Air conditioning system optimization configuration and control method based on twenty-four solar terms time sequence characteristics

Abstract

The application provides a kind of air conditioning system optimization method based on twenty-four solar terms time sequence characteristics, comprising: obtaining historical load-multidimensional weather data set containing twenty-four solar terms label;Using historical load-multidimensional weather data set containing twenty-four solar terms label to construct weather-solar terms coupled load prediction model;The historical load-multidimensional weather data set containing twenty-four solar terms label is input into weather-solar terms coupled load prediction model to obtain annual dynamic load prediction value hour by hour;Statistical annual dynamic load prediction value hour by hour and obtain minimum load value, peak load value, probability density function and cumulative distribution function;Optimization analysis is carried out to minimum load value, peak load value, probability density function and cumulative distribution function, and unit configuration strategy is obtained.The application can improve the rationality of air conditioning system configuration.

Description

Technical Field

[0001] This invention belongs to the field of HVAC control technology, and specifically relates to an air conditioning system optimization method based on the time sequence characteristics of the 24 solar terms. Background Technology

[0002] Current load forecasting and equipment selection for HVAC systems are typically based on static calculations using instantaneous peak loads on the "design date." This method has the following significant drawbacks: First, it ignores the continuity and stages of climate change, making it difficult to accurately reflect the unique climatic patterns corresponding to China's "Twenty-Four Solar Terms" (such as the alternating high temperature, high humidity, and thunderstorms during "Great Heat"). Second, it leads to excessive redundancy in system equipment selection, resulting in partial load operation for more than 95% of the year's operating time, leading to low overall energy efficiency and the risk of low-load surge. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an air conditioning system optimization method based on the time sequence characteristics of the 24 solar terms, which can improve the rationality of air conditioning system configuration.

[0004] This invention provides, in one aspect, a method for optimizing an air conditioning system based on the temporal characteristics of the twenty-four solar terms, comprising:

[0005] Obtain a historical load-multidimensional meteorological dataset containing labels for the 24 solar terms;

[0006] A meteorological-solar term coupled load prediction model was constructed using the historical load-multidimensional meteorological dataset containing the 24 solar terms labels.

[0007] The historical load-multidimensional meteorological dataset containing the 24 solar terms labels is input into the meteorological-solar term coupled load prediction model to obtain the hourly dynamic load prediction value for the whole year.

[0008] The hourly dynamic load forecast values ​​for the whole year are statistically analyzed to obtain the minimum load value, peak load value, probability density function and cumulative distribution function;

[0009] The minimum load value, the peak load value, the probability density function, and the cumulative distribution function are used to perform optimization analysis to obtain the unit configuration strategy.

[0010] Furthermore, obtaining the historical load-multidimensional meteorological dataset containing the 24 solar terms labels includes:

[0011] Collect historical load data and local meteorological data;

[0012] The historical load data and the local meteorological data are preprocessed to obtain the required historical load data and the required local meteorological data.

[0013] The required historical load data and the required local meteorological data are constructed into a multidimensional meteorological-load time series based on an annual time axis.

[0014] The multidimensional meteorological-load time series was preprocessed using a seasonal autoregressive integral moving average model, and the annual time axis was mapped to the time series slices of the twenty-four solar terms and labeled to obtain the historical load-multidimensional meteorological dataset containing the labels of the twenty-four solar terms.

[0015] Furthermore, the meteorological-solar term coupled load prediction model is constructed using the historical load-multidimensional meteorological dataset containing the 24 solar terms labels, specifically including:

[0016] The stationarity of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels was detected;

[0017] The historical load data of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels is decomposed into three parts: trend, seasonality, and residual by using a seasonal autoregressive integral moving average model to capture the seasonal fluctuations of the historical load data and obtain a time series model.

[0018] By embedding the 24 solar terms as seasonal variables into the time series model, a solar term time series model is obtained.

[0019] Weather type features are extracted from the historical load-multidimensional meteorological dataset containing the 24 solar terms labels, and the weather type features are integrated into the solar term time series model to obtain the meteorological-solar term coupled load prediction model.

[0020] Furthermore, the solar term time series model can calculate the corresponding typical daily load pattern for each solar term, obtain the correspondence between the twenty-four solar terms and the typical daily load pattern, which is the second correspondence, and provide the second correspondence as input features to the meteorological-solar term coupled load prediction model.

[0021] Furthermore, the meteorological-solar term coupled load prediction model includes the correlation and correspondence relationship between the twenty-four solar terms, weather type characteristics, and historical load data, namely the third correspondence relationship. The third correspondence relationship can automatically align the "hourly load" and "daily weather" and automatically identify which solar term each day belongs to.

[0022] Furthermore, the meteorological-solar term coupled load prediction model can also output the weather load scissor difference, which can quantify the differential impact of different weather types on load characteristic data during the same solar term.

[0023] Furthermore, the time series model includes a first inertial correspondence, a first noise smoothing correspondence, a first reverberation correspondence, and a first burst noise smoothing correspondence, specifically including:

[0024] The historical load data of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels is decomposed into a trend part, a seasonal part, and a residual part. The autoregressive parameter, moving average parameter, seasonal autoregressive parameter, and seasonal moving average parameter in the seasonal part are selected using the Akaike Information Criterion.

[0025] The autoregressive parameters reflect the first inertial correspondence;

[0026] The moving average parameters reflect the first noise smoothing correspondence;

[0027] The seasonal autoregressive parameters reflect the first response correspondence;

[0028] The seasonal moving average parameter reflects the smooth correspondence of the first burst noise.

[0029] Furthermore, the first inertial correspondence predicts the current load value using the load value at a past time; the first noise smoothing correspondence predicts the current load value using a weighted average of the prediction errors of the load values; the first echo correspondence predicts the current value using the load value at the same time in the previous cycle; and the first burst noise smoothing correspondence corrects the current prediction using the prediction error at the same time in the previous cycle.

[0030] Furthermore, the hourly dynamic load forecast values ​​for the entire year are statistically analyzed to obtain the minimum load value, peak load value, probability density function, and cumulative distribution function, specifically including:

[0031] Statistical analysis of hourly dynamic load forecasts throughout the year yields minimum and peak load values.

[0032] The frequency of the hourly dynamic load forecast values ​​throughout the year is counted, and the hourly dynamic load forecast values ​​throughout the year are divided into 24 subsets that contain parts of the hourly dynamic load forecast values ​​throughout the year according to the solar terms.

[0033] Interval statistics are performed on the hourly dynamic load forecast values ​​for the whole year within the subset;

[0034] Set the span of the annual hourly dynamic load forecast interval to ΔL, and count the frequency Nj ​​of the annual hourly dynamic load forecast falling in the first interval [Lj, Lj+ΔL].

[0035] The probability density function is obtained by statistically analyzing the probability of the first load value Lj occurring.

[0036] The cumulative distribution function is obtained by statistically analyzing the total probability of the first load value Lj occurring.

[0037] Furthermore, the unit configuration strategy includes base load capacity;

[0038] The optimization analysis of the minimum load value, the peak load value, the probability density function and the cumulative distribution function specifically includes: selecting the curve interval of the cumulative distribution function, and the dynamic load prediction value corresponding to the curve interval is used as the first base load capacity;

[0039] The first base load capacity is compared with the minimum load value. If the first base load capacity is greater than the minimum load value, the first base load capacity is selected as the second base load capacity. If the first base load capacity is less than the minimum load value, the first base load capacity is adjusted to obtain the second base load capacity, so that the second base load capacity is greater than the minimum load value.

[0040] Determine whether the second base load capacity meets the minimum load rate standard. If the second base load capacity meets the minimum load rate standard, then select the second base load capacity as the base load capacity. If the second base load capacity does not meet the minimum load rate standard, then reselect the curve interval of the cumulative distribution function.

[0041] Furthermore, the minimum load factor standard is calculated from the minimum load value and the second base load capacity, specifically including:

[0042] Minimum load factor standard = minimum load value / second base load capacity

[0043] Furthermore, the unit configuration strategy also includes peak-shaving units;

[0044] The selection of peak-shaving units is determined by the capacity of the peak-shaving units, which is calculated using the base load capacity and the peak load value.

[0045] In another aspect, the present invention provides an air conditioning system equipped with the aforementioned air conditioning system optimization method based on the time sequence characteristics of the twenty-four solar terms, comprising:

[0046] The data acquisition module is used to acquire historical load-multidimensional meteorological datasets containing labels for the 24 solar terms;

[0047] The load forecasting module constructs a meteorological-solar term coupled load forecasting model and obtains hourly dynamic load forecasts for the whole year through the meteorological-solar term coupled load forecasting model;

[0048] The optimization configuration module is used to statistically analyze the hourly dynamic load forecast values ​​for the whole year and obtain the minimum load value, peak load value, probability density function and cumulative distribution function; perform optimization analysis on the minimum load value, the peak load value, the probability density function and the cumulative distribution function; and obtain the unit configuration strategy.

[0049] The control execution module receives the unit configuration policy and executes it.

[0050] In another aspect, the present invention provides an electronic product including the above-described air conditioning system for controlling air conditioning.

[0051] In another aspect, the present invention provides a computer-readable storage medium for storing a computer program that, when executed by a processor, implements the above-described air conditioning system optimization method based on the timing characteristics of the twenty-four solar terms.

[0052] In another aspect, the present invention provides a computer program product, wherein the computer program product stores a program or instructions, which, when executed by a processor, implement the above-described air conditioning system optimization method based on the timing characteristics of the twenty-four solar terms.

[0053] As can be seen from the above solutions, the advantages of the present invention are:

[0054] By decomposing historical load-multidimensional meteorological datasets tagged with the 24 solar terms and capturing the seasonal fluctuations of historical load data to obtain a time series model, and further integrating the 24 solar terms and weather type characteristics into the solar term time series model, a meteorological-solar term coupled load prediction model is obtained. This model can quantify load disturbances under different weather conditions and solar terms, improving the accuracy of load prediction and, consequently, the rationality of air conditioning system configuration. Using the SARIMA model combined with the variables of the 24 solar terms and weather type characteristics, a meteorological-solar term coupled load prediction model is constructed, outputting hourly dynamic load prediction values ​​for the entire year, thus improving the accuracy of load prediction. This approach can solve the problems of low load prediction accuracy, unreasonable equipment configuration, and coarse operation control in existing technologies. Attached Figure Description

[0055] Figure 1 A flowchart illustrating an air conditioning system optimization method based on the temporal characteristics of the 24 solar terms provided by this invention;

[0056] Figure 2 A flowchart illustrating the process of obtaining a historical load-multidimensional meteorological dataset containing labels for the 24 solar terms;

[0057] Figure 3A flowchart illustrating the process of constructing a meteorological-solar term coupled load prediction model using a historical load-multidimensional meteorological dataset containing labels of the 24 solar terms;

[0058] Figure 4 A flowchart illustrating the process of calculating hourly dynamic load forecasts for the entire year and obtaining minimum load, peak load, probability density function, and cumulative distribution function;

[0059] Figure 5 This is a flowchart illustrating another air conditioning system optimization method based on the time sequence characteristics of the 24 solar terms.

[0060] Figure 6 A flowchart for unit configuration strategy based on cumulative distribution function;

[0061] Figure 7 A box plot of load distribution based on the 24 solar terms;

[0062] Figure 8 This is a schematic diagram of a typical daily load curve model;

[0063] Figure 9 A schematic diagram illustrating the load characteristics under various meteorological conditions throughout the 24 solar terms in 2025;

[0064] Figure 10 This is a configuration optimization diagram based on the probability distribution of all operating conditions and the timing of the twenty-four solar terms;

[0065] Figure 11 This is another configuration optimization diagram based on the probability distribution of all operating conditions and the timing of the twenty-four solar terms;

[0066] Figure 12 This is a schematic diagram of an air conditioning system.

[0067] Figure 13 This is a schematic diagram of another air conditioning system;

[0068] Figure 14 This is a schematic diagram of the structure of an electronic product.

[0069] In the attached figures, the following labels are used:

[0070] 10-Air conditioning system;

[0071] 11-Data Acquisition Module;

[0072] 111 - Data Acquisition Module;

[0073] 112 - Data Cleaning Module;

[0074] 113 - Slicing and Marking Module;

[0075] 114 - Load forecasting module;

[0076] 12-Load Forecasting Module;

[0077] 121-Decomposition Module;

[0078] 122 - Seasonal variable embedding module;

[0079] 123 - Weather Feature Fusion Module;

[0080] 124 - Quantification module for the scissors difference effect between sunny and rainy weather;

[0081] 125 - Dynamic load forecast generation module;

[0082] 13-Optimize configuration module;

[0083] 14-Control Execution Module;

[0084] 131 - Load frequency statistics module;

[0085] 132 - Curve Calculation Module;

[0086] 133 - Optimization Analysis Module;

[0087] 134 - Configuration strategy output module;

[0088] 20-Electronic products. Detailed Implementation

[0089] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments to further understand the purpose, solution and effect of the present invention, but it is not intended to limit the scope of protection of the appended claims.

[0090] References to "embodiment," "another embodiment," "this embodiment," etc., in the specification refer to embodiments that may include specific features, structures, or characteristics, but not every embodiment must include these specific features, structures, or characteristics. Furthermore, such expressions do not refer to the same embodiment. Moreover, when describing specific features, structures, or characteristics in conjunction with embodiments, whether or not explicitly described, it is indicated that incorporating such features, structures, or characteristics into other embodiments is within the knowledge of those skilled in the art.

[0091] The specification and subsequent claims use certain terms to refer to specific components or parts. Those skilled in the art will understand that users or manufacturers may use different names or terms to refer to the same component or part. This specification and claims do not distinguish components or parts by differences in name, but rather by differences in function. The terms "comprising" and "including" used throughout the specification and claims are open-ended and should be interpreted as "including but not limited to". Furthermore, the term "connection" here includes any direct and indirect electrical connection means. Indirect electrical connection means include connections via other means.

[0092] Figures 1 to 11 The air conditioning system optimization method based on the time sequence characteristics of the 24 solar terms provided in this embodiment includes:

[0093] S1: Obtain a historical load-multidimensional meteorological dataset containing labels for the 24 solar terms;

[0094] S2: Construct a meteorological-solar term coupled load prediction model using a historical load-multidimensional meteorological dataset containing labels of the 24 solar terms;

[0095] S3: Input the historical load-multidimensional meteorological dataset containing the 24 solar terms labels into the meteorological-solar term coupled load prediction model to obtain the hourly dynamic load prediction value for the whole year;

[0096] S4: Statistically predict hourly dynamic load values ​​throughout the year and obtain minimum load value, peak load value, probability density function and cumulative distribution function;

[0097] S5: Perform optimization analysis on the minimum load value, peak load value, probability density function and cumulative distribution function to obtain the unit configuration strategy.

[0098] A historical load-multidimensional meteorological dataset with labels for the 24 solar terms was acquired, and a meteorological-solar term coupled load prediction model was constructed using this dataset. Considering the continuity and stages of climate change, and combining the unique climate patterns corresponding to the 24 solar terms, the model can quantify load disturbances under different weather conditions and solar terms. By inputting the historical load-multidimensional meteorological dataset with labels for the 24 solar terms into the model, hourly dynamic load prediction values ​​for the entire year were obtained, improving the accuracy of load prediction. The hourly dynamic load prediction values ​​for the entire year were statistically analyzed to obtain the minimum load value, peak load value, probability density function, and cumulative distribution function. An optimization analysis was conducted to select appropriate unit configuration strategies. Based on the highly accurate hourly dynamic load prediction values ​​for the entire year, optimization analysis ensured the selection of a unit configuration with reasonable overall energy efficiency, improving the rationality of air conditioning system configuration.

[0099] In one embodiment, such as Figure 2 As shown, obtaining the historical load-multidimensional meteorological dataset containing labels for the 24 solar terms includes:

[0100] S11: Collect historical load data and local meteorological data;

[0101] S12: Preprocess historical load data and local meteorological data to obtain the required historical load data and local meteorological data;

[0102] In one embodiment, the collected historical load data and local meteorological data are standardized. Before data cleaning, it is necessary to determine the type of data problem, identify and process errors, inconsistencies, missing, duplicate or invalid records in the original data, and improve the accuracy and consistency of the collected historical load data and local meteorological data.

[0103] S13: Construct the required historical load data and required local meteorological data into a multi-dimensional meteorological-load time series based on the annual time axis;

[0104] S14: The multidimensional meteorological-load time series is preprocessed using a seasonal autoregressive integral moving average model. The entire year time axis is mapped to the time series slices of the twenty-four solar terms and labeled to obtain a historical load-multidimensional meteorological dataset with labels of the twenty-four solar terms.

[0105] By slicing and labeling multidimensional meteorological-load time series based on the annual time axis, a historical load-multidimensional meteorological dataset with labels of the 24 solar terms is obtained. Combined with the unique climate patterns corresponding to the 24 solar terms, the continuity and stage of climate change are considered, so that the meteorological-solar term coupled load prediction model can quantify load disturbances under different meteorological conditions and different solar terms, improve the accuracy of load prediction, and further improve the rationality of air conditioning system configuration.

[0106] In one embodiment, local meteorological data includes data such as temperature, humidity, solar radiation, and weather type.

[0107] In one embodiment, the historical load data includes indoor environmental data, which mainly calculates the loads of human beings, lighting, equipment heat dissipation, etc., and is a part of the historical load data.

[0108] In one embodiment, SARIMA (Seasonal Autoregressive Integral Moving Average Model) is used to preprocess the multidimensional meteorological-load time series based on the annual time axis, and the annual time axis is mapped to the twenty-four solar terms time series slices.

[0109] In one embodiment, a meteorological-solar term coupled load prediction model is constructed using a historical load-multidimensional meteorological dataset containing labels for the 24 solar terms, such as... Figure 3 As shown, it specifically includes:

[0110] S21: Detect the stationarity of a historical load-multidimensional meteorological dataset containing labels for the 24 solar terms;

[0111] S22: Using a seasonal autoregressive integral moving average model, the historical load data of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels is decomposed into three parts: trend, seasonality, and residual. The seasonal fluctuations of the historical load data are captured to obtain a time series model.

[0112] S23: Embed the 24 solar terms as seasonal variables into the time series model to obtain the solar term time series model;

[0113] S24: Extract weather type features from the historical load-multidimensional meteorological dataset containing the 24 solar terms labels, and integrate the weather type features into the solar term time series model to obtain the meteorological-solar term coupled load prediction model.

[0114] By decomposing historical load-multidimensional meteorological datasets tagged with the 24 solar terms and capturing the seasonal fluctuations of historical load data to obtain a time series model, and further integrating the 24 solar terms and weather type characteristics into the solar term time series model, a meteorological-solar term coupled load prediction model is obtained. This model can quantify load disturbances under different weather conditions and solar terms, improving the accuracy of load prediction and, consequently, the rationality of air conditioning system configuration. Using the SARIMA model combined with the variables of the 24 solar terms and weather type characteristics, a meteorological-solar term coupled load prediction model is constructed, outputting hourly dynamic load prediction values ​​for the entire year, thus improving the accuracy of load prediction. This approach can solve the problems of low load prediction accuracy, unreasonable equipment configuration, and coarse operation control in existing technologies.

[0115] In one embodiment, the meteorological-solar term coupled load forecasting model embeds the 24 solar terms as seasonal variables into the time series model. For each solar term, load characteristics under different weather types are extracted, the nonlinear disturbance of load by different weather under the same solar term is quantified, and the typical daily load curves corresponding to the 24 solar terms are clustered to generate hourly dynamic load forecasts for the whole year, ensuring the accuracy of hourly dynamic load forecasts for the whole year.

[0116] In one embodiment, load characteristics refer to the characteristics exhibited by the air conditioning system under a specific combination of solar term (e.g., Great Heat) and weather type (e.g., sunny day). These are a set of mathematical indicators describing the shape, amplitude, volatility, and responsiveness of the load curve to the environment. For example: Baseline characteristics: The average daily load curve for the "Great Heat" solar term (including the average of all the above indicators). Sunny day characteristics: Compared to the baseline, the peak load is 20% higher, the peak time is 1 hour earlier, and the ramp-up rate is faster. Rainy day characteristics: Compared to the baseline, the peak load is 30% lower, the peak-to-valley difference is smaller, and the latent heat load proportion increases.

[0117] In one embodiment, different weather types include: sunny, cloudy, rain, thunder, etc.

[0118] In one embodiment, the solar term time series model can calculate the corresponding typical daily load pattern for each solar term, obtain the correspondence between the twenty-four solar terms and the typical daily load pattern, which is the second correspondence, and provide the second correspondence as input features to the meteorological-solar term coupled load prediction model to improve the accuracy of the meteorological-solar term coupled load prediction model in generating hourly dynamic load prediction values ​​for the whole year.

[0119] In one embodiment, the solar term time series model can calculate the corresponding typical daily load pattern for each solar term, obtain the correspondence between the twenty-four solar terms and the typical daily load pattern, which is the second correspondence, and the typical daily load curve clustering. The second correspondence and the typical daily load curve clustering are provided as input features to the meteorological-solar term coupled load prediction model to improve the accuracy of the meteorological-solar term coupled load prediction model in generating hourly dynamic load prediction values ​​for the whole year.

[0120] In one embodiment, the stationarity of a historical load-multidimensional meteorological dataset containing the 24 solar terms is detected using the SARIMA (Seasonal Autoregressive Integral Moving Average Model) algorithm via a unit root test (such as the ADF test).

[0121] In one embodiment, the solar term time series model can calculate the corresponding typical daily load pattern for each solar term, obtain the correspondence between the twenty-four solar terms and the typical daily load pattern, which is the second correspondence, and the typical daily load curve clustering. The second correspondence or the typical daily load curve clustering is provided as input features to the meteorological-solar term coupled load prediction model to improve the accuracy of the meteorological-solar term coupled load prediction model in generating hourly dynamic load prediction values ​​for the whole year.

[0122] In one embodiment, the meteorological-solar term coupled load prediction model includes the correlation and correspondence between the twenty-four solar terms, weather type characteristics, and historical load data, namely the third correspondence. The third correspondence can automatically align "hourly load" and "daily weather" and automatically identify which solar term each day belongs to.

[0123] In one embodiment, the meteorological-solar term coupled load prediction model can also output the weather load scissor difference, which can quantify the differential impact of different weather types on load characteristic data during the same solar term, and can also be called the "sunny-rainy scissor difference" effect.

[0124] In one embodiment, the time series model includes a first inertial correspondence, a first noise smoothing correspondence, a first echo correspondence, and a first sudden noise smoothing correspondence. Specifically, it includes: decomposing historical load data of a historical load-multidimensional meteorological dataset containing 24 solar term labels into a trend part, a seasonal part, and a residual part; using the Akaike Information Criterion to select autoregressive parameters, moving average parameters, seasonal autoregressive parameters, and seasonal moving average parameters in the seasonal part; the autoregressive parameters reflect the first inertial correspondence; the moving average parameters reflect the first noise smoothing correspondence; the seasonal autoregressive parameters reflect the first echo correspondence; and the seasonal moving average parameters reflect the first sudden noise smoothing correspondence.

[0125] In one embodiment, the SARIMA (Seasonal Autoregressive Integral Moving Average Model) algorithm is used to decompose the historical load data of the historical load-multidimensional meteorological dataset containing the 24 solar terms into three parts: trend, seasonality, and residuals, to capture the seasonal fluctuations of the load. Then, the optimal autoregressive parameter AR, moving average parameter MA, seasonal autoregressive parameter seasonal AR, and seasonal moving average parameter seasonal MA are selected using criteria such as AIC (Akaike Information Criterion).

[0126] In one embodiment, the trend refers to the long-term upward, downward, or stable trend of the data over a relatively long period. This reflects non-periodic long-term changes. For example, as office building occupancy increases year by year, or as equipment ages and energy efficiency decreases, the base load of air conditioning may slowly increase year by year. The SARIMA model requires the data to be "stationary," and extracting the trend and differencing it is to eliminate the effects of such long-term growth or decline, so that the load data fluctuates around the mean.

[0127] In one embodiment, the seasonal component refers to periodic fluctuations in the data that repeat over time with a fixed frequency. This invention particularly emphasizes the "Twenty-Four Solar Terms," ​​using them as a scale for seasonal cycles to capture the periodic patterns of specific time nodes such as "Start of Summer" and "Great Heat." This helps the model remember what happened during "last year's Great Heat" or "yesterday at noon," thereby predicting the situation during "this year's Great Heat" or "today at noon."

[0128] In one embodiment, the residual refers to the random fluctuations remaining after removing "trends" and "seasonality." It represents sudden, unpredictable factors that cannot be explained by any pattern. Examples include a sudden thunderstorm causing a sharp drop in temperature, a large, impromptu meeting leading to a surge in attendees, or sensor measurement errors. Ideally, the residual should be "white noise" (random numbers with no discernible pattern). If patterns are still discernible in the residual, it indicates that the model has not been fully extracted. SARIMA's goal is to make the remaining residual as small and random as possible.

[0129] In one embodiment, the first inertial correspondence predicts the current load value using the load value at a past time; the first noise smoothing correspondence predicts the current load value using a weighted average of the prediction errors of the load values; the first echo correspondence predicts the current value using the load value at the same time in the previous cycle; and the first burst noise smoothing correspondence corrects the current prediction using the prediction error at the same time in the previous cycle.

[0130] In one embodiment, the hourly dynamic load forecast values ​​for the entire year are statistically analyzed to obtain the minimum load value, peak load value, probability density function, and cumulative distribution function, such as... Figure 4 As shown, it specifically includes:

[0131] S41: Calculate the hourly dynamic load forecast values ​​for the whole year and obtain the minimum load value and peak load value;

[0132] S42: Count the frequency of the hourly dynamic load forecast values ​​throughout the year and divide the hourly dynamic load forecast values ​​throughout the year into 24 subsets that contain some of the hourly dynamic load forecast values ​​throughout the year according to the solar terms;

[0133] S43: Perform interval statistics on the hourly dynamic load forecast values ​​for the whole year within the subset;

[0134] S44: Set the interval span of the annual hourly dynamic load forecast values ​​to ΔL, and count the annual hourly dynamic load forecast values ​​falling within the first interval [L]. j ,L j The frequency N of +ΔL] j ;

[0135] S45: Obtain the probability density function by calculating the probability of the first load value Lj occurring;

[0136] S46: The cumulative distribution function is obtained by statistically analyzing the total probability of the occurrence of the first load value Lj.

[0137] In one embodiment, such as Figure 5As shown, the process of the air conditioning system optimization method based on the time series characteristics of the 24 solar terms is as follows: First, historical load data and local meteorological data are collected. Then, the data is standardized and cleaned. Next, the 24 solar terms are sliced ​​and labeled to obtain a historical load-multidimensional meteorological dataset containing the 24 solar terms labels. The historical load-multidimensional meteorological dataset containing the 24 solar terms labels is decomposed using the SARIMA model and seasonal variables of the solar terms are embedded, weather type features are fused, and the sunny and rainy scissors difference effect is quantified. The meteorological-solar term coupled load prediction model is constructed using the historical load-multidimensional meteorological dataset containing the 24 solar terms labels, and the hourly dynamic load prediction value for the whole year is generated. Then, the load frequency statistics (binning) of the hourly dynamic load prediction value for the whole year are performed. The PDF probability density and CDF cumulative distribution are further calculated. The two-way constraint optimization algorithm is used to obtain the unit configuration strategy, complete the configuration capacity optimization, and output the configuration scheme. The configuration scheme includes the baseload unit capacity, peak-shaving unit capacity, and extreme condition absorption strategy.

[0138] In one embodiment, such as Figure 6 As shown, the unit configuration strategy process based on Cumulative Probability (CDF) is as follows: Input the hourly dynamic load forecast values ​​for the whole year, and statistically calculate the minimum load value Lmin and the peak load value Lmax, generating a CDF curve. Then, based on the CDF golden section interval, initially select the base load capacity C_base, and determine whether C_base > Lmin. If C_base > Lmin, determine whether the minimum load rate > 60%. Otherwise, increase C_base. If the minimum load rate > 60%, determine the base load capacity, with the goal of 80% full load time. Otherwise, readjust, and again initially select the base load capacity C_base based on the CDF golden section interval, and repeat the above operation to complete the base load unit selection. After determining the base load capacity, calculate the remaining fluctuating load L_rem = Lmax - C_base to determine the peak-shaving unit capacity C_peak, and select the appropriate type. For wet May / thunderstorms, variable frequency / magnetic levitation can be selected to complete the peak-shaving unit selection. Extreme peak handling (CDF>99.5%) utilizes the building's thermal inertia to naturally absorb the final output unit configuration strategy.

[0139] In one embodiment, the probability density function (PDF) represents the number of hours (frequency) the air conditioning system operated within a specific load range, or the percentage of the total annual time (probability / frequency). This corresponds to the "probability" column in the table, indicating "how likely the load is to operate between 10% and 20%." (For example, 5% of the time the system operates within this range). The cumulative distribution function, also known as the cumulative probability distribution (CDF), represents the total probability that the load is less than or equal to a certain value (or the upper limit of the range). It is obtained by accumulating the PDF row by row downwards from the smallest interval.

[0140] In one embodiment, the minimum load value and peak load value are obtained by statistically analyzing historical load data.

[0141] In one embodiment, the set of hourly dynamic load forecasts for the whole year is... Divided into 24 subsets These correspond to the twenty-four solar terms, from the Beginning of Spring to Rain Water and Great Cold. Simultaneously, each solar term subset... within Perform interval statistics (Binning). Let the idle span of the load zone be... (For example, 10kW), calculate the load falling within a certain range. frequency and a certain load value The probability of occurrence, a certain load value This can be referred to as the first load value. .

[0142] In one embodiment, the unit configuration strategy includes base load capacity; optimization analysis is performed on the minimum load value, peak load value, probability density function, and cumulative distribution function, specifically including: selecting a curve interval of the cumulative distribution function, and the dynamic load forecast value corresponding to the curve interval as the first base load capacity; comparing the first base load capacity with the minimum load value (Lmin), if the first base load capacity is greater than the minimum load value, then the first base load capacity is selected as the second base load capacity; if the first base load capacity is less than the minimum load value, then the first base load capacity is adjusted to obtain the second base load capacity, so that the second base load capacity is greater than the minimum load value; determining whether the second base load capacity meets the minimum load rate standard, if the second base load capacity meets the minimum load rate standard, then the second base load capacity is selected as the base load capacity, if the second base load capacity does not meet the minimum load rate standard, then the curve interval of the cumulative distribution function is reselected.

[0143] In one embodiment, the optimization analysis can adopt the "minimum load constraint-cumulative probability optimization" strategy, select the load value corresponding to the segment with the largest slope in the CDF curve or the golden section interval (such as 40%-50% of the peak load) as the base load capacity, and verify whether it is greater than the minimum load of the whole year, i.e. the minimum load value, to ensure that the base load unit is in the full load and high efficiency zone for most of the year and does not surge.

[0144] In one embodiment, the minimum load factor criterion is calculated from the minimum load value and the second base load capacity, specifically including:

[0145]

[0146] In one embodiment, the unit configuration strategy also includes peak-shaving units; the selection of peak-shaving units is determined by their capacity, which is calculated using the base load capacity and peak load value. The remaining peak load is calculated using the base load capacity and peak load value, and variable frequency peak-shaving units are configured to handle the remaining peak load above the base load capacity, especially for latent heat load fluctuations during the hot and humid summer season.

[0147] In one embodiment, the specific calculation process for the remaining fluctuating load is as follows: Remaining fluctuating load (L_rem) = Peak load value (Lmax) - Base load capacity (C_base).

[0148] In one embodiment, the unit configuration strategy also includes extreme peak loads; for extreme peak loads with a probability of occurrence below a preset threshold (e.g., 0.5%), they are not included in the installed capacity of the equipment and are absorbed through building thermal inertia. It should be noted that extreme peak loads can also be referred to as extreme operating conditions.

[0149] In one embodiment, taking an office building as an example, the actual application process of the present invention is demonstrated:

[0150] After obtaining the historical load-multidimensional meteorological dataset with 24 solar terms labels, the stationarity of the dataset was tested. A seasonal autoregressive integral moving average (SRM) model was used to decompose the historical load data into trend, seasonality, and residual components, capturing the seasonal fluctuations of the historical load data and obtaining a time series model. During the processing of the historical load-multidimensional meteorological dataset with 24 solar terms labels using the SRM model, the following results were obtained: Figure 7 The load distribution box plot shown is based on the 24 solar terms. (Analysis is needed for accurate translation.) Figure 7 Data analysis revealed significant load fluctuations during high-temperature solar terms: for example, the Summer Solstice and Great Heat solar terms showed high median loads and wide box-shaped ranges, indicating a substantial impact of high temperatures on air conditioning cooling loads and increased load volatility. Low-temperature or transitional solar terms, such as the Winter Solstice and Minor Cold solar terms, exhibited more stable load distributions, suggesting relatively stable user demand during these periods. This example extracted the hourly average load for each solar term and highlighted key data points such as... Figure 8 The daily load curves for four representative solar terms (Spring Equinox, Summer Solstice, Autumn Equinox, and Winter Solstice) are shown. These typical daily curves for solar terms are similar to the daily curve decomposition method commonly used in building HVAC energy consumption modeling. It uses historical load data to extract regular features from intraday load patterns, providing reliable input for load forecasting and HVAC system control strategy design.

[0151] from Figure 8As can be seen, the summer solstice typically sees a significant peak: the load curve reaches its peak between 12:00 and 16:00, corresponding to the highest solar radiation intensity and indoor air-cooling demand. The spring / autumn equinox shows a similar pattern: the load curves for these transitional solar terms are generally relatively stable, with lower intraday peak values. The winter solstice shows lower and less volatile loads: this reflects a lower proportion of cooling load or relatively moderate heating load demand during winter.

[0152] The meteorological-solar term coupled load forecasting model identified the "scissors difference between sunny and rainy weather": under the same "Great Heat" solar term, the predicted load for sunny weather is approximately 950kW, while the predicted load for rainy weather is reduced to 780kW due to the attenuation of solar radiation, a decrease of 17.8%. Based on this, the model was used to correct the load forecasts for "wet May" and the "Great Heat thunderstorm season," thus improving the forecast accuracy.

[0153] Figure 9 The red bars indicate sunny weather, the blue bars indicate rain / thunder, the green bars indicate snow, and the gray bars indicate cloudy / overcast weather. Figure 9 A significant "scissors difference" effect is evident, namely the height difference between the red (sunny) and blue (rainy) columns under the same solar term. In summer (Minor Heat / Major Heat), there is a huge difference (approximately 200kW~300kW) between the average load on sunny days (approximately 900kW+) and the average load on rainy days (approximately 650kW-700kW). This quantifies the thermal disturbance effect of solar radiation on the building. Sensible heat load increases dramatically on sunny days; while on rainy days, although latent heat (dehumidification) load increases, the decrease in sensible heat due to cloud cover and precipitation cooling is more significant, resulting in a lower total load. This "scissors difference" proves that selecting a single "design date" is not precise enough. The system must be able to adjust from 900kW (sunny) to 600kW (rainy) in a short time, verifying the necessity of reasonable unit configuration and variable frequency peak shaving.

[0154] In addition, there is the special operating condition verification of "wet May". Observing the data from "Start of Summer to Grain Full" (May), although the load on sunny days (red bars) is already relatively high, due to the rainy season in May 2025, the actual operation will frequently experience low load conditions as shown in the blue bars. This results in the actual average load in May being lower and more volatile than in previous years. Therefore, the selection logic of base-load units needs to cover the low load on rainy days.

[0155] By collecting 8760 hours of dynamic load data throughout the year, the minimum annual cooling load was found to be 303kW, and the peak load was 1016kW. Traditional solutions typically select turbines based on peak loads or simple averaging, potentially configuring two 500kW units, resulting in one unit operating below 60% load during the transition season. It should be noted that the 8760 hours of dynamic load data throughout the year can refer to hourly dynamic load forecasts or historical load data.

[0156] This embodiment uses dynamic load simulation data from 8760 hours throughout the year and employs a method that couples the probability density function (PDF) and cumulative distribution function (CDF) with the time-series characteristics of the 24 solar terms to perform an optimization analysis of the capacity configuration of the cold and heat source system. It should be noted that the capacity configuration of the cold and heat source system is the same as the unit configuration strategy.

[0157] 1) Data boundaries: The minimum annual cooling load is 303kW, the peak cooling load is 1016kW, and the peak-valley load fluctuation ratio is approximately 3.35 times.

[0158] 2) Optimization objective: Find the optimal capacity point of the baseload chiller to maximize the full-load operation time throughout the year while covering the minimum load, thereby improving the overall system energy efficiency (IPLV).

[0159] Combination Figure 10 , Figure 11 The two figures show that this configuration scheme has significant advantages in "energy efficiency matching under all operating conditions" and "preventing low-load surge".

[0160] from Figure 10 The blue line represents the cumulative load distribution (CDF) throughout the year, and the light-colored area represents the recommended base load range (400-500kW). Figure 10 CDF curve analysis shows that for approximately 64% of the year (mainly the transitional season and winter), the building load is less than 457kW. During this time, only one baseload unit needs to be operated, with its operating load rate within the optimal energy efficiency range of 66.3% to 100%, independently handling the load without the need for frequent start-stop intervention from a second unit. During the remaining 36% of the summer peak season, the load exceeds 457kW. At this time, the baseload unit operates at 100% full load, providing the most stable cooling output, while fluctuations are handled by peak-shaving units. The red arrow indicates the algorithm-recommended baseload: 457kW (45% of peak load), with a cumulative fulfillment rate of 66.2%.

[0161] This configuration strategy ensures that the base-mounted units are always in the high-efficiency zone of medium to high load throughout the year, without any inefficient waste caused by "overloading".

[0162] from Figure 11The box plot analysis shows that the gray dashed line represents the minimum annual load of 10 kW, while the red dashed line indicates a recommended base load of 457 kW. The minimum operating load rate is 66.3%, indicating perfect minimum load matching for the unit. This means that even during the lowest load periods of the year (such as the "Minor Cold" or "Major Cold" solar terms), the base load rate remains at 66.3%, perfectly avoiding the low-load surge zone (typically <30%) and low-efficiency degradation zone (<50%) of centrifugal / screw turbines. In contrast, if an 800 kW unit were used, the minimum load rate would only be 37%, resulting in a significant decrease in energy efficiency.

[0163] =303 / 457≈66.3%

[0164] Furthermore, based on the analysis of the box plots for adaptability to the 24 solar terms, the red dashed line (recommended base load line) precisely divides the load boxes for different seasons. In summer (from the Beginning of Summer to the End of Heat), the load box is entirely above the red line (457kW), indicating that both units must operate simultaneously, with the base load unit operating at full load and the peak load unit responsible for frequency conversion regulation. In winter (from the Beginning of Winter to the Great Cold), the load box is completely below the red line, with its lower edge close to the gray line (303kW), indicating that only the base load unit needs to be operated, and its load rate fluctuates between 66% and 100%, precisely at the peak of the partial load characteristic curve (IPLV) of the frequency conversion / magnetic levitation unit.

[0165] The data (frequency, probability) in Table 1 are based on Figure 10 , Figure 11 The analysis yielded a "64.0% cumulative satisfaction rate" (meaning that 64% of the time load is less than 450kW, which can be satisfied by a single unit; 36% of the time load is greater than 450kW, which requires the operation of two units) which was then used to construct the system.

[0166] Table 1. Statistical Table of Unit Capacity Matching Based on Annual Load Probability Distribution

[0167]

[0168] Based on the above statistical analysis and the "double-peak" characteristic of measured meteorological data in 2025, and considering the high humidity fluctuations during the five-month rainy season from May to September 2025, peak-shaving units need to have an extremely wide adjustment range. The chiller configuration logic needs to be redefined:

[0169] 1) Base load configuration: A 450kW unit (approximately 45% of peak capacity) was selected as the base load unit. This capacity is greater than the minimum load of 303kW (minimum operating load rate of 66%), ensuring that the base load unit operates at the power frequency without surge throughout the year, and operates at 100% full load for 83.6% of the time, achieving energy saving by operating at the power frequency in the high-efficiency range.

[0170] 2) Peak-shaving unit: A 560kW variable frequency unit was selected. This unit only starts when the load exceeds 450kW. It uses variable frequency technology to precisely adjust the remaining 0-566kW fluctuating load, perfectly coping with the sudden load changes during summer thunderstorms. The specific configuration is shown in Table 2 below.

[0171] 3) Extreme operating conditions: For the extremely low probability (<0.5%) extreme peak of 1010kW-1016kW, the thermal inertia of the building envelope is used to absorb the load naturally without increasing the installed capacity.

[0172] Table 2 Annual Unit Matching Scheme

[0173]

[0174] To verify the superiority of the "dual-unit scheme (450kW fixed frequency + 560kW variable frequency)" in this embodiment, this study, in conjunction with engineering practice, conducted a multi-dimensional quantitative comparison with the traditional "three-unit cascade scheme (450kW fixed frequency + 280kW fixed frequency + 280kW variable frequency)".

[0175] 1) Scheme definition:

[0176] A. Implementation Scheme (Scheme A): A configuration of "1 base load (fixed frequency) + 1 wide-range regulation (variable frequency)" is adopted. This scheme is calculated through an algorithm and aims to achieve the optimal balance between a minimalist system architecture and total lifecycle cost (LCC).

[0177] B. Comparative Scheme (Scheme B): This scheme uses a configuration of "1 base load (fixed frequency) + 1 intermediate stage (fixed frequency) + 1 fine adjustment (variable frequency)". This scheme aims to achieve theoretically maximum energy efficiency by increasing the proportion of fixed frequency units in operation.

[0178] 2) The key performance indicator comparison table is as follows:

[0179] Table 3 Comparison of key indicators between the two schemes

[0180]

[0181] 3) In-depth comparative analysis

[0182] (1) Economic efficiency and marginal benefit analysis

[0183] According to the calculations based on the algorithm of this invention, although Scheme B increases the full-load operation time of the power frequency unit by splitting the unit, it can theoretically reduce the operating energy consumption by about 1.5% to 2.5% per year. However, adding a third unit and its supporting facilities (pumps, towers, pipelines, and automation) will lead to a significant increase in the initial investment (CAPEX).

[0184] Based on the principle of diminishing marginal returns, Option B's increased payback period is expected to exceed 8 years, far exceeding the expectations of typical commercial projects. In contrast, Option A successfully locked in the "inflection point" of return on investment, representing a more economically viable globally optimal solution.

[0185] (2) Performance coverage of frequency conversion technology

[0186] Traditional design principles hold that "full-load efficiency at line frequency is higher than partial-load efficiency at variable frequency," a principle that holds true for older units. However, considering the characteristics of modern magnetic levitation or variable-frequency centrifugal chillers, their COP (Coefficient of Performance) in the 40%–75% load range is often higher than that at 100% full load. Therefore, the 560kW variable-frequency chiller in Scheme A, when operating under medium to low loads (e.g., 200–400kW), is actually not less efficient than the 280kW line-frequency chiller in Scheme B.

[0187] (3) System reliability and control robustness

[0188] Engineering reliability theory indicates that the more system components there are, the higher the probability of failure. Option B introduces more switching points, which can easily lead to frequent start-ups, shutdowns, or oscillations, especially when load fluctuations pass through the critical points of 280kW and 730kW. Option A, on the other hand, has only one switching logic point (450kW), and the variable frequency unit has a wide-range stepless adjustment capability of 10%~100%, enabling smooth transitions across all operating conditions.

[0189] 4) Optimal Conclusion

[0190] In summary, although the three-unit scheme (Scheme B) has some reference value in terms of a single energy efficiency dimension, the optimization algorithm proposed in this invention, after comprehensively considering the initial investment, operation and maintenance complexity, land space and total life cycle cost, determines that the two-unit scheme (Scheme A) is the optimal configuration under the background of this embodiment.

[0191] This further verifies the advanced nature of the selection method described in this patent: it is not a simple superposition of energy efficiency, but a multi-objective global optimization based on probability distribution and equipment characteristics.

[0192] In one embodiment, such as Figure 12As shown, an air conditioning system 10 is also provided, which is equipped with the above-mentioned air conditioning system optimization method based on the time sequence characteristics of the twenty-four solar terms, including: a data acquisition module 11, used to acquire historical load-multidimensional meteorological datasets containing twenty-four solar term tags; a load prediction module 12, used to construct a meteorological-solar term coupled load prediction model, and obtain the hourly dynamic load prediction value for the whole year through the meteorological-solar term coupled load prediction model; an optimization configuration module 13, used to statistically analyze the hourly dynamic load prediction value for the whole year and obtain the minimum load value, peak load value, probability density function and cumulative distribution function; perform optimization analysis on the minimum load value, peak load value, probability density function and cumulative distribution function; and obtain the unit configuration strategy; and a control execution module 14, used to receive the unit configuration strategy and execute it.

[0193] In one embodiment, such as Figure 13 As shown, the data acquisition module 11 includes: a data acquisition module 111, a data cleaning module 112, and a slicing and labeling module 113; the load forecasting module 12 includes: a decomposition module 121, a seasonal variable embedding module 122, a weather feature fusion module 123, a sunny / rainy scissor difference effect quantification module 124, and a dynamic load forecast value generation module 125; the optimization configuration module 13 includes: a load frequency statistics module 131, a curve calculation module 132, an optimization analysis module 133, and a configuration strategy output module 134.

[0194] The process of the air conditioning system optimization method based on the time-series characteristics of the 24 solar terms is as follows: The data acquisition module 111 collects historical load data and local meteorological data. Then, the data cleaning module 112 performs data standardization and cleaning. Next, the slicing and labeling module 113 performs time-series slicing and labeling of the 24 solar terms to ensure that the data acquisition module 11 completes the acquisition of historical load-multidimensional meteorological datasets containing the 24 solar term labels. The decomposition module 121 decomposes the historical load-multidimensional meteorological dataset containing the 24 solar term labels using the SARIMA model. The seasonal variable embedding module 122, the weather feature fusion module 123, and the sunny-rainy scissor difference effect quantification module 124 embed the seasonal variables of the solar terms, fuse the weather type features, and quantify the sunny-rainy scissor difference effect. The dynamic load prediction value generation module 125 generates hourly dynamic load prediction values ​​for the whole year, so that the load prediction module 12 can use the historical load-multidimensional meteorological dataset containing the 24 solar term labels to construct a meteorological-solar term coupled load prediction model and generate hourly dynamic load prediction values ​​for the whole year. The load frequency statistics module 131 performs load frequency statistics (Binning) on ​​the hourly dynamic load forecast values ​​throughout the year. The curve calculation module 132 further calculates the PDF probability density and CDF cumulative distribution. The optimization analysis module 133 uses a two-way constraint optimization algorithm to obtain the unit configuration strategy, completes the configuration capacity optimization, and outputs the configuration scheme through the configuration strategy output module 134. The configuration scheme includes the base load unit capacity, peak shaving unit capacity, and extreme condition absorption strategy.

[0195] This system embodiment can be implemented in conjunction with the implementation methods described above. The relevant technical details mentioned in the implementation methods of the above embodiments remain valid in the implementation methods of this device embodiment, and will not be repeated here to avoid repetition.

[0196] In one embodiment, an electronic product 20 is also provided, such as... Figure 14 As shown, the electronic product 20 includes the aforementioned air conditioning system 10 for controlling the air conditioner.

[0197] In one embodiment, the electronic product 20 may be configured as a robot, intelligent agent, unmanned device, microprocessor, server, etc. The electronic product 20 includes an air conditioning system 10 and is used to execute the received unit configuration strategy and control the air conditioning.

[0198] In one embodiment, the electronic product 20 can be placed in scenarios such as automobiles, trains, mobile large vehicles, office parks, public buildings, commercial buildings, industrial production workshops, office buildings, and residences to control air conditioning and optimize the configuration of the air conditioning system.

[0199] This electronic product embodiment can be implemented in conjunction with the implementation methods described above. The relevant technical details mentioned in the implementation methods of the above embodiments remain effective in the implementation methods of this device embodiment, and will not be repeated here to avoid repetition.

[0200] This invention also provides a computer-readable storage medium for storing a computer program that, when executed by a processor, implements the aforementioned air conditioning system optimization method based on the timing characteristics of the 24 solar terms, and achieves the same technical effect.

[0201] This invention also provides a computer program product, which stores a program or instructions. When the program or instructions are executed by a processor, they implement the steps of the above-described air conditioning system optimization method based on the timing characteristics of the 24 solar terms, and achieve the same technical effect.

[0202] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for selecting air conditioning systems based on the temporal characteristics of the 24 solar terms, used for load prediction and equipment selection of HVAC systems, characterized in that, Include: Historical load data and local meteorological data were collected and constructed into a multidimensional meteorological-load time series based on an annual time axis. The multidimensional meteorological-load time series was sliced ​​and labeled to obtain a historical load-multidimensional meteorological dataset containing labels of the 24 solar terms. The twenty-four solar terms are embedded as seasonal variables into the time series model to obtain the solar term time series model. Weather type characteristics are then incorporated into the solar term time series model to construct a meteorological-solar term coupled load prediction model. The historical load-multidimensional meteorological dataset containing the 24 solar terms labels is input into the meteorological-solar term coupled load prediction model to obtain the hourly dynamic load prediction value for the whole year. The hourly dynamic load forecast values ​​for the whole year are statistically analyzed to obtain the minimum load value, peak load value, probability density function and cumulative distribution function; The minimum load value, the peak load value, the probability density function, and the cumulative distribution function are analyzed using a "minimum load constraint-cumulative probability optimization" strategy. Specifically, Select the curve interval of the cumulative distribution function, and use the dynamic load forecast value corresponding to the curve interval as the first base load capacity; The first base load capacity is compared with the minimum load value. If the first base load capacity is greater than the minimum load value, the first base load capacity is selected as the second base load capacity. If the first base load capacity is less than the minimum load value, the first base load capacity is adjusted to obtain the second base load capacity, so that the second base load capacity is greater than the minimum load value. Determine whether the second base load capacity meets the minimum load rate standard. If the second base load capacity meets the minimum load rate standard, then select the second base load capacity as the base load capacity. If the second base load capacity does not meet the minimum load rate standard, then reselect the curve interval of the cumulative distribution function. The remaining peak load is obtained by using the base load capacity and the peak load value, and the variable frequency peak-shaving unit is configured to undertake the remaining peak load above the base load capacity; The resulting configuration scheme includes the selection of unit configurations for baseload units and peak-shaving units.

2. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 1, characterized in that, Obtaining the historical load-multidimensional meteorological dataset containing the 24 solar terms labels includes: Collect historical load data and local meteorological data; The historical load data and the local meteorological data are preprocessed to obtain the required historical load data and the required local meteorological data. The required historical load data and the required local meteorological data are constructed into a multidimensional meteorological-load time series based on an annual time axis. The multidimensional meteorological-load time series was preprocessed using a seasonal autoregressive integral moving average model, and the annual time axis was mapped to the time series slices of the twenty-four solar terms and labeled to obtain the historical load-multidimensional meteorological dataset containing the labels of the twenty-four solar terms.

3. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 1, characterized in that, The meteorological-solar term coupled load prediction model is constructed using the historical load-multidimensional meteorological dataset containing the 24 solar terms labels, specifically including: The stationarity of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels was detected; The historical load data of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels is decomposed into three parts: trend, seasonality, and residual by using a seasonal autoregressive integral moving average model to capture the seasonal fluctuations of the historical load data and obtain a time series model. By embedding the 24 solar terms as seasonal variables into the time series model, a solar term time series model is obtained. Weather type features are extracted from the historical load-multidimensional meteorological dataset containing the 24 solar terms labels, and the weather type features are integrated into the solar term time series model to obtain the meteorological-solar term coupled load prediction model.

4. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 3, characterized in that, The solar term time series model can calculate the corresponding typical daily load pattern for each solar term, obtain the correspondence between the twenty-four solar terms and the typical daily load pattern, which is the second correspondence, and provide the second correspondence as input features to the meteorological-solar term coupled load prediction model.

5. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 3, characterized in that, The meteorological-solar term coupled load prediction model includes the correlation and correspondence between the twenty-four solar terms, weather type characteristics, and historical load data, namely the third correspondence. The third correspondence can automatically align "hourly load" and "daily weather" and automatically identify which solar term each day belongs to.

6. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 3, characterized in that, The meteorological-solar term coupled load prediction model can also output the weather load scissor difference, which can quantify the differential impact of different weather types on load characteristic data during the same solar term.

7. The method for selecting an air conditioning system based on the temporal characteristics of the twenty-four solar terms according to claim 3, 4, 5, or 6, is characterized in that, The time series model includes a first inertial correspondence, a first noise smoothing correspondence, a first reverberation correspondence, and a first burst noise smoothing correspondence, specifically including: The historical load data of the historical load-multidimensional meteorological dataset containing the 24 solar terms labels is decomposed into a trend part, a seasonal part, and a residual part. The autoregressive parameter, moving average parameter, seasonal autoregressive parameter, and seasonal moving average parameter in the seasonal part are selected using the Akaike Information Criterion. The autoregressive parameters reflect the first inertial correspondence; The moving average parameters reflect the first noise smoothing correspondence; The seasonal autoregressive parameters reflect the first response correspondence; The seasonal moving average parameter reflects the smooth correspondence of the first burst noise.

8. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 7, characterized in that, The first inertial correspondence predicts the current load value using the load value at a past time; the first noise smoothing correspondence predicts the current load value using a weighted average of the prediction errors of the load values; the first echo correspondence predicts the current value using the load value at the same time in the previous cycle; and the first burst noise smoothing correspondence corrects the current prediction using the prediction error at the same time in the previous cycle.

9. The method for selecting an air conditioning system based on the temporal characteristics of the twenty-four solar terms according to claim 1, 2, 3, 4, 5, or 8, characterized in that, The system statistically analyzes the hourly dynamic load forecast values ​​for the entire year and obtains the minimum load value, peak load value, probability density function, and cumulative distribution function, specifically including: Statistical analysis of hourly dynamic load forecasts throughout the year yields minimum and peak load values. The frequency of the hourly dynamic load forecast values ​​throughout the year is counted, and the hourly dynamic load forecast values ​​throughout the year are divided into 24 subsets that contain parts of the hourly dynamic load forecast values ​​throughout the year according to the solar terms. Interval statistics are performed on the hourly dynamic load forecast values ​​for the whole year within the subset; Set the range of the annual hourly dynamic load forecast value interval to ΔL, and count the annual hourly dynamic load forecast values ​​falling within the first interval [L]. j ,L j The frequency N of +ΔL] j ; Statistical analysis of the first load value L j The probability density function is obtained from the probability of occurrence. Statistical analysis of the first load value L j The total probability of occurrence yields the cumulative distribution function.

10. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 9, characterized in that, The minimum load factor standard is calculated from the minimum load value and the second base load capacity, specifically including: Minimum load rate standard = minimum load value / second base load capacity.

11. The air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms according to claim 9, characterized in that, The unit configuration selection also includes peak-shaving units; The selection of peak-shaving units is determined by the capacity of the peak-shaving units, which is calculated using the base load capacity and the peak load value.

12. An air conditioning system, incorporating the air conditioning system selection method based on the time sequence characteristics of the twenty-four solar terms as described in any one of claims 1 to 11, characterized in that, include: The data acquisition module is used to acquire historical load-multidimensional meteorological datasets containing labels for the 24 solar terms; The load forecasting module constructs a meteorological-solar term coupled load forecasting model and obtains hourly dynamic load forecasts for the whole year through the meteorological-solar term coupled load forecasting model; The optimization configuration module is used to statistically analyze the hourly dynamic load forecast values ​​for the whole year and obtain the minimum load value, peak load value, probability density function, and cumulative distribution function; perform optimization analysis on the minimum load value, peak load value, probability density function, and cumulative distribution function; and obtain the unit configuration selection. The control execution module receives and selects the configuration of the receiving unit and executes it.

13. An electronic product, characterized in that, Includes the air conditioning system as described in claim 12, used for controlling air conditioning.

14. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed by a processor, implements the air conditioning system selection method based on the timing characteristics of the twenty-four solar terms as described in any one of claims 1 to 11.

15. A computer program product, characterized in that, The computer program product stores a program or instruction that, when executed by a processor, implements the air conditioning system selection method based on the timing characteristics of the twenty-four solar terms as described in any one of claims 1 to 11.

Images