An AI-based central air conditioning system energy-saving optimization simulation platform

Abstract

This invention relates to the field of AI-based energy-saving optimization technology for central air conditioning systems, and discloses an AI-based simulation platform for energy-saving optimization of central air conditioning systems. The platform sets a uniform sampling period and reads the rated temperature difference, temperature and heat exchange sensor resolution, and maximum heat exchange from the instruction manual. It initializes the thermal resistance and heat capacity double-exponential smoothing parameters and sets non-zero protection. It collects condenser temperature, evaporator temperature, and heat exchange in real time, checking and eliminating anomalies at limits and lower limits. It calculates the temperature difference and its first-order rate of change, and estimates the instantaneous thermal resistance and heat capacity accordingly. It adaptively generates smoothing coefficients based on the continuous period parameter differences, using double-exponential smoothing for iterative updates. It physically corrects the smoothed parameters and stores historical data to ensure reliable results. Finally, it calculates and corrects the recommended heat exchange based on the latest smoothed thermal resistance, serving as a closed-loop control reference for the refrigerant flow or compressor load in the next cycle.

Description

Technical Field

[0001] This invention relates to the field of AI-based energy-saving optimization technology for central air conditioning systems, specifically to an AI-based simulation platform for energy-saving optimization of central air conditioning systems. Background Technology

[0002] Central air conditioning systems are among the most energy-consuming devices in large public buildings, commercial complexes, and industrial plants, often accounting for over 30% of the building's total energy consumption. To minimize energy consumption while maintaining indoor comfort, academia and industry have proposed various energy-saving optimization methods. Traditional methods mainly include setpoint control based on empirical rules, classic PID (proportional-integral-derivative) control, and optimization strategies based on model predictive control (MPC).

[0003] Empirical rule-based control typically sets refrigerant flow rate or compressor speed to a few fixed levels and roughly adjusts them based on outdoor temperature or indoor load. It requires no online calculation and is simple to implement, but it is prone to problems such as operational lag, energy waste, and decreased comfort. PID control automatically adjusts the actuator input through feedback, enabling it to follow system dynamics to some extent. However, because PID parameters usually require repeated manual adjustments, it is prone to overshoot or oscillation when faced with sensor measurement noise, system time delays, and changes in operating conditions in central air conditioning systems, making it difficult to achieve an ideal balance between energy consumption and comfort. Model predictive control (MPC) establishes a mathematical model of the system to achieve online optimization of system behavior over a future period, possessing good foresight and robustness. However, to fully utilize the advantages of MPC, accurate modeling of the system's thermodynamic characteristics, heat exchange processes, and piping hysteresis is required, along with high-frequency data acquisition and numerical solving, resulting in large computational loads, complex algorithms, and high dependence on model accuracy. In practical engineering, MPC solutions are often limited by computational resources or incomplete models, hindering their widespread application. Furthermore, existing dynamic parameter estimation methods, such as thermal resistance and thermal capacity identification based on recursive least squares and Kalman filtering, require pre-setting multiple hyperparameters, including the covariance matrix and noise intensity. The filter order and initial values ​​significantly impact convergence speed and estimation accuracy. Once operating conditions change abruptly or the measurement noise level changes, existing identification methods often struggle to adaptively adjust, resulting in inaccurate characterization of the system's transient characteristics. At the sensor level, both temperature and heat transfer sensors have resolution and range limitations. Ignoring situations where the denominator might be zero or data jumps during calculations can lead to computational anomalies or distorted estimations. Some existing algorithms simply use amplitude limiting or directly discard abnormal data, lacking a unified denominator protection strategy and data reliability verification process.

[0004] Therefore, this study aims to propose an AI-based energy-saving optimization method and simulation platform for central air conditioning systems. First, the system parameters are automatically initialized using equipment manual parameters and non-zero protection constants to ensure the robustness of the algorithm during startup. Then, real-time data on condenser and evaporator temperatures and heat exchange are collected, and anomalies are eliminated through boundary and lower limit checks to ensure input reliability. Next, the temperature difference and its rate of change are calculated to estimate instantaneous thermal resistance and heat capacity, reflecting the dynamic characteristics of system heat exchange and heat storage. Based on this, the smoothing coefficient is adaptively adjusted using the parameter difference between two consecutive periods, and a double exponential smoothing algorithm is used to iteratively update thermal resistance and heat capacity, balancing response speed and anti-interference capability. Subsequently, the updated parameters undergo physical rationality correction and historical data is continuously stored to support the next iteration. Finally, the recommended heat exchange is calculated based on the latest smoothed thermal resistance result, and range correction is performed, using this as a closed-loop control reference to achieve precise adjustment of refrigerant flow or compressor load. Summary of the Invention

[0005] This invention provides an AI-based simulation platform for energy-saving optimization of central air conditioning systems, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: an AI-based energy-saving optimization method for central air conditioning systems, comprising:

[0007] Set the data sampling period, obtain and set the system rated temperature difference, temperature sensor resolution, heat exchange sensor resolution and system rated maximum heat exchange from the equipment manual, initialize the bi-exponential smoothing parameters of thermal resistance and heat capacity, and set non-zero protection constants for temperature difference, thermal resistance and heat capacity calculations.

[0008] Collect the condenser temperature, evaporator temperature and instantaneous heat exchange of the central air conditioning system, and perform out-of-bounds and lower-limit checks on the collected data. Record alarms and skip the current cycle for data that exceeds the sensor range or is below the set threshold.

[0009] The temperature difference between the condenser and the evaporator is calculated based on the collected temperature data, and the first-order rate of change is calculated based on two consecutive temperature difference data, while a minimum rate of change threshold is set.

[0010] The instantaneous thermal resistance and instantaneous heat capacity of the system are estimated using temperature difference and heat exchange data, respectively.

[0011] The increment is calculated based on the difference between instantaneous thermal resistance and heat capacity in two consecutive periods, and the smoothing coefficients of thermal resistance and heat capacity are adaptively adjusted accordingly.

[0012] A double exponential smoothing algorithm is used to iteratively update the thermal resistance and heat capacity parameters;

[0013] The updated thermal resistance and thermal capacity parameters are physically corrected for rationality, and historical data is stored in a rolling manner according to time sequence.

[0014] The recommended heat exchange rate is calculated based on the latest smoothed thermal resistance results, and the results are corrected for range. The corrected heat exchange rate is then used as a closed-loop control reference for the refrigerant flow rate or compressor load in the next cycle.

[0015] Optionally, the setting of the data sampling period involves obtaining and setting the system's rated temperature difference, temperature sensor resolution, heat exchange sensor resolution, and system's rated maximum heat exchange from the equipment manual, initializing the bi-exponential smoothing parameters for thermal resistance and heat capacity, and setting non-zero protection constants for the calculation of temperature difference, thermal resistance, and heat capacity. Specifically, this includes:

[0016] The sampling period is set to Δt;

[0017] Obtain the rated temperature difference ΔT from the equipment manual. nom Temperature sensor resolution δ T δ resolution of heat transfer sensor Q and the system's rated maximum heat exchange Q max ;

[0018] Set a zero-prevention constant: Where, ε R This is for use in thermal resistance calculations where the denominator is non-zero; For use in preventing zero-limiting of temperature difference rate of change; ε C This is used for protecting the denominator from being non-zero in heat capacity calculations;

[0019] Initial smoothing parameters: Among them, R init To initialize the system's smooth thermal resistance; C init The system's initial heat capacity is smoothed.

[0020] For all n values ​​of -2, -1, 0, set:

[0021]

[0022] Where R[n] is the smoothed thermal resistance estimate after the nth sampling; C[n] is the instantaneous thermal resistance estimate for the nth sampling; C[n] is the smoothed thermal capacity estimate after the nth sampling. This is the instantaneous heat capacity estimate for the nth sampling; ΔT[n] is the temperature difference for the nth sampling; denoted as the first-order rate of change of temperature difference in the nth sampling; n is the sampling index, with negative values ​​indicating earlier historical data in the time series.

[0023] Optionally, the process involves collecting the condenser temperature, evaporator temperature, and instantaneous heat exchange of the central air conditioning system, and performing out-of-range and lower-limit checks on the collected data. For data exceeding the sensor's range or falling below a set threshold, an alarm is recorded and the current processing cycle is skipped. Specifically, this includes:

[0024] For each sampling index n≥1, perform the following steps:

[0025] S101, Measurement Data:

[0026] Obtain the condenser temperature T of the nth sample. cond [n], the evaporator temperature T during the nth sampling. evap [n] and the instantaneous heat transfer Q measured in the nth sampling. meas [n];

[0027] S102, Boundary and Lower Limit Checks:

[0028] If T cond [n] or T evap If [n] exceeds the range of its respective sensor, an alarm is recorded and the current cycle is skipped;

[0029] If Q meas If [n]≤0, then Q meas [n]←δ Q .

[0030] Optionally, the step of calculating the temperature difference between the condenser and evaporator based on the collected temperature data, and calculating the first-order rate of change based on two consecutive temperature difference data points, while setting a minimum rate of change threshold, specifically includes:

[0031] Calculate the condenser-evaporator temperature difference ΔT[n] = T at the nth sampling time. cond [d]-T evap [n];

[0032] Calculate the rate of change of temperature difference at the nth sampling.

[0033] Set the minimum rate of change Where max(a,b) is the larger of a and b.

[0034] Optionally, estimating the instantaneous thermal resistance and instantaneous heat capacity of the system using temperature difference and heat transfer data specifically includes:

[0035] Calculate the instantaneous thermal resistance estimate for the nth sampling.

[0036] Calculate the instantaneous heat capacity estimate for the nth sample.

[0037] Optionally, the step of calculating the increment based on the difference between instantaneous thermal resistance and heat capacity in two consecutive periods, and adaptively adjusting the smoothing coefficients of thermal resistance and heat capacity accordingly, specifically includes:

[0038] Calculate the two-level increment:

[0039]

[0040] Where ΔR1[n] and ΔR2[n] are the differences in instantaneous thermal resistance between two consecutive periods; ΔC1[n] and ΔC2[n] are the differences in instantaneous heat capacity between two consecutive periods;

[0041] Set the thermal resistance smoothing factor to

[0042] Set the heat capacity smoothing factor to

[0043] Optionally, the step of iteratively updating the thermal resistance and heat capacity parameters using a double exponential smoothing algorithm specifically includes:

[0044] Constructing a thermal resistance smoothing function:

[0045] Construct a heat capacity smoothing function:

[0046] Optionally, the step of performing physical rationality correction on the updated thermal resistance and thermal capacity parameters and storing historical data in a rolling manner over time specifically includes:

[0047] Perform positive value correction: R[n]←max(R[n],ε R ), C[n]←max(C[n],ε C );

[0048] The data is stored sequentially and rolled forward for the next iteration, specifically:

[0049]

[0050] ΔT[n-2]←ΔT[n-1], ΔT[n-1]←ΔT[n];

[0051]

[0052] Optionally, the step of calculating the recommended heat exchange rate based on the latest smoothed thermal resistance results, performing range correction on the results, and using the corrected heat exchange rate as a closed-loop control reference for the refrigerant flow rate or compressor load in the next cycle specifically includes:

[0053] The recommended heat transfer is calculated based on the latest smoothed thermal resistance.

[0054] Range correction: Q *[n]←min(max(Q * [n],δ Q ),Q max );

[0055] Q * [n] serves as a reference for the refrigerant flow rate or compressor load in the next cycle, completing one closed-loop optimization setting.

[0056] A simulation platform for implementing the AI-based energy-saving optimization method for central air conditioning systems includes:

[0057] Data acquisition module: used to acquire condenser temperature, evaporator temperature, and heat exchange data;

[0058] Data processing module: connected to the data acquisition module, used to perform temperature difference and temperature difference change rate processing, instantaneous thermal resistance and heat capacity estimation, and incremental calculation and adaptive smoothing coefficient determination;

[0059] Smoothing module: connected to the data processing module, used to implement double exponential smoothing updates and physical corrections and historical shifts;

[0060] Optimization calculation module: connected to the smoothing processing module, used to calculate and correct the optimal heat exchange.

[0061] Control output module: Connected to the optimization calculation module, it is used to output control signals suitable for refrigerant flow or compressor load regulation in the simulation environment.

[0062] The present invention has the following beneficial effects:

[0063] 1. By automatically reading the system's rated temperature difference, sensor resolution, and rated maximum heat exchange from the equipment manual, and combining this with the user-defined data sampling period, the system's basic parameters are uniformly configured. Simultaneously, a set of non-zero protection constants is set to protect the denominator in the calculations of temperature difference, thermal resistance, and heat capacity, preventing algorithm instability caused by zero or extremely small values. Deeply integrating equipment design parameters with the operating algorithm automates and refines parameter acquisition, eliminating the previous reliance on manual input or experience-based estimation; improves the reliability and consistency of system initialization, reducing the risk of human error; solves the common problem in existing technologies of "inaccurate algorithm startup parameters leading to large initial estimation deviations"; and compared to traditional fixed-parameter or manual parameter tuning methods, shortens deployment time and lays a solid foundation for stable subsequent algorithm operation.

[0064] 2. The system acquires the condenser temperature, evaporator temperature, and instantaneous heat exchange of the central air conditioning system in real time, and performs out-of-range and lower-limit checks on the collected data. For measurements exceeding the sensor range or falling below the set threshold, the system automatically records an alarm and skips the current cycle to remove invalid or abnormal data; boundary values ​​are limited to ensure the data remains within a reasonable range. Combining multiple verification mechanisms with real-time monitoring preserves boundary value information to increase data utilization while effectively eliminating extreme errors; it improves the reliability of input data and effectively prevents the accumulation of calculation errors caused by sensor drift or malfunction; it solves the problems of coarse handling of abnormal data and susceptibility to algorithm interference in existing technologies; compared to simple threshold limits or manual data screening methods, it achieves fully automatic, continuous, and seamless data cleaning, ensuring the accuracy and stability of subsequent estimations and optimizations.

[0065] 3. Based on the calibrated temperature data, the temperature difference between the condenser and evaporator is calculated. Furthermore, the first-order rate of change of the temperature difference is calculated based on two consecutive temperature difference data points, while a minimum change threshold is set to suppress minor fluctuations. Combining the rate of change of temperature difference with the minimum threshold protection mechanism not only promptly captures sudden changes in system state but also filters out measurement noise; it improves the accuracy of dynamic feature extraction and enhances the stability of subsequent heat capacity estimation; it solves the problem that existing technologies that simply calculate the temperature difference or directly calculate the rate of change are easily affected by small fluctuations; compared with traditional methods that rely solely on temperature difference or simple differential calculations, this method balances response speed and anti-interference capabilities, providing a more reliable foundation for subsequent performance parameter estimation.

[0066] 4. The instantaneous thermal resistance and instantaneous heat capacity of the system are estimated using the calculated temperature difference and heat exchange data to reflect the heat exchange efficiency and heat storage capacity of the central air conditioning system at a specific moment. By simultaneously estimating these two core thermal performance indicators, thermal resistance and heat capacity, a more comprehensive dynamic model of the system's thermal performance is constructed. This model can reflect the changes in the system's thermal characteristics under different operating conditions in real time, providing a data foundation for refined energy-saving strategies. It solves the problem of one-sided and non-real-time estimation of single thermal resistance or heat capacity in existing technologies. Compared with traditional offline testing or empirical model estimation, this method achieves online, dynamic, and synchronous dual-parameter estimation, improving the accuracy and timeliness of thermal performance monitoring.

[0067] 5. By comparing the difference between instantaneous thermal resistance and instantaneous heat capacity over two consecutive periods, the parameter change increment is calculated, and an adaptive smoothing coefficient is dynamically generated accordingly. The smoothing coefficient is automatically adjusted based on the magnitude of the parameter increment, achieving an adaptive trade-off between rapid response to sudden changes in operating conditions and noise suppression in stable operating conditions. This maintains the algorithm's sensitivity to rapid changes in state while avoiding the lag or excessive oscillation caused by a fixed smoothing coefficient. It solves the problems of traditional exponential smoothing requiring manual parameter tuning and difficulty in accommodating multiple operating conditions. Compared with fixed smoothing factors or manual adjustment based on experience, this method improves the robustness and adaptability of the algorithm, ensuring that parameter updates are neither distorted nor excessively delayed.

[0068] 6. A double exponential smoothing algorithm is employed to iteratively update thermal resistance and thermal capacity parameters, preserving data trends while suppressing random noise. Applying the double exponential smoothing strategy to thermal performance parameters enables the system to track changes in thermal state in real time while ensuring smooth transitions; it improves the stability and reliability of parameter estimation, avoiding misjudgments caused by sudden noise jumps; it solves the problems of slow response or oversensitivity of single exponential smoothing in existing technologies; compared with traditional single exponential smoothing or mean filtering techniques, this method balances rapid trend capture with effective noise filtering, providing more accurate and smoother parameter curves for subsequent optimization decisions.

[0069] 7. Physical feasibility correction is performed on the updated smoothed thermal resistance and heat capacity parameters, forcing the parameters to be no lower than the design-preset minimum values. All historical parameters are stored in chronological order to provide context information for the next iteration. Physical boundary constraints are introduced to ensure that the algorithm output always conforms to the actual system engineering situation; non-physical interpretations caused by algorithm drift are eliminated, effectively improving safety and interpretability; the problems of improper historical data management and difficulty in correcting parameter drift or accumulated errors in existing technologies are solved; compared with unconstrained rolling storage or offline boundary checking methods, this method realizes real-time, automatic, and online physical correction and historical management, further enhancing the stability and application credibility of the algorithm.

[0070] 8. Based on the latest smoothed thermal resistance results, the recommended heat exchange is calculated and range correction is performed to meet equipment resolution and limit constraints. The corrected heat exchange is then used as a closed-loop control reference for the refrigerant flow or compressor load in the next cycle. Thermal performance parameters are directly mapped to executable control commands, achieving seamless integration between algorithm output and hardware control. This ensures the optimization results are practically feasible, effectively saving energy while avoiding over- or under-adjustment. It solves the problem of existing technologies where optimization objectives and execution mechanisms are disconnected, and results are difficult to implement directly. Compared to traditional methods that only provide optimization suggestions or require secondary conversion, this method forms an end-to-end closed-loop optimization, directly driving the operation of the execution equipment, improving energy efficiency and system automation. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Example, refer to Figure 1 An AI-based energy-saving optimization method for central air conditioning systems includes:

[0074] Set the data sampling period, obtain and set the system rated temperature difference, temperature sensor resolution, heat exchange sensor resolution and system rated maximum heat exchange from the equipment manual, initialize the bi-exponential smoothing parameters of thermal resistance and heat capacity, and set non-zero protection constants for temperature difference, thermal resistance and heat capacity calculations.

[0075] Collect the condenser temperature, evaporator temperature and instantaneous heat exchange of the central air conditioning system, and perform out-of-bounds and lower-limit checks on the collected data. Record alarms and skip the current cycle for data that exceeds the sensor range or is below the set threshold.

[0076] The temperature difference between the condenser and the evaporator is calculated based on the collected temperature data, and the first-order rate of change is calculated based on two consecutive temperature difference data, while a minimum rate of change threshold is set.

[0077] The instantaneous thermal resistance and instantaneous heat capacity of the system are estimated using temperature difference and heat exchange data, respectively.

[0078] The increment is calculated based on the difference between instantaneous thermal resistance and heat capacity in two consecutive periods, and the smoothing coefficients of thermal resistance and heat capacity are adaptively adjusted accordingly.

[0079] A double exponential smoothing algorithm is used to iteratively update the thermal resistance and heat capacity parameters;

[0080] The updated thermal resistance and thermal capacity parameters are physically corrected for rationality, and historical data is stored in a rolling manner according to time sequence.

[0081] The recommended heat exchange rate is calculated based on the latest smoothed thermal resistance results, and the results are corrected for range. The corrected heat exchange rate is then used as a closed-loop control reference for the refrigerant flow rate or compressor load in the next cycle.

[0082] By setting a unified data sampling period and automatically acquiring design parameters such as the system's rated temperature difference, sensor resolution, and maximum heat exchange, and by introducing non-zero protection constants in the calculations of temperature difference, thermal resistance, and heat capacity, a reliable benchmark is ensured during the algorithm's startup phase. Real-time acquisition of condenser temperature, evaporator temperature, and heat exchange data, combined with out-of-bounds and lower-limit checks, eliminates abnormal or invalid data, ensuring the accuracy of subsequent calculation inputs. Temperature difference and its first-order rate of change are calculated based on two consecutive temperature difference data points, and the instantaneous thermal resistance and heat capacity of the system are estimated using temperature difference and heat exchange information, reflecting the system's dynamic heat exchange characteristics in real time. A smoothing coefficient is adaptively generated by comparing the thermal resistance and heat capacity differences over consecutive periods, and a double-exponential smoothing algorithm is used to smoothly update parameters, achieving effective noise suppression and rapid response to sudden operating conditions. Physical rationality correction of the smoothed parameters and rolling storage of historical data ensure that the calculation results always conform to engineering realities. The recommended heat exchange is calculated and corrected based on the latest smoothed thermal resistance results, forming an executable refrigerant flow or compressor load closed-loop control reference, directly driving equipment execution. This technical solution solves the problems caused by traditional manual parameter tuning, such as large deviations in startup estimation, abnormal measurement data and severe noise interference, difficulty in balancing response speed and stability in algorithms, and difficulty in directly implementing optimization results. It improves the automation level, calculation accuracy and system reliability of energy-saving optimization.

[0083] The set data sampling period involves obtaining and setting the system's rated temperature difference, temperature sensor resolution, heat exchange sensor resolution, and system's rated maximum heat exchange from the equipment manual; initializing the bi-exponential smoothing parameters for thermal resistance and heat capacity; and setting non-zero protection constants for the calculation of temperature difference, thermal resistance, and heat capacity. Specifically, this includes:

[0084] Set the sampling period to Δt; define a fixed time interval for system data acquisition to provide a unified benchmark for all timing operations;

[0085] Obtain the rated temperature difference ΔT from the equipment manual. nom Temperature sensor resolution δ T δ resolution of heat transfer sensor Q and the system's rated maximum heat exchange Q max Read the rated temperature difference, sensor resolution, and maximum heat exchange of the system from the equipment manual to provide design reference values ​​for subsequent parameter calculations;

[0086] Set a zero-prevention constant: Where, ε R This is for use in thermal resistance calculations where the denominator is non-zero; For use in preventing zero-limiting of temperature difference rate of change; ε C To protect the denominator from being zero in heat capacity calculations; to provide denominator protection for calculations of thermal resistance, rate of change of temperature difference, and heat capacity, respectively, to avoid division by zero or numerical instability;

[0087] Initial smoothing parameters: Among them, R init To initialize the system's smooth thermal resistance; C init The system's initial heat capacity is smoothed.

[0088] For all n values ​​of -2, -1, 0, set:

[0089]

[0090] Where R[n] is the smoothed thermal resistance estimate after the nth sampling; C[n] is the instantaneous thermal resistance estimate for the nth sampling; C[n] is the smoothed thermal capacity estimate after the nth sampling. This is the instantaneous heat capacity estimate for the nth sampling; ΔT[n] is the temperature difference for the nth sampling; is the first-order rate of change of temperature difference in the nth sampling; n is the sampling index, and a negative value indicates earlier historical data in time series; the smoothing parameter and instantaneous estimation register are initialized with nominal values ​​to ensure that there is a reasonable benchmark and no historical legacy error when the algorithm starts.

[0091] By defining a unified sampling period and automatically retrieving the system's rated temperature difference, sensor resolution, and maximum heat exchange from the equipment manual, accurate and reliable design reference values ​​are provided for the algorithm, avoiding errors caused by previous experience-based estimations or manual input. By setting a dedicated zero-prevention constant, the denominator in the calculations of temperature difference, thermal resistance, and heat capacity is protected, completely eliminating the instability caused by division by zero or excessively small values. By initializing the double exponential smoothing parameters and assigning reasonable nominal initial values ​​to all variables, the algorithm has a good initial state at startup, avoiding the uncertainty of historical data. Through the above steps, the problems of algorithm initialization relying on manual intervention, lack of protection mechanisms, and large initial estimation deviations in existing technologies are solved, providing solid and stable benchmark support for subsequent dynamic calculations in each stage, and improving the repeatability and reliability of the entire energy-saving optimization process of the system.

[0092] The system collects the condenser temperature, evaporator temperature, and instantaneous heat exchange of the central air conditioning system, and performs out-of-range and lower-limit checks on the collected data. Data exceeding the sensor's range or falling below a set threshold is recorded with an alarm and the current processing cycle is skipped. Specifically, this includes:

[0093] For each sampling index n≥1, perform the following steps:

[0094] S101, Measurement Data:

[0095] Obtain the condenser temperature T of the nth sample. cond [n], the evaporator temperature T during the nth sampling. evap[n] and the instantaneous heat transfer Q measured in the nth sampling. meas [n];

[0096] Real-time acquisition of condenser / evaporator temperature and instantaneous heat exchange is used as input for subsequent estimation;

[0097] S102, Boundary and Lower Limit Checks:

[0098] If T cond [n] or T evap If [n] exceeds the range of its respective sensor, an alarm is recorded and the current cycle is skipped;

[0099] If Q meas If [n]≤0, then Q meas [n]←δ Q ;

[0100] Abnormal measurements are eliminated and the heat exchange is ensured to be positive to prevent invalid or negative data from affecting the calculation.

[0101] By measuring condenser temperature, evaporator temperature, and heat exchange data in real time, and combining this with out-of-bounds and lower-limit verification mechanisms, abnormal data exceeding the sensor's effective range or falling below the threshold is automatically eliminated. This solves the problems of false alarms and invalid values ​​being included in calculations due to sensor drift, malfunction, or external interference in traditional data acquisition processes. By recording alarms and skipping abnormal cycles, the cumulative impact of abnormal data on parameter estimation is avoided, ensuring the continuity and accuracy of subsequent calculations. By applying amplitude limiting to boundary values, the heat exchange data is ensured to always be positive, avoiding physical meaning conflicts caused by negative values. These steps improve the quality of input data, enhance the robustness and stability of the entire energy-saving optimization algorithm, and provide a highly reliable data foundation for subsequent thermal performance estimation and optimization decisions.

[0102] The process involves calculating the temperature difference between the condenser and evaporator based on the collected temperature data, calculating the first-order rate of change based on two consecutive temperature difference data points, and setting a minimum rate of change threshold. Specifically, this includes:

[0103] Calculate the condenser-evaporator temperature difference ΔT[n] = T at the nth sampling time. cond [n]-T evap [n]; Calculates the instantaneous temperature difference of the system, providing basic data for thermal resistance estimation;

[0104] Calculate the rate of change of temperature difference at the nth sampling. Calculate the rate of change of temperature difference over time to provide basic data for heat capacity estimation;

[0105] Set the minimum rate of change Among them, max(a,b) is to take the larger of a and b; the lower limit of the rate of change of temperature difference is protected to avoid the calculation being unstable due to the value being too small.

[0106] By calculating the real-time temperature difference between the condenser and evaporator, an accurate input basis is provided for thermal resistance estimation, solving the deviation problem that may be caused by directly using uncalibrated temperature data. By calculating the first-order rate of change of the temperature difference from two consecutive temperature difference data, dynamic rate information is provided for heat capacity estimation, solving the problem that static temperature difference cannot reflect the transient heat storage capacity of the system. By setting a minimum rate of change threshold, the influence of measurement noise and small fluctuations is suppressed, avoiding numerical instability or excessive amplification of the temperature difference rate of change when it is extremely small. The above steps enable the system to achieve a balance between capturing rapid changes in operating conditions and suppressing measurement noise, improving the accuracy and stability of subsequent thermal performance estimation.

[0107] The method of estimating the instantaneous thermal resistance and instantaneous heat capacity of the system using temperature difference and heat exchange data specifically includes:

[0108] Calculate the instantaneous thermal resistance estimate for the nth sampling. The instantaneous thermal resistance is calculated based on the current temperature difference and heat exchange to reflect the transient heat exchange characteristics of the system.

[0109] Calculate the instantaneous heat capacity estimate for the nth sample. Instantaneous heat capacity is calculated based on the current heat exchange and temperature difference change rate to reflect the dynamics of the system's heat capacity.

[0110] By estimating instantaneous thermal resistance using calculated temperature difference and measured heat transfer data, the algorithm can capture the system's heat exchange resistance under different operating conditions in real time, solving the problems of lag and low granularity in estimation by traditional offline experiments or empirical models. By estimating instantaneous heat capacity using temperature difference change rate and heat transfer data, the algorithm can reflect the system's real-time heat storage dynamics, solving the problem that relying solely on thermal resistance cannot reflect heat storage characteristics. Simultaneous estimation of instantaneous thermal resistance and heat capacity provides comprehensive and accurate physical measures for subsequent adaptive smoothing and optimization calculations, improving the reliability and pertinence of energy-saving optimization strategies. This step enhances the system's ability to perceive dynamic changes in thermal performance, providing a solid data foundation for achieving refined energy-saving control.

[0111] The calculation of the increment based on the difference between instantaneous thermal resistance and heat capacity in two consecutive periods, and the adaptive adjustment of the smoothing coefficients of thermal resistance and heat capacity accordingly, specifically includes:

[0112] Calculate the two-level increment:

[0113]

[0114] Where ΔR1[n] and ΔR2[n] are the differences in instantaneous thermal resistance between two consecutive periods; ΔC1[n] and ΔC2[n] are the differences in instantaneous heat capacity between two consecutive periods;

[0115] The difference between the current instantaneous estimate and the previous two periods is calculated to measure the trend of parameter change.

[0116] Set the thermal resistance smoothing factor to

[0117] Set the heat capacity smoothing factor to

[0118] The exponential smoothing factor is dynamically generated based on the increment size to achieve automatic trade-offs between sudden changes and stable operating conditions.

[0119] By incrementally calculating the difference between instantaneous thermal resistance and heat capacity over two consecutive periods, the drastic change in the system's current operating condition is accurately measured, solving the problem of slow response or excessive oscillation under abrupt changes in a fixed smoothing factor. Through adaptive adjustment of the smoothing coefficients of thermal resistance and heat capacity, the algorithm can respond quickly during abrupt changes and effectively suppress noise during stable conditions, improving the flexibility and reliability of parameter updates. This mechanism avoids the tedious manual parameter tuning, ensuring that the algorithm maintains the optimal smoothing-response balance under various operating conditions. Furthermore, it enhances the adaptability of the energy-saving optimization method to dynamic system changes, strengthening the algorithm's robustness and universality.

[0120] The method of iteratively updating the thermal resistance and heat capacity parameters using a double exponential smoothing algorithm specifically includes:

[0121] Constructing a thermal resistance smoothing function: The instantaneous thermal resistance is exponentially smoothed to suppress noise and preserve abrupt change information;

[0122] Construct a heat capacity smoothing function: The instantaneous heat capacity is exponentially smoothed to balance response speed and stability.

[0123] By employing a double exponential smoothing algorithm to perform trend smoothing and rapid response processing on instantaneous thermal resistance and heat capacity parameters respectively, the problem of single exponential smoothing being difficult to balance between trend capture and noise suppression is solved. This algorithm effectively suppresses the interference of data noise on parameter estimation while preserving the trend of system thermal performance changes, improving the smoothness and reliability of parameter curves. The iterative update mechanism of double exponential smoothing enables the algorithm to dynamically track changes in system state, avoiding excessive lag under sudden operating conditions. This provides a more stable and accurate input for subsequent physical rationality correction and heat exchange optimization calculations, thereby ensuring the stability and executability of the energy-saving control strategy.

[0124] The process of performing physical rationality correction on the updated thermal resistance and heat capacity parameters, and storing historical data in a rolling manner over time, specifically includes:

[0125] Perform positive value correction: R[n]←max(R[n],ε R), C[n]←max(C[n],ε C The forced smoothing parameter should not be lower than the zero-prevention constant to ensure physical feasibility.

[0126] The data is stored sequentially and rolled forward for the next iteration, specifically:

[0127]

[0128] ΔT[n-2]←ΔT[n-1], ΔT[n-1]←ΔT[n];

[0129]

[0130] Update the historical data register to provide the latest context information for the next iteration.

[0131] By enforcing lower limit protection on the smoothly updated thermal resistance and thermal capacity parameters, the system avoids parameters falling below reasonable physical boundaries due to algorithm drift or sudden anomalies, thus resolving non-physical interpretations and safety hazards caused by unconstrained updates. By rolling the storage of historical parameters and providing context data for the next iteration, the system addresses the problem of lack of historical support and the tendency for cumulative errors in single iterations. These steps ensure that the parameter output always conforms to engineering realities and has good traceability. They also improve the stability and interpretability of the algorithm's long-term operation, providing a strong guarantee for the safe and reliable implementation of continuous energy-saving optimization.

[0132] The process of calculating the recommended heat exchange rate based on the latest smoothed thermal resistance results, and then performing range correction on the results, using the corrected heat exchange rate as a closed-loop control reference for the refrigerant flow rate or compressor load in the next cycle, specifically includes:

[0133] The recommended heat transfer is calculated based on the latest smoothed thermal resistance. Based on the latest smooth thermal resistance and nominal temperature difference, calculate the optimal heat exchange reference for the next cycle;

[0134] Range correction: Q * [n]←min(max(Q * [n],δ Q ),Q max ); Apply resolution and device limits to the recommended values ​​to ensure executability;

[0135] Q * [n] serves as a reference for the refrigerant flow or compressor load in the next cycle, completing a closed-loop optimization setting; the calculation results are then pushed to the actuator to achieve actual energy-saving control of the system.

[0136] By utilizing the latest smoothed thermal resistance parameters to calculate the optimal heat exchange, the recommended heat exchange accurately reflects the current thermal performance of the system, solving the problems of traditional optimization methods being out of touch with actual operating conditions and the optimization results being difficult to apply. By imposing resolution and equipment limit constraints on the calculation results, it is ensured that the recommended values ​​are within the actual executable range, avoiding secondary energy waste caused by excessive or insufficient adjustment. The corrected heat exchange is directly used as a closed-loop control reference for the refrigerant flow or compressor load in the next cycle, realizing an end-to-end closed loop from data calculation to control execution, solving the pain point of the disconnect between optimization suggestions and execution. This improves the real-time performance and execution efficiency of energy-saving optimization, reduces the cost of manual intervention, and enhances system automation and energy-saving benefits.

[0137] This embodiment also provides a simulation platform for an AI-based energy-saving optimization method for central air conditioning systems, including:

[0138] Data acquisition module: used to acquire condenser temperature, evaporator temperature, and heat exchange data;

[0139] Data processing module: connected to the data acquisition module, used to perform temperature difference and temperature difference change rate processing, instantaneous thermal resistance and heat capacity estimation, and incremental calculation and adaptive smoothing coefficient determination;

[0140] Smoothing module: connected to the data processing module, used to implement double exponential smoothing updates and physical corrections and historical shifts;

[0141] Optimization calculation module: connected to the smoothing processing module, used to calculate and correct the optimal heat exchange.

[0142] Control output module: Connected to the optimization calculation module, it is used to output control signals suitable for refrigerant flow or compressor load regulation in the simulation environment.

[0143] By modularly integrating data acquisition, processing, smoothing updates, optimization calculations, and control output functions into the simulation platform, a complete verification environment for energy-saving optimization methods is achieved, solving the problem of providing only methods but lacking simulation verification tools. Through modular design, the simulation platform can flexibly replace or upgrade various functional modules to meet the testing needs of different scenarios and equipment, solving the problems of poor versatility and high customization costs of existing simulation tools. The platform's closed-loop control output module can feed back the optimization calculation results to the simulation environment in real time, enabling visualized adjustment of refrigerant flow or compressor load, solving the problem of disconnect between optimization results and simulation execution. It provides a reusable and scalable verification and testing tool, reducing the risk assessment cost before the implementation of energy-saving optimization solutions and accelerating the transformation of algorithms from design to engineering applications.

[0144] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0145] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An AI-based energy-saving optimization method for central air conditioning systems, characterized in that, include: Set the data sampling period, obtain and set the system rated temperature difference, temperature sensor resolution, heat exchange sensor resolution and system rated maximum heat exchange from the equipment manual, initialize the bi-exponential smoothing parameters of thermal resistance and heat capacity, and set non-zero protection constants for temperature difference, thermal resistance and heat capacity calculations. Collect the condenser temperature, evaporator temperature and instantaneous heat exchange of the central air conditioning system, and perform out-of-bounds and lower-limit checks on the collected data. Record alarms and skip the current cycle for data that exceeds the sensor range or is below the set threshold. The temperature difference between the condenser and the evaporator is calculated based on the collected temperature data, and the first-order rate of change is calculated based on two consecutive temperature difference data, while a minimum rate of change threshold is set. The instantaneous thermal resistance and instantaneous heat capacity of the system are estimated using temperature difference and heat exchange data, respectively. The increment is calculated based on the difference between instantaneous thermal resistance and heat capacity in two consecutive periods, and the smoothing coefficients of thermal resistance and heat capacity are adaptively adjusted accordingly. A double exponential smoothing algorithm is used to iteratively update the thermal resistance and heat capacity parameters; The updated thermal resistance and thermal capacity parameters are physically corrected for rationality, and historical data is stored in a rolling manner according to time sequence. The recommended heat exchange rate is calculated based on the latest smoothed thermal resistance results, and the results are corrected for range. The corrected heat exchange rate is then used as a closed-loop control reference for the refrigerant flow rate or compressor load in the next cycle.

2. The energy-saving optimization method for a central air conditioning system based on AI according to claim 1, characterized in that, The set data sampling period involves obtaining and setting the system's rated temperature difference, temperature sensor resolution, heat exchange sensor resolution, and system's rated maximum heat exchange from the equipment manual; initializing the bi-exponential smoothing parameters for thermal resistance and heat capacity; and setting non-zero protection constants for the calculation of temperature difference, thermal resistance, and heat capacity. Specifically, this includes: The sampling period is set to Δt; Obtain the rated temperature difference ΔT from the equipment manual. nom Temperature sensor resolution δ T Heat exchange sensor resolution δ Q and the system's rated maximum heat exchange Q max ; Set a zero-prevention constant: Where, ε R This is for use in thermal resistance calculations where the denominator is non-zero; For use in preventing zero-limiting of temperature difference rate of change; ε C This is used for protecting the denominator from being non-zero in heat capacity calculations; Initial smoothing parameters: Among them, R init To initialize the system's smooth thermal resistance; C init The system's initial heat capacity is smoothed. For all n values ​​of -2, -1, 0, set: Where R[n] is the smoothed thermal resistance estimate after the nth sampling; C[n] is the instantaneous thermal resistance estimate for the nth sampling; C[n] is the smoothed thermal capacity estimate after the nth sampling. This is the instantaneous heat capacity estimate for the nth sampling; ΔT[n] is the temperature difference for the nth sampling; denoted as the first-order rate of change of temperature difference in the nth sampling; n is the sampling index, with negative values ​​indicating earlier historical data in the time series.

3. The AI-based energy-saving optimization method for a central air conditioning system according to claim 2, characterized in that, The system collects the condenser temperature, evaporator temperature, and instantaneous heat exchange of the central air conditioning system, and performs out-of-range and lower-limit checks on the collected data. Data exceeding the sensor's range or falling below a set threshold is recorded with an alarm and the current processing cycle is skipped. Specifically, this includes: For each sampling index n≥1, perform the following steps: S101, Measurement Data: Obtain the condenser temperature T of the nth sample. cond [n], the evaporator temperature T during the nth sampling. evap [n] and the instantaneous heat transfer Q measured in the nth sampling. meas [n]; S102, Boundary and Lower Limit Checks: If T cond [n] or T evap If [n] exceeds the range of its respective sensor, an alarm is recorded and the current cycle is skipped; If Q meas If [n]≤0, then Q meas [n]←δ Q .

4. The AI-based energy-saving optimization method for a central air conditioning system according to claim 3, characterized in that, The process involves calculating the temperature difference between the condenser and evaporator based on the collected temperature data, calculating the first-order rate of change based on two consecutive temperature difference data points, and setting a minimum rate of change threshold. Specifically, this includes: Calculate the condenser-evaporator temperature difference ΔT[n] = T at the nth sampling time. cond [n]-T evap [n]; Calculate the rate of change of temperature difference at the nth sampling. Set the minimum rate of change Where max(a,b) is the larger of a and b.

5. The AI-based energy-saving optimization method for a central air conditioning system according to claim 4, characterized in that, The method of estimating the instantaneous thermal resistance and instantaneous heat capacity of the system using temperature difference and heat exchange data specifically includes: Calculate the instantaneous thermal resistance estimate for the nth sampling. Calculate the instantaneous heat capacity estimate for the nth sample.

6. The energy-saving optimization method for a central air conditioning system based on AI according to claim 5, characterized in that, The calculation of the increment based on the difference between instantaneous thermal resistance and heat capacity in two consecutive periods, and the adaptive adjustment of the smoothing coefficients of thermal resistance and heat capacity accordingly, specifically includes: Calculate the two-level increment: Where ΔR1[n] and ΔR2[n] are the differences in instantaneous thermal resistance between two consecutive periods; ΔC1[n] and ΔC2[n] are the differences in instantaneous heat capacity between two consecutive periods; Set the thermal resistance smoothing factor to Set the heat capacity smoothing factor to 7. The AI-based energy-saving optimization method for a central air conditioning system according to claim 6, characterized in that, The method of iteratively updating the thermal resistance and heat capacity parameters using a double exponential smoothing algorithm specifically includes: Constructing a thermal resistance smoothing function: Construct a heat capacity smoothing function:

8. The AI-based energy-saving optimization method for a central air conditioning system according to claim 7, characterized in that, The process of performing physical rationality correction on the updated thermal resistance and heat capacity parameters, and storing historical data in a rolling manner over time, specifically includes: Perform positive value correction: R[n]←max(R[n],ε R ), C[n]←max(C[n],ε C ); The data is stored sequentially and rolled forward for the next iteration, specifically: ΔT[n-2]←ΔT[n-1], ΔT[n-1]←ΔT[n]; 9. The energy-saving optimization method for a central air conditioning system based on AI according to claim 8, characterized in that, The process of calculating the recommended heat exchange rate based on the latest smoothed thermal resistance results, and then performing range correction on the results, using the corrected heat exchange rate as a closed-loop control reference for the refrigerant flow rate or compressor load in the next cycle, specifically includes: The recommended heat transfer is calculated based on the latest smoothed thermal resistance. Range correction: Q * [n]←min(max(Q * [n],δ Q ),Q max ); Q * [n] serves as a reference for the refrigerant flow rate or compressor load in the next cycle, completing one closed-loop optimization setting.

10. A simulation platform employing the AI-based energy-saving optimization method for central air conditioning systems as described in claim 9, characterized in that, include: Data acquisition module: used to acquire condenser temperature, evaporator temperature, and heat exchange data; Data processing module: connected to the data acquisition module, used to perform temperature difference and temperature difference change rate processing, instantaneous thermal resistance and heat capacity estimation, and incremental calculation and adaptive smoothing coefficient determination; Smoothing module: connected to the data processing module, used to implement double exponential smoothing updates and physical corrections and historical shifts; Optimization calculation module: connected to the smoothing processing module, used to calculate and correct the optimal heat exchange. Control output module: Connected to the optimization calculation module, it is used to output control signals suitable for refrigerant flow or compressor load regulation in the simulation environment.

Images