a specified time window (For example, 60 seconds). First, all RSSI measurement values ​​that belong to the time window and correspond to the MAC address of a specific active device are filtered out from the basic sensor data set to form an RSSI value list. For example, for the device with MAC address AA:BB:CC:DD:EE:FF, the RSSI value sequence collected in the past 60 seconds is [-55dBm, -58dBm, -56dBm, -57dBm, -55dBm, -59dBm]. Then, the mean and standard deviation of the sequence are calculated. is the number of RSSI measurements of the device within the time window. Calculation example: For the sequence [-55, -58, -56, -57, -55, -59], the average dBm, standard deviation dBm, therefore, the standard deviation of the signal strength of this device in this time window It is 1.49dBm.

[0061] The parameter acquisition step is as follows: the parameter represents the preset time window length. For example, by analyzing the data of 24 consecutive hours in an office environment, it is found that the calculated time window is 60 seconds. The index is highly consistent with the actual personnel entry and exit and activity status, and can effectively avoid the pseudo peak caused by short-term signal interference. Therefore, the seconds as the preset time window length.

[0062] The parameter is obtained as follows: the parameter represents the absolute value of the change in the number of active device MAC addresses detected in two consecutive time windows. It reflects the dynamic changes in the entry and exit of people holding Wi-Fi devices in the environment or the switch status of devices. For example, in the current 60-second time window The system detected 5 active MAC addresses, namely , and in the previous 60-second time window The system detected 3 active MAC addresses, namely , then the absolute value of the number of MAC address changes in adjacent time windows is ,This value indicates that the net change in the number of active devices in the environment is 2 during the past time window period.

[0063] The steps for obtaining the parameter are as follows: This parameter represents the total number of active device MAC addresses detected in the current Wi-Fi channel and the current time window. It is the upper limit of the summation term in the formula for calculating the dynamic index of personnel, and also indirectly reflects the device density or potential number of personnel in the current environment. It is directly obtained from the basic sensor data set in the current time window. Statistics within, for example, in the current 60-second time window, a total of 8 signals from different MAC addresses are received, including MAC_1 (RSSI = -60, number of connections = 10), MAC_2 (RSSI = -70, number of connections = 5), MAC_3 (RSSI = -88, number of connections = 6), MAC_4 (RSSI = -65, number of connections = 8), MAC_5 (RSSI = -75, number of connections = 4), MAC_6 (RSSI = -90, number of connections = 2), MAC_7 (RSSI = -72, number of connections = 12), MAC_8 (RSSI = -80, number of connections = 1). According to the active condition (RSSI > -85dBm and number of connections > 3), MAC_3, MAC_6, and MAC_8 do not meet the conditions. Therefore, the active devices are MAC_1, MAC_2, MAC_4, MAC_5, and MAC_7, a total of 5. .

[0064] Substituting the formula into the calculation, the result is 0. This result shows that the personnel dynamic index calculated based on the fluctuation of Wi-Fi signals and the change in the number of devices detected in the current time window is 0, which means that the human activity level in the current environment is low, or although there is equipment ( ), but its signal strength is relatively stable ( The value is small), and the entry and exit of personnel or equipment does not change much ( ), it is comprehensively judged to be in a low dynamic state. The value of this indicator will serve as one of the important bases for judging indoor load changes in subsequent steps. A value such as 0 or 1 means that the air conditioning system can be maintained at a lower operating load or energy-saving mode, while a higher A value greater than or equal to 3 indicates that cooling or heating output needs to be increased to cope with the potential load increase.

[0065] Based on the personnel dynamic indicators calculated in the previous step As well as the basic sensor data set, we first extract the MAC addresses of all active devices and their corresponding average RSSI values ​​within multiple consecutive time windows (for example, the last five time windows, each window length is 60 seconds) from the basic sensor data set, convert the average RSSI value of each MAC address in each time window into an approximate distance estimate between it and the Wi-Fi signal receiver, and combine it with the fixed position coordinates of the receiver (for example, (0, 0)) to infer the possible position coordinates of each device in each time window. Next, we apply the Gaussian kernel density estimation (GKDE) method to the set of all inferred device position coordinate points in these five consecutive time windows, and select a two-dimensional Gaussian function as the kernel function with a bandwidth of The selection is based on personnel dynamic indicators The specific setting method is to establish a piecewise function relationship between bandwidth and dynamic indicators. For example, when When the activity is small and the location is relatively stable, a smaller bandwidth is selected. (e.g. 0.5 m) to obtain finer position resolution, the calculation formula is: ,in is the basic bandwidth (for example, 0.8 meters), through or ,when When the activity is moderate, choose medium bandwidth (e.g. 1.0 m), the calculation formula is ,Right now meters, when When the activity is intense, the position changes quickly and the uncertainty is high, a larger bandwidth is selected (e.g. 1.5 meters) to smooth out the noise and capture the main gathering areas, calculated as ,For example hour After GKDE calculation, a kernel density estimation map covering the entire indoor space is obtained. Then, local peak points with density values ​​higher than the preset density threshold are found on this map. The spatial coordinates (x, y) of these peak points and their corresponding peak density values ​​are calculated, and the number of peak points is recorded as the core parameter of the personnel location distribution. For example, if three peak points are found with coordinates of (2.1, 3.5), (4.8, 2.0), and (2.5, 3.8), it is considered that there may be three main personnel gathering areas in the room. Finally, the number and coordinate information of these peak points (personnel location distribution) are combined with the measured indoor temperature, outdoor temperature, and humidity values ​​corresponding to the current timestamp in the basic sensor dataset to form a structured vector, such as [indoor temperature, outdoor temperature, humidity, number of peak points, peak point 1 coordinate x, peak point 1 coordinate y, ..., peak point N coordinate y, personnel dynamic index Dp], to construct a multidimensional environment and load state vector.

[0066] The steps to obtain the basic comfort deviation value are:

[0067] Extracting the measured indoor temperature value, the measured outdoor temperature value, and the measured humidity value from the multidimensional environment and load state vector, and reading the user-set temperature value and the upper and lower limits of the user-preset humidity range from the user configuration file, and establishing a five-tuple parameter set including the measured indoor temperature value, the user-set temperature value, the measured humidity value, the lower limit of the humidity range, and the upper limit of the humidity range;

[0068] Based on the measured indoor temperature value and the user-set temperature value in the five-tuple parameter set, the absolute value of the difference between the two is calculated as the original temperature deviation value. At the same time, based on the measured humidity value and the lower limit value and upper limit value of the humidity range, the absolute value of the measured humidity deviation from the median of the preset humidity range is calculated as the original humidity deviation value.

[0069] The original temperature deviation value is divided by the preset maximum allowable temperature deviation threshold to achieve normalization processing, and the original humidity deviation value is divided by the preset maximum allowable humidity deviation threshold to achieve normalization processing. The normalized temperature deviation value and humidity deviation value are weighted and summed according to the weight ratio of 6:4 to generate the basic comfort deviation value.

[0070] Specifically, from the multi-dimensional environment and load state vector constructed in the previous step, locate and extract the current indoor temperature measured value, such as reading the value 26.5 degrees Celsius from the specified index position 0 of the vector, and at the same time extract the outdoor temperature measured value, such as reading the value 31.0 degrees Celsius from the index position 1, and the humidity measured value, such as reading the value 65% relative humidity from the index position 2. Then, access the user profile stored in the system, which is usually a JSON or XML format document associated with the currently activated user account, and search and read the user's personalized air-conditioning control parameters, including the user-set temperature value, such as reading the user's preferred target room temperature of 24.0 degrees Celsius and the user's preset acceptable humidity. Range, the range includes a lower limit value and an upper limit value. For example, the user-set comfortable humidity range is read as 50% to 60% relative humidity, that is, the user-set humidity range lower limit value is 50%, and the user-set humidity range upper limit value is 60%. The five extracted and read key parameter values, namely the indoor temperature measured value (26.5), the outdoor temperature measured value (31.0), the humidity measured value (65), the user-set temperature value (24.0), the user-set humidity range lower limit value (50) and the user-set humidity range upper limit value (60), are organized in a predefined order to establish a five-tuple parameter set including the indoor temperature measured value, the user-set temperature value, the humidity measured value, the humidity range lower limit value and the humidity range upper limit value.

[0071] Based on the five-tuple parameter set established in the previous step, which includes the measured indoor temperature value (26.5), the user-set temperature value (24.0), the measured humidity value (65), the lower limit of the humidity range (50), and the upper limit of the humidity range (60), the deviation calculation is performed. First, for the temperature dimension, the absolute value of the difference between the measured indoor temperature value and the user-set temperature value is calculated, and this absolute difference is used as the original temperature deviation value. The specific calculation is: Celsius, which directly reflects the degree to which the current room temperature deviates from the user's desired target. Then, for the humidity dimension, first calculate the median of the user's preset humidity range, using the lower limit value (50) and upper limit value (60) of the humidity range in the quintuple, and calculate the median value to be %, then calculate the absolute value of the difference between the measured humidity value (65) and the median of this humidity range (55), and use this absolute difference as the original humidity deviation value. The specific calculation is: %, which represents the degree to which the current ambient humidity deviates from the center of the user-set comfortable humidity range. Combining these two steps, we obtain two basic measurement values ​​representing the deviation of the current temperature and humidity from the user-set comfortable state, namely the original temperature deviation value of 2.5 degrees Celsius and the original humidity deviation value of 10%.

[0072] The original temperature deviation value (2.5 degrees Celsius) and the original humidity deviation value (10%) calculated in the previous step are normalized respectively. The threshold value based on which the normalization is performed is the preset maximum allowable temperature deviation threshold value ( ) and the preset maximum allowable humidity deviation threshold ( ), the setting of these two thresholds refers to the widely accepted indoor thermal comfort standards, such as ASHRAE Standard 55, and is combined with the statistical analysis of a large number of users' comfort feedback data in different environments to set The process is as follows: Analysis shows that when the indoor temperature deviates from the set temperature by more than 3 degrees Celsius, more than 80% of users will report obvious discomfort. Set to 3.0 degrees Celsius, set The process is as follows: Analysis shows that when the relative humidity deviates from the median of the comfortable range (usually between 40% and 60%) by more than 15 percentage points, users generally feel too dry or too humid. Set to 15.0%, the normalization calculation uses the formula ,in is the original deviation value, is the corresponding maximum allowable deviation threshold. Use the min function to ensure that the normalized result does not exceed 1.0. Calculate the normalized temperature deviation value: , calculate the normalized humidity deviation value: ,Then, the normalized temperature deviation value and humidity deviation value are weighted and summed according to the preset weight ratio, and the weight is set to 60% for temperature ( ), humidity accounts for 40% ( ), the basis for this weight ratio is that multiple comfort studies have shown that in a typical indoor air-conditioned environment, the impact of temperature changes on the human body's immediate thermal sensation is usually greater than that of humidity changes. Through multivariate linear regression analysis of historical operating data and user comfort scores, it was determined that the ratio of the regression coefficient of the temperature deviation term to the regression coefficient of the humidity deviation term is close to 6:4, so this weight ratio is set. and , calculate the basic comfort deviation value , substitute the values: , round the calculation result to generate the basic comfort deviation value.

[0073] The steps to obtain the multi-objective performance indicator set are:

[0074] Read the current wind speed from the air conditioner in real time , get the basic comfort deviation value , analyze the number of starts and stops in the last hour in the compressor operation record Total runtime , collect the current air conditioner voltage With current , generate the original parameter set;

[0075] Based on the original parameter set, the wind discomfort index is calculated using the following formula:

[0076] ;

[0077] in, It is an index of wind discomfort. is the basic comfort deviation value, is the current wind speed, For the The average wind speed in the time window, is the interval between the current moment and the last wind speed change;

[0078] Based on the number of starts and stops in the last hour Total runtime , calculate the start-stop frequency index per unit time ,in Prevent division by zero, voltage-based With current Calculate the current power value ,Integrate the blowing discomfort index, start-stop frequency index and current power value to generate a set of multi-objective performance indicators.

[0079] Specifically, the wind speed sensor built into the air conditioner indoor unit, such as a hot wire anemometer or a Hall effect sensor, reads the current wind speed at the air outlet once per second to obtain the real-time wind speed value. For example, the current wind speed is read as 2.5 meters per second. At the same time, the current basic comfort deviation value generated from the previous calculation step is obtained. For example, if the value obtained is 0.767, then access the compressor operation log stored in the air conditioning control unit memory or associated database. This log records the timestamps of the compressor start and stop events. Parse these records and filter out all start and stop events in the last hour (3600 seconds back from the current time). Count the total number of start events to get the start and stop count. For example, if the compressor is started 3 times in the last hour, the total running time of the compressor is calculated (accumulating the time interval from each start to stop) to get the total running time. For example, the total running time is calculated to be 1800 seconds (i.e. 0.5 hours). In addition, the voltage sensor and current transformer connected to the air conditioner power input terminal are used to collect the current AC voltage RMS of the air conditioner main circuit in real time. and the effective value of current For example, if the voltage is 230 volts and the current is 4.5 amps, the current wind speed will be obtained in real time. (2.5), the previously calculated basic comfort deviation value (0.767), the number of starts and stops in the last hour (3) Total running time in the last hour (0.5 hours), current voltage (230), and the current (4.5) These values ​​are aggregated to generate a structured data record containing these specific parameters, namely the original parameter set.

[0080] formula: The usefulness of the formula is that it takes into account multiple key factors that cause the discomfort of blowing wind, including not only wind speed ( ) and the discomfort level of the current thermal and humid environment (basic comfort deviation value ) These two direct influencing factors (through This shows that the squared wind speed term amplifies the effect of high-speed winds).

[0081] The parameter acquisition step is as follows: the parameter represents the current real-time wind speed at the air outlet of the air conditioner indoor unit. The value is directly measured by the wind speed sensor installed in the air duct or air outlet of the air conditioner. For example, at the current moment, the control unit reads the signal from the wind speed sensor and obtains the current real-time wind speed value after processing. m / s.

[0082] The step of obtaining the parameter is as follows: the parameter represents the basic comfort deviation value. For example, according to the above calculation process, the basic comfort deviation value obtained is .

[0083] The steps to obtain the parameter are as follows: The average wind speed in a time window. For example, if the current time is T and the time window is 60 seconds, then: for The average wind speed in the interval is calculated m / s; for The average wind speed in the interval is calculated m / s; for The average wind speed in the interval is calculated m / s. In addition, the formula also requires ,when When you need value, represent The previous time window (the fourth time window from the last) ) average wind speed, for example, we can calculate m / s.

[0084] The steps for obtaining the parameter are as follows: the parameter represents the time interval between the current moment and the last time the wind speed changed significantly, and is used to measure the duration of the current wind speed stability. Its acquisition requires the system to record the time point of each adjustment of the fan speed (which in turn affects the wind speed). For example, the current time is 10:30:00, and the system record shows that the last wind speed gear adjustment occurred at 10:28:00, then Second.

[0085] Substitute the parameters into the formula to calculate the wind discomfort index The result is about -0.930. The result shows that the wind discomfort index calculated based on the current indoor environment deviation, wind speed, wind speed stabilization time and recent wind speed fluctuations is is about -0.930, The larger the value (or the closer it is to a positive value), the stronger the discomfort caused by the wind. A negative value may mean that under the current conditions, the discomfort caused by the wind is relatively low. Here, -0.930 indicates that the current discomfort caused by the wind is relatively low.

[0086] Formula 1: , Formula 2: , formula 1 Used to calculate the start-stop frequency index per unit time, it quantifies the average number of times the compressor starts per unit operating time and reflects the frequency of the compressor's operation. This is meaningful for evaluating equipment loss and system stability. Frequent start-stops will increase equipment wear and may affect cooling efficiency and temperature stability. Formula 2 Used to calculate the current instantaneous power value of the air conditioner, which directly reflects the current energy consumption level of the equipment and is a key parameter for evaluating the system energy efficiency and operating costs. Total running time in the last hour Hourly zero prevention constant Hourly current voltage V Current A

[0087] Calculate the start-stop frequency index per unit time :

[0088] ;

[0089] ;

[0090] ;

[0091] ;

[0092] (times / hour running time);

[0093] Calculate the current power value :

[0094] ;

[0095] ;

[0096] (Watt, W);

[0097] Integration index: The calculated wind discomfort index , start-stop frequency index per unit time , and the current power value W, combined into a vector or set, forming a multi-objective performance indicator set. Multi-objective performance indicator set = [ , , ]=[-0.930, 5.393, 1035].

[0098] The steps for obtaining the candidate control parameter list are:

[0099] The measured indoor temperature and humidity values, as well as the occupant distribution density, are extracted from the multi-dimensional environment and load state vector. The wind discomfort index, start-stop frequency index, and current power value are extracted from the multi-objective performance index set. These values ​​are then mapped one-to-one with the temperature deviation upper threshold, humidity deviation upper threshold, wind speed safety threshold, start-stop frequency limit, and power limit in the preset operating rules to generate an initial candidate parameter combination set.

[0100] Based on the initial set of candidate parameter combinations, all preset operating rule entries are traversed. If a rule requires that the measured temperature deviation value is less than or equal to the temperature deviation upper threshold and the start-stop frequency index is less than or equal to the start-stop frequency limit, the parameter combinations that meet all the conditions in the rule are screened. If at least one rule is not met, the parameter combination is eliminated to generate a valid parameter combination subset;

[0101] According to each parameter combination in the valid parameter combination subset, the corresponding compressor frequency preset gear value, fan speed gradient value and air guide plate angle adjustment value are extracted, and the parameters are sorted in ascending order of compressor frequency and descending order of fan speed to generate a candidate control parameter list containing frequency-speed-angle triplets.

[0102] Specifically, from the multi-dimensional environment and load state vector obtained in the previous step, the current indoor temperature measured value (e.g., 26.5 degrees Celsius), humidity measured value (e.g., 65%), and the personnel location distribution density obtained by Gaussian kernel density estimation (e.g., the number of peak points is 3, or a scalar value representing the overall density is 0.7) are extracted. At the same time, the wind discomfort index (-0.930), start-stop frequency index (5.393 times / hour of operation), and current power value (1035 watts) are extracted from the multi-objective performance index set [-0.930, 5.393, 1035] generated in the previous stage. Next, the visit The preset operation rule base stores a series of operation constraints, which are expressed as specific thresholds or limits, including the upper temperature deviation threshold (for example, 3.0 degrees Celsius, based on the recommendations on acceptable temperature deviation in ASHRAE Standard 55, calculated by analyzing the temperature range corresponding to 80% acceptability recommended in the standard), the upper humidity deviation threshold (for example, 15%, based on the fact that exceeding the center value of the typical comfortable humidity range by 15% usually causes user discomfort, and is set through statistical analysis of user feedback survey data), the wind speed safety threshold (for example, the maximum allowable wind speed at the indoor unit outlet is 5.0 meters per second, based on the According to the product design specifications and the research settings to avoid excessive noise or blowing feeling, refer to ISO7730's guidance on indoor air velocity limits, and conduct actual test verification), start and stop frequency limit (for example, 10.0 times / hour, based on the service life recommendations and reliability data analysis provided by the compressor manufacturer, to avoid too frequent start and stop shortening the equipment life), and power limit (for example, 1500 watts, set according to the rated power of the air conditioner, the maximum load capacity of the connected circuit or a specific energy-saving target), the currently extracted various states and performance index values ​​(temperature deviation needs to be calculated based on the measured value and the set value, for example, 2.5 degrees Celsius; humidity ... The difference also needs to be calculated, for example, 10%) and associated with the corresponding limit value in the rule base to preliminarily evaluate whether the current system state is within a safe and reasonable operating range. These state values ​​and performance index values ​​are combined with a set of predefined basic control action sets covering the main adjustable range of the air conditioner (for example, all combinations of compressor frequency range [30Hz, 40Hz, 50Hz], fan speed gear [1, 2, 3, 4, 5], and air guide plate angle [0 degrees, 30 degrees, 45 degrees, 60 degrees, 90 degrees]) to generate an initial candidate parameter combination set. Each element in the set contains the current state indicator and a set of potential control parameters.

[0103] Based on the initial candidate parameter combination set generated in the previous step, which includes the current environmental status, performance indicators, and all potential control action combinations, the system begins to go through each rule entry stored in the preset operation rule library one by one. These rule entries define the constraints that the air conditioner must comply with during operation. For example, Rule 1: "At any time, the start-stop frequency index ( ) shall not exceed the start-stop frequency limit (10.0 times / hour running time)", Rule 2: "At any time, the current power value ( ) shall not exceed the power limit (1500 watts)", Rule 3: "When the absolute value of the indoor temperature deviation exceeds the temperature deviation upper limit threshold (3.0 degrees Celsius), the lowest wind speed gear (gear 1) is not allowed to be selected", Rule 4: "When the density of personnel distribution in the area directly below the air outlet (for example, based on coordinate judgment, the (x, y) coordinates are in the range of [1.0-2.0, 2.0-3.0]) is higher than 0.8, the wind speed shall not exceed half of the wind speed safety threshold (2.5 meters per second)", For each parameter combination in the initial candidate parameter combination set (representing a potential control action), the system will adjust the control action based on the current actual status and performance indicators (such as , , temperature deviation = 2.5 degrees Celsius, humidity deviation = 10%, for example, the density of people does not trigger the conditions of Rule 4, the wind speed in the potential action is 3.0 meters per second), check against all rule entries in the rule base, check Rule 1: Current , meet, check rule 2: current , meet, check rule 3: current temperature deviation , the rule premise is not met, so this rule is not applied for restriction. Check Rule 4: The premise is not met, so it is not applied. If the current state corresponding to a parameter combination or the action implied by the combination itself (for example, Rule 3 restricts the selection of the lowest wind speed gear action) violates the requirements of any rule (that is, the rule condition judgment result is false), then the parameter combination is removed from the set. After traversing and screening all initial parameter combinations and all rule entries, the retained parameter combinations constitute the valid parameter combination subset.

[0104] According to the valid parameter combination subset obtained by screening in the previous step, each retained parameter combination represents a potential feasible control option that meets all the preset operating rule constraints in the current state. Now it is necessary to clearly extract the specific control instruction parameters from these valid parameter combinations. For each parameter combination in the subset, the system parses the control action information it contains and extracts the corresponding compressor frequency preset gear value (for example, determine a specific frequency such as 40Hz from discrete gears or continuous values ​​such as 30Hz, 40Hz, 50Hz), fan speed gradient value (for example, determine a gear value from gears 1, 2, 3, 4, 5, etc. , such as gear 4), and the wind deflector angle adjustment value (for example, determining a target angle, such as 60 degrees, from preset angles such as 0 degrees, 30 degrees, 45 degrees, 60 degrees, and 90 degrees), thereby converting each valid parameter combination into a specific "frequency-speed-angle" triplet control instruction. For example, the valid subset may include multiple triplets such as [40Hz, gear 5, 30 degrees], [30Hz, gear 4, 60 degrees], [40Hz, gear 3, 45 degrees], and [30Hz, gear 5, 0 degrees]. Next, these extracted triplets are sorted. The purpose of sorting is to have a priority parameter when selecting the optimal instruction later. For the purpose of this study, the sorting rule is set as follows: first, sort in ascending order according to the compressor frequency (first element) (low frequency is preferred for energy saving). If the compressor frequencies are the same, sort in descending order according to the fan speed (second element) (high wind speed is preferred at the same frequency to achieve the effect quickly, or this secondary sorting rule can be adjusted according to other optimization goals). The deflector angle (third element) is not used as the main sorting basis, but is kept in the triplet. This sorting rule is applied to sort the above example triplet list: first compare the frequency, 30Hz is ranked before 40Hz, and the groups {[30Hz, gear 4, 60 degrees], [30Hz, gear 5, 0 degrees]} and {[40 Hz, gear 5, 30 degrees], [40 Hz, gear 3, 45 degrees]}; then sort each frequency group in descending order by wind speed, the first group becomes [30 Hz, gear 5, 0 degrees], [30 Hz, gear 4, 60 degrees], the second group becomes [40 Hz, gear 5, 30 degrees], [40 Hz, gear 3, 45 degrees], and finally merge to obtain the sorted list: [[30 Hz, gear 5, 0 degrees], [30 Hz, gear 4, 60 degrees], [40 Hz, gear 5, 30 degrees], [40 Hz, gear 3, 45 degrees]], generating a candidate control parameter list containing frequency-speed-angle triplets.

[0105] The steps to obtain the collaborative control instruction set are:

[0106] Extract the compressor frequency setting value, fan speed setting value, and air guide vane angle setting value of each candidate item from the candidate control parameter list, obtain the frequency stability deviation, speed fluctuation rate, and angle adjustment response time of the corresponding item in the historical operation data, and generate a candidate parameter performance feature set;

[0107] Based on the candidate parameter performance feature set, the comprehensive performance score is calculated using the following formula:

[0108] ;

[0109] in, For comprehensive performance rating, To prevent division by zero constant, is the compressor frequency stability deviation, is the fan speed fluctuation rate, Adjust the response time for the deflector angle;

[0110] All candidate items are sorted in descending order according to their comprehensive performance scores, and the compressor frequency setting value, fan speed setting value, and air guide plate angle setting value corresponding to the highest comprehensive performance score are selected to generate a collaborative control instruction set.

[0111] Specifically, from the candidate control parameter list [[30Hz, gear 5, 0 degrees], [30Hz, gear 4, 60 degrees], [40Hz, gear 5, 30 degrees], [40Hz, gear 3, 45 degrees]] generated in the previous step, each candidate item (i.e., frequency-speed-angle triplet) is processed in turn. For example, the candidate item [30Hz, gear 5, 0 degrees] is first processed to extract the compressor frequency setting value (30Hz), the fan speed setting value (gear 5), and the air guide angle setting value (0 degrees). Then, the system accesses the stored historical operation database, which records the past period of time (for example, the most recent 7 days) during the operation of the air conditioner, the time when the control command was issued, the command content (frequency, speed, angle setting value), and the corresponding actual sensor readings (actual compressor frequency, actual fan speed, air deflector encoder angle reading, command completion timestamp, etc.). The system queries the database and retrieves all historical operation segments that completely match the setting value of the current candidate [30Hz, gear 5, 0 degrees] or are within a predetermined tolerance range (for example, the frequency error is less than ±1Hz, the speed is at the same gear, and the angle error is less than ±5 degrees). For each matching historical segment retrieved, the performance characteristics corresponding to the segment are calculated: Calculate the frequency stability deviation ( ), specifically by calculating the root mean square error (RMSE) of the difference between the actual compressor frequency reading sequence and the set value (30Hz) within the segment, using the formula: , where N is the number of frequency readings in the segment; calculate the speed fluctuation rate ( ), specifically by calculating the standard deviation of the sequence of actual fan speed readings (converted to RPM) within the segment; calculating the angle adjustment response time ( ), the specific method is to calculate the average time from the issuance of the angle adjustment command to the actual time the wind deflector reaches the target angle (0 degrees), and to use the performance characteristic values ​​of multiple historical segments calculated for a candidate (for example, the performance characteristic values ​​of 5 historical segments calculated for 0.4, 0.6, 0.5, 0.45, 0.55) and average them to get the final performance characteristic value of the candidate (for example, the average Hz, average RPM, average seconds), repeat this process for all candidates in the candidate control parameter list, and compare each candidate with its corresponding calculated performance characteristic ( , , ) to generate a set of candidate parameter performance characteristics.

[0112] formula: The benefit of the formula is that it provides a method to quantitatively evaluate the comprehensive performance of candidate control parameter combinations, which reflects the three key negative indicators of control stability and response speed (frequency stability deviation , speed fluctuation rate , Angle adjustment response time ) is converted into a positive scoring item by using the inverse form , so that parameters with smaller deviations, smaller fluctuations, and shorter response times (i.e., parameters with better performance) can obtain higher scores. The additive combination method (direct addition of each item) realizes a simple multi-objective fusion, integrating the performance of three different dimensions into a single comprehensive performance score. .

[0113] The parameter acquisition step is as follows: the parameter represents the compressor frequency stability deviation. For example, for the candidate [30 Hz, gear 5, 0 degrees], the corresponding frequency stability deviation is obtained from the candidate parameter performance feature set: Hz.

[0114] The parameter acquisition step is as follows: the parameter represents the fan speed fluctuation rate. For example, for the candidate item [30Hz, gear 5, 0 degrees], the corresponding speed fluctuation rate is obtained from the candidate parameter performance feature set. RPM.

[0115] The parameter acquisition step is as follows: the parameter represents the response time of the wind deflector angle adjustment. For example, for the candidate [30Hz, gear 5, 0 degrees], the corresponding angle adjustment response time is obtained from the candidate parameter performance feature set: Second.

[0116] The steps to obtain the parameter are as follows: the parameter is a zero-proof constant, which is a small positive value. .

[0117] Substitute the parameters into the formula to calculate The result is about 3.444, which shows that the candidate [30Hz, gear 5, 0 degrees] has a comprehensive performance score calculated based on its historical performance. Approximately 3.444.

[0118] Based on the comprehensive performance score calculated for each candidate [frequency setting value, speed setting value, angle setting value] in the candidate parameter performance characteristic set in the previous step For example, if the list [[30Hz, gear 5, 0 degrees]: 3.444, [30Hz, gear 4, 60 degrees]: 3.120, [40Hz, gear 5, 30 degrees]: 3.680, [40Hz, gear 3, 45 degrees]: 3.300] is calculated, the system will perform a sorting operation on this set of candidate items and their corresponding scores. The sorting is based on the comprehensive performance score. The values ​​of the parameters are sorted in descending order, that is, the highest-scoring candidates are placed at the front and the lowest-scoring candidates are placed at the back. A standard sorting algorithm (such as quick sort or merge sort) is used to process the set. The sorted list becomes: [[40Hz, gear 5, 30 degrees]: 3.680, [30Hz, gear 5, 0 degrees]: 3.444, [40Hz, gear 3, 45 degrees]: 3.300, [30Hz, gear 4, 60 degrees]: 3.120], completing the sorting. After sorting, the system selects the candidate ranked first in the list, that is, the one with the highest comprehensive performance score, as the control instruction to be finally adopted. In this example, the candidate with a score of 3.680 [40Hz, gear 5, 30 degrees] is selected. The compressor frequency setting value (40Hz), fan speed setting value (gear 5) and air guide plate angle setting value (30 degrees) corresponding to this selected item are extracted, and these three values ​​are combined to generate the final output of this control decision, that is, the collaborative control instruction set.

[0119] The steps to obtain the device driver signal to be executed are:

[0120] Extract the compressor frequency setting value, fan speed setting value, and air deflector angle setting value from the collaborative control instruction set, convert the frequency setting value into a PWM duty cycle control signal for the compressor drive module, map the speed setting value into a 0-10V analog voltage output signal of the fan inverter, and convert the air deflector angle setting value into the number of stepper motor control pulses to generate the original drive signal set;

[0121] Based on the PWM duty cycle control signal, 0-10V analog voltage output signal, and stepper motor control pulse number in the original drive signal set, the system verifies whether the duty cycle exceeds the maximum frequency limit allowed by the compressor, whether the voltage signal is within the rated operating range of the fan, and whether the pulse number matches the mechanical limit angle of the air deflector, thereby generating a compliant drive signal set.

[0122] According to the PWM duty cycle control signal, 0-10V analog voltage output signal and stepper motor control pulse number in the compliant drive signal set, they are encapsulated into a complete message frame including the device address code, function code, data field and CRC check code according to the Modbus-RTU protocol to generate the device drive signal to be executed.

[0123] Specifically, from the collaborative control instruction set [40Hz, gear 5, 30 degrees] generated in the previous stage, the specific control target value, that is, the compressor frequency setting value, is extracted. Hz, fan speed setting value Gear position 5 and wind deflector angle setting value Then, the operation of converting these logical setting values ​​into physical drive signals that can be recognized by the underlying hardware is performed. For the compressor frequency setting value (40Hz), according to the mapping relationship between frequency and PWM (pulse width modulation) duty cycle defined in the compressor inverter specification, for example, it is known that the operating frequency range of the inverter is Hz to Hz, the corresponding PWM duty cycle range is %arrive %, using linear mapping formula Convert and calculate %, generating a PWM duty cycle control signal of 26%; for the fan speed setting value (gear 5), according to the system's preset gear to voltage mapping table, which is established based on the calibration test of the fan and controller characteristics, for example {gear 1: 2V, gear 2: 4V, gear 3: 6V, gear 4: 8V, gear 5: 10V}, the voltage value corresponding to gear 5 is 10.0V, and the generated 0-10V analog voltage output signal is 10.0V; for the air guide angle setting value (30 degrees), based on the parameter calibration of the stepper motor and its drive mechanism, it is known that the air guide plate needs to drive the stepper motor to rotate 1000 pulses in total from 0 degrees to the maximum angle of 120 degrees (mechanical limit range), and a linear proportional relationship is adopted. Convert and calculate The three calculated or mapped specific signal values ​​(PWM duty cycle 26%, analog voltage 10.0V, control pulse number 250) are combined to generate the original drive signal set.

[0124] Based on the original drive signal set generated in the previous step, which includes the PWM duty cycle control signal (26%), the 0-10V analog voltage output signal (10.0V), and the stepper motor control pulse number (250), a series of verification operations are performed. First, the PWM duty cycle control signal (26%) is verified and converted back to the equivalent compressor frequency through the inverse mapping relationship. Then, the calculated frequency is compared with the maximum frequency limit allowed by the compressor read from the device specification library; then, the 0-10V analog voltage output signal (10.0V) is verified and compared with the rated operating voltage range required by the fan driver (for example, [0V, 10V], which is determined according to the fan controller hardware manual). The judgment condition is , here The condition is met; finally, the stepper motor control pulse number (250) is checked and compared with the total pulse number range allowed by the wind deflector mechanical structure (for example, from 0 pulses corresponding to 0 degrees to 1000 pulses corresponding to 120 degrees, this range is determined by the initialization calibration when the equipment is installed). The judgment condition is , here If all check items satisfy their respective constraints, the signal values ​​in the original drive signal set do not need to be modified and are directly used as compliant signals. If any check item does not satisfy the conditions, for example, the calculated number of pulses exceeds the range, the signal value is clamped to its allowed boundary value (for example, if 1100 pulses are calculated, it is corrected to 1000 pulses), and the corrected signal value is output together with other compliant signal values. After these checks and possible corrections, a compliant drive signal set is generated.

[0125] According to the compliance drive signal set verified and confirmed in the previous step, which includes PWM duty cycle control signal (26%), 0-10V analog voltage output signal (10.0V) and stepper motor control pulse number (250), the system encapsulates these signal values ​​into a standard message frame according to the Modbus-RTU protocol standard widely used in industrial fieldbus communication. First, the device address code (SlaveID) of the target device (air conditioning control unit) in the Modbus network is determined. The address code is set during system installation or configuration, for example, it is set to 0x01. Then, the appropriate Modbus function code is selected to perform the write operation. To write multiple control parameters at the same time, select function code 0x10 (WriteMultipleHoldingRegisters). Next, define the mapping address and format of the data in the Modbus register. For example, write the PWM duty cycle (multiplied by 100 to express the integer form of percentage, that is, 2600) to the holding register with the starting address of 40001, write the analog voltage (multiplied by 100 to retain two decimal places in the integer form, that is, 1000) to the register with the address of 40002, and write the number of pulses (250) to the register with the address of 40003. Based on this, construct the data domain: specify the starting register address (4000 1. The Modbus address is represented by 0x9C40), the number of registers to be written (3), the number of data bytes (3 registers * 2 bytes / register = 6 bytes), and the specific data values ​​are arranged in the order of high byte first and low byte last (PWM: 2600 = 0x0A28; voltage: 1000 = 0x03E8; number of pulses: 250 = 0x00FA), forming the data sequence [0x0A, 0x28, 0x03, 0xE8, 0x00, 0xFA]. Then, the CRC checksum of all bytes from the device address code to the end of the data field is calculated using the Modbus standard CRC-16 algorithm ( The polynomial is 0xA001), and a 16-bit CRC value is calculated. Finally, the device address code, function code, start address, number of registers, number of data bytes, data content, and the calculated CRC checksum (low byte first, high byte last) are combined in sequence to form a complete Modbus-RTU message frame, for example: [0x01, 0x10, 0x9C, 0x40, 0x00, 0x03, 0x06, 0x0A, 0x28, 0x03, 0xE8, 0x00, 0xFA, CRC_Lo, CRC_Hi] (the CRC value needs to be calculated based on the actual bytes), generating the device drive signal to be executed.

[0126] The steps to obtain the current operating parameters of the air conditioning equipment are as follows:

[0127] The device drive signal to be executed is sent to the air conditioning control unit via the communication bus. The Modbus-RTU protocol message is parsed to determine the compressor PWM duty cycle setting value, the fan analog voltage signal value, and the air deflector pulse count setting value. The system then drives the compressor inverter to adjust the output frequency, the fan speed control module to adjust the supply voltage, and the air deflector stepper motor to rotate to the target pulse count. The system also records the start timestamp and completion status of each execution action to generate a device drive execution record set.

[0128] Based on the device driver execution record set, the high-pressure side pressure sensor, low-pressure side pressure sensor, evaporator inlet temperature sensor and condenser outlet temperature sensor of the refrigerant circulation pipeline are used to collect real-time pressure and temperature values, and the refrigerant volume flow rate data is synchronously read to generate a refrigerant circulation state parameter set;

[0129] Monitor the surface temperature distribution of the condenser heat sink, detect the thickness of frost on the evaporator fins, integrate the refrigerant cycle status parameter set with the condenser and evaporator working status data, and generate the current operating parameters of the air conditioning equipment.

[0130] Specifically, through the specified communication bus, usually the RS-485 serial bus, the communication parameters are configured such as the baud rate is 9600 bits per second, 8 data bits, no parity bit, and 1 stop bit, and the device drive signal to be executed generated in the previous step, that is, the encapsulated Modbus-RTU message frame [0x01, 0x10, 0x9C, 0x40, 0x00, 0x03, 0x06, 0x0A, 0x28, 0x03, 0xE8, 0x00, 0xFA, CRC_Lo, CRC_Hi], is sent by the main controller to the bus with the address code 0x0 1, after receiving the message, the air conditioning control unit first performs a CRC check to verify the data integrity. After the check passes, the message content is parsed to identify the function code 0x10 (write multiple holding registers) and the target register address and data value in the data field, that is, the value of address 40001 is 0x0A28 (corresponding to PWM duty cycle 2600, i.e. 26%), the value of address 40002 is 0x03E8 (corresponding to voltage 1000, i.e. 10.0V), and the value of address 40003 is 0x00FA (corresponding to pulse number 250). The control unit then reads the message based on the parsed The set value executes the corresponding drive operation and outputs a PWM signal with a duty cycle of 26% to the compressor inverter. The inverter adjusts the AC frequency of the compressor motor to about 40Hz. At the same time, a 10.0V analog voltage signal is generated by the digital-to-analog converter (DAC) and output to the fan speed control module. The module adjusts the voltage supplied to the fan motor to make the fan reach the speed corresponding to gear 5. In addition, 250 control pulse signals are sent to the driver of the air guide stepper motor to drive the stepper motor to accurately rotate the corresponding number of steps, driving the air guide plate to reach the target position of 30 degrees. When issuing an instruction (frequency adjustment, speed adjustment, angle adjustment), the system records the current operation type, target set value, and the start timestamp accurate to milliseconds. It also determines whether the action is successfully completed by monitoring the actuator status feedback signal (for example, the inverter reports that the frequency is stable, and the stepper motor driver reports that the movement is completed) or setting a reasonable timeout time (for example, the angle adjustment timeout is set to 3 seconds). The system records the completion status (for example, success, failure, timeout) and completion timestamp, and aggregates these record entries containing the start time, instruction content, completion status, and end time to generate a device driver execution record set.

[0131] Based on the time reference and action background provided by the device driver execution record set generated in the previous step, the system starts the real-time data collection process for the core parameters of the refrigerant cycle, and obtains data through sensors installed at specific locations in the refrigerant cycle pipeline, including: a high-pressure side pressure sensor (for example, a diffused silicon pressure transmitter with a range of 0-4 MPa) installed on the high-pressure side pipeline (between the compressor outlet and the condenser inlet) for measuring the refrigerant exhaust pressure; a low-pressure side pressure sensor (for example, a ceramic capacitor pressure sensor with a range of 0-2 MPa) installed on the low-pressure side pipeline (between the evaporator outlet and the compressor inlet) for measuring the refrigerant suction pressure; an evaporator inlet temperature sensor (for example, an NTC thermistor or a PT1000 platinum resistance thermometer) installed on the outer wall or inside the refrigerant pipe at the evaporator inlet for measuring the refrigerant temperature entering the evaporator; a condenser outlet temperature sensor (same type as above) installed on the outer wall or inside the refrigerant pipe at the condenser outlet for measuring the refrigerant temperature leaving the condenser These sensors operate continuously at a preset sampling frequency (e.g., once every 2 seconds), converting the physical quantities they sense into electrical signals (such as voltage, current, or resistance). The air conditioning control unit reads these electrical signals through an analog-to-digital converter (ADC) and converts the raw readings into standard engineering units (pressure unit: MPa, temperature unit: Celsius) using sensor calibration curve data stored in non-volatile memory. Simultaneously, the system obtains refrigerant volume flow rate data in the pipeline through a volume flow meter (such as an ultrasonic flow meter or turbine flow meter) installed on the refrigerant liquid or gas pipe, or indirectly estimates it through a built-in refrigerant physical property model based on the current operating frequency of the compressor and suction and exhaust state parameters. The system integrates the high-pressure side pressure, low-pressure side pressure, evaporator inlet temperature, condenser outlet temperature, and refrigerant volume flow rate data collected at the same sampling time point into a timestamped data record. These records are continuously collected and accumulated to generate a set of refrigerant cycle state parameters.

[0132] While collecting the core parameters of the refrigerant cycle, the system monitors the working status of the key heat exchange components of the air conditioner (condenser and evaporator). For the condenser, the temperature distribution information on the condenser surface is obtained by deploying a temperature sensor array on the surface of its heat sink (for example, a 3x3 chip thermistor network is arranged along the airflow direction and vertical direction) or using a non-contact infrared thermal imager (at a lower frequency, for example, once every 5 minutes). If it is a sensor array, the temperature value of each point is read, and the average temperature, the maximum and minimum temperature difference, or the standard deviation of the temperature distribution is calculated; if it is a thermal imager, the acquired thermal image is analyzed to extract the temperature statistical characteristics of the key areas. These data reflect the heat dissipation efficiency of the condenser and whether there is local blockage or uneven heat dissipation. For the evaporator, the focus is on monitoring the frost on its fin surface. For example, the method used is to install temperature sensors at different positions on the evaporator fins (especially the frost-prone areas in the middle and lower parts of the windward and leeward sides). When the temperature measured by the sensor is continuously lower than the preset frost judgment threshold (for example, the low When temperatures exceed -2°C for more than 5 minutes (this threshold is determined experimentally, meaning significant frost is observed under these conditions, impacting heat exchange efficiency), the evaporator is determined to have begun frosting. The severity of the frost is preliminarily assessed by the number of sensors below the threshold or the lowest temperature value. Alternatively, the system measures the pressure differential between the evaporator inlet and outlet. A significant increase in the pressure differential (for example, exceeding 150% of the baseline pressure differential during normal operation; this baseline pressure differential and the increase threshold are set through calibration of operating data under frost-free conditions and frosting experiments) also indicates possible frost formation or filter blockage. The monitored condenser fin surface temperature distribution data (for example, an average temperature of 35°C with a temperature difference of 8°C) and evaporator fin frost status information (for example, light frost) are integrated with a set of refrigerant cycle state parameters (including pressure, temperature, and flow rate) acquired during the same time period. All of this data reflecting the current operating status of the air conditioning system is organized into a structured data record, generating the current operating parameters of the air conditioning equipment, which includes information on refrigeration cycle parameters, heat exchanger status, and other aspects.

[0133] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A deep learning-based air-cooling air conditioning control method, characterized in that: The following steps are involved: Obtain sensor readings of indoor temperature, outdoor temperature, and humidity, as well as Wi-Fi channel status information, to determine the number of people and their activity status. Combined with the time window, the distribution of people's locations is estimated, and the multi-dimensional environment and load state vectors are integrated. Based on the multi-dimensional environment and load state vector and the user-set temperature, a temperature deviation value and a humidity deviation value are calculated to obtain a basic comfort deviation value. Based on the basic comfort deviation value, a wind discomfort index is calculated in combination with the current wind speed value, a start-stop frequency index is calculated by correlating the compressor operation record, and a current power value is calculated by correlating the energy consumption data to obtain a set of multi-objective performance indicators. Based on the multi-dimensional environment and load state vector and the multi-objective performance indicator set, and in accordance with preset operating rules, potential control action combinations that meet the conditions are screened to obtain a list of candidate control parameters. Based on the list of candidate control parameters, a set of instructions specifying the compressor frequency, the fan speed, and the air deflector angle is determined to generate a collaborative control instruction set. Based on the collaborative control instruction set, an adjustment signal for the compressor frequency, fan speed, and air guide plate angle is generated and output to the air-cooled air conditioning control unit, and a drive signal for the device to be executed is established. Based on the drive signal for the device to be executed, the compressor, fan, and air guide motor are driven to perform an action, the refrigerant circulation state and the changes in the working state of the condenser and evaporator are monitored, and the current operating parameters of the air conditioning equipment are obtained; The steps for obtaining the basic comfort deviation value are as follows: Extracting the measured indoor temperature value, the measured outdoor temperature value, and the measured humidity value from the multidimensional environment and load state vector, and reading the user-set temperature value and the upper and lower limits of the user-preset humidity range from the user configuration file, and establishing a five-tuple parameter set including the measured indoor temperature value, the user-set temperature value, the measured humidity value, the lower limit of the humidity range, and the upper limit of the humidity range; Based on the indoor temperature measured value and the user-set temperature value in the five-tuple parameter set, the absolute value of the difference between the two is calculated as the original temperature deviation value. At the same time, based on the humidity measured value and the lower limit value and the upper limit value of the humidity range, the absolute value of the measured humidity deviating from the median of the preset humidity range is calculated as the original humidity deviation value; The original temperature deviation value is divided by the preset maximum allowable temperature deviation threshold value to achieve normalization processing, and the original humidity deviation value is divided by the preset maximum allowable humidity deviation threshold value to achieve normalization processing, and the normalized temperature deviation value and humidity deviation value are weighted and summed according to a weight ratio of 6:4 to generate a basic comfort deviation value; The steps for obtaining the multi-objective performance indicator set are: Read the current air speed from the air conditioner in real time, obtain the basic comfort deviation value, analyze the number of starts and stops and the total operating time in the last hour from the compressor operation record, collect the current air conditioner voltage and current, and generate the original parameter set; Calculating a wind discomfort index based on the original parameter set; Based on the number of starts and stops and the total operating time in the last hour, the start-stop frequency index per unit time is calculated. The current power value is calculated based on the voltage and current. The wind discomfort index, start-stop frequency index and current power value are integrated to generate a multi-objective performance index set. The steps for obtaining the candidate control parameter list are: Extracting the measured indoor temperature value, the measured humidity value, and the occupant location distribution density from the multi-dimensional environment and load state vector, extracting the wind discomfort index, the start-stop frequency index, and the current power value from the multi-objective performance index set, and performing one-to-one mapping with the temperature deviation upper limit threshold, the humidity deviation upper limit threshold, the wind speed safety threshold, the start-stop frequency limit, and the power limit in the preset operating rules to generate an initial candidate parameter combination set; Based on the initial candidate parameter combination set, all preset operating rule entries are traversed. If a rule requires that the measured temperature deviation value is less than or equal to the temperature deviation upper limit threshold and the start-stop frequency index is less than or equal to the start-stop frequency limit, then the parameter combination that meets all the conditions in the rule is screened. If at least one rule is not met, the parameter combination is eliminated to generate a valid parameter combination subset; According to each parameter combination in the valid parameter combination subset, the corresponding compressor frequency preset gear value, fan speed gradient value and air guide plate angle adjustment value are extracted, and they are sorted in ascending order of compressor frequency and descending order of fan speed to generate a candidate control parameter list containing frequency-speed-angle triplets.

2. The air-cooling air conditioning control method based on deep learning according to claim 1 is characterized in that: The steps for obtaining the multi-dimensional environment and load state vector are: The temperature sensor, humidity sensor, and Wi-Fi signal receiver collect real-time values ​​of indoor temperature, outdoor temperature, and humidity, respectively. The active connection times and average signal strength of the device MAC address in the Wi-Fi channel status information are analyzed to generate a basic sensor data set. Calculating personnel dynamic indicators based on the basic sensor data set; Based on the personnel dynamic indicators, Gaussian kernel density estimation is performed on the equipment connection positions within the continuous time window, and the spatial coordinates of the kernel density peak points are counted to generate the personnel position distribution. Combined with the basic sensor data set, a multidimensional environment and load state vector is constructed.

3. The air-cooling air conditioning control method based on deep learning according to claim 1, characterized in that: The steps for obtaining the collaborative control instruction set are: Extracting the compressor frequency setting value, fan speed setting value, and air guide plate angle setting value of each candidate item from the candidate control parameter list, obtaining the frequency stability deviation, speed fluctuation rate, and angle adjustment response time of the corresponding item in historical operation data, and generating a candidate parameter performance feature set; Calculating a comprehensive performance score based on the candidate parameter performance feature set; All candidate items are sorted in descending order according to their comprehensive performance scores, and the compressor frequency setting value, fan speed setting value, and air guide plate angle setting value corresponding to the highest comprehensive performance score are selected to generate a collaborative control instruction set.

4. The air-cooling air conditioning control method based on deep learning according to claim 1, characterized in that: The steps of obtaining the device drive signal to be executed are: Extracting the compressor frequency setting value, the fan speed setting value, and the air deflector angle setting value from the collaborative control instruction set, converting the frequency setting value into a PWM duty cycle control signal of the compressor drive module, mapping the speed setting value into a 0-10V analog voltage output signal of the fan inverter, and converting the air deflector angle setting value into a stepper motor control pulse number to generate an original drive signal set; Based on the PWM duty cycle control signal, the 0-10V analog voltage output signal, and the number of stepper motor control pulses in the original drive signal set, respectively checking whether the duty cycle exceeds the maximum frequency limit allowed by the compressor, whether the voltage signal is within the rated operating range of the fan, and whether the number of pulses matches the mechanical limit angle of the air deflector, to generate a compliant drive signal set; According to the PWM duty cycle control signal, 0-10V analog voltage output signal and stepper motor control pulse number in the compliant drive signal set, they are encapsulated into a complete message frame including the device address code, function code, data field and CRC check code according to the Modbus-RTU protocol to generate the device drive signal to be executed.

5. The air-cooling air conditioning control method based on deep learning according to claim 1, characterized in that: The steps for obtaining the current operating parameters of the air conditioning equipment are as follows: The device drive signal to be executed is sent to the air conditioning control unit via the communication bus, and the compressor PWM duty cycle setting value, the fan analog voltage signal value, and the air guide plate pulse number setting value in the Modbus-RTU protocol message are parsed to respectively drive the compressor inverter to adjust the output frequency, the fan speed control module to adjust the supply voltage, and the air guide stepper motor to execute the target pulse number rotation. The start timestamp and completion status of each execution action are recorded to generate a device drive execution record set; Based on the device driver execution record set, real-time pressure and temperature values ​​are collected through the high-pressure side pressure sensor, low-pressure side pressure sensor, evaporator inlet temperature sensor, and condenser outlet temperature sensor of the refrigerant circulation pipeline, and the volume flow rate data of the refrigerant is synchronously read to generate a refrigerant circulation state parameter set; Monitor the surface temperature distribution of the condenser heat sink, detect the thickness of frost on the evaporator fins, integrate the refrigerant cycle status parameter set with the condenser and evaporator working status data, and generate the current operating parameters of the air conditioning equipment.

6. The air-cooling air-conditioning control system according to any one of claims 1 to 5, characterized in that: include: The environmental data processing module is used to integrate indoor and outdoor temperature and humidity sensor data with Wi-Fi channel status information, and combine it with the location density of people within the time window to generate a multidimensional state vector that reflects environmental dynamics and heat load distribution; The control index calculation module is used to calculate the temperature deviation, humidity deviation, and wind speed discomfort index based on the multi-dimensional state vector and the user-set temperature value, and correlate the compressor start and stop frequency with the real-time energy consumption power to construct a multi-objective performance index set that quantifies the balance between comfort and energy efficiency; A collaborative optimization module is used to screen control parameter combinations that meet preset rules based on a multi-dimensional state vector and a set of multi-objective performance indicators. It generates a list of candidate control parameters that balance efficiency and stability through priority sorting, and optimizes the collaborative control instructions for the compressor, fan, and air deflector. A drive signal generation module is used to convert the compressor frequency, fan speed, and air deflector angle in the candidate parameter list into PWM duty cycle signals, analog voltage signals, and stepper motor pulse instructions to generate a standardized control signal set that can directly drive the air conditioner actuator; The equipment execution monitoring module is used to execute the driving signal and monitor the refrigerant circulation pressure, condenser and evaporator temperatures and fin frost thickness parameters in real time, and feedback the equipment operation status data to form a key status monitoring link for closed-loop control.

Images