Method and system for predicting photo-thermal efficiency of solar heat collection system

Abstract

The invention provides a solar heat collection system photo-thermal efficiency prediction method and system, and relates to the technical field of solar heat utilization, and the method comprises the steps: firstly collecting historical meteorological data and photo-thermal efficiency data, constructing a general data set, and predicting the photo-thermal efficiency through a multivariate nonlinear regression model; then marking the terrain and continuously collecting data, when the solar heat collection system in the same terrain in a fixed region reaches a collection day number threshold value and a quantity threshold value, integrating historical meteorological data, the photo-thermal efficiency and the inclination angle parameter to construct a region data set, and predicting the photo-thermal efficiency by using a random forest model; and calculating a root-mean-square error in real time, and when the number of times exceeds a root-mean-square error threshold value in a specific time window, switching to the multivariate nonlinear regression model for prediction until the random forest model completes new data updating and is re-enabled after a collection day number threshold value is reached, thereby realizing dynamic prediction of the photo-thermal efficiency under different topographic conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of solar thermal utilization, and in particular to a method and system for predicting the photothermal efficiency of a solar thermal collection system. Background Art

[0002] Over the past few decades, solar thermal technology has gradually become an important direction for energy transformation and sustainable development due to its clean and renewable characteristics. However, the solar thermal efficiency of solar thermal systems is complexly affected by weather, climate factors and equipment performance, and its prediction and optimization have always been technical difficulties. Traditional methods mainly rely on manual measurement and theoretical model calculations, which are not only costly, but also lead to large prediction errors due to the inability to effectively deal with the uncertainty of meteorological conditions. For example, the solar thermal efficiency of a linear Fresnel solar collector system needs to be comprehensively evaluated in combination with the efficiency of the reflector, the performance of the collector tube and real-time meteorological parameters. However, in existing technologies, the dynamic coupling relationship of parameters is difficult to accurately model, and data collection is difficult. Especially when the sample size is limited, the generalization ability of the model is significantly reduced.

[0003] In the prior art, patent publication number CN109539596A proposes a method for predicting the photothermal efficiency of solar thermal systems based on a genetic algorithm-optimized generalized regression neural network. By optimizing the smoothing factor of the GRNN and combining it with the global optimization capabilities of the genetic algorithm, this method improves prediction accuracy under small sample data. However, this method still has limitations: First, the input variables only select three indicators: relative humidity, surface temperature, and normal direct radiation intensity, ignoring the nonlinear effects of key factors such as sunshine duration and the tilt angle of the solar thermal system on photothermal efficiency; second, model training relies on the integrity and representativeness of historical meteorological data, which is insufficiently adaptable in extreme weather or new solar thermal system deployment scenarios; third, the lack of a dynamic data update mechanism makes it impossible to adjust the prediction model according to real-time meteorological changes, limiting the stability of long-term predictions.

[0004] The above problems make it difficult for existing prediction methods to meet actual needs in complex meteorological conditions or new application scenarios. For example, the photothermal efficiency responses in high temperature difference areas and low temperature difference areas are significantly different, and the existing models do not perform differentiated modeling for such regional characteristics. Therefore, there is an urgent need for a photothermal efficiency prediction method that can integrate multi-dimensional meteorological parameters, dynamically update model parameters and adapt to complex scenarios to improve the operating efficiency and economy of solar thermal collection systems.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for predicting the photothermal efficiency of a solar thermal collection system, so as to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The method for predicting the photothermal efficiency of a solar thermal collection system includes the following specific steps:

[0009] Step 1: Collect historical meteorological data and the solar thermal efficiency of the solar thermal system at the location of the solar thermal system. The meteorological data includes daily average temperature, daily temperature range, solar radiation intensity, and sunshine duration. Construct a common dataset based on the historical meteorological data and solar thermal efficiency based on the daily average temperature and daily temperature range.

[0010] Step 2: Extract daily average temperature, daily solar radiation intensity, and sunshine duration from each common data set as input variables, and use photothermal efficiency as the output variable to construct a multiple nonlinear regression model. Calculate the sunshine duration using an astronomical formula, and input the daily average temperature, solar radiation intensity, and sunshine duration into the multiple nonlinear regression model to predict photothermal efficiency.

[0011] Step 3: Divide a fixed area and classify the terrain where the solar thermal system is located. When the solar thermal efficiency collected under the same terrain in the fixed area reaches the preset collection day threshold and quantity threshold, the historical meteorological data, solar thermal efficiency, and the tilt angle of the solar thermal system are used to construct a regional dataset.

[0012] Step 4: Extract historical meteorological data, solar thermal efficiency, and tilt angle from the regional dataset, build a random forest model to predict solar thermal efficiency, and calculate the root mean square error (RMS). Set a RMS error threshold. When the number of times the RMS error exceeds the preset RMS error threshold within a preset time window exceeds the preset threshold, switch to a multivariate nonlinear regression model to predict solar thermal efficiency.

[0013] Step 5: After triggering the switch of the multivariate nonlinear regression model, when the cumulative number of days reaches the collection day threshold, the random forest model is re-enabled for prediction.

[0014] Furthermore, we collected historical meteorological data and the solar thermal efficiency of the solar thermal system in the location of the solar thermal system. We obtained the optical efficiency and heat loss coefficient by consulting the standard product specifications of the solar thermal system. To address the lack of measured data on solar thermal efficiency in some areas, we used a physical model of the solar thermal collector combined with meteorological data to calculate the solar thermal efficiency instead.

[0015] Calculate the average collector temperature:

[0016]

[0017] Where, Tm Indicates the average collector temperature, T max Indicates the maximum daily temperature, T min Indicates the daily minimum temperature;

[0018] Constructing a mathematical calculation model for photothermal efficiency:

[0019]

[0020] Where η0 represents the optical efficiency, a1 and a2 represent the heat loss coefficient, and T a represents the daily average temperature, and G represents the daily solar radiation intensity, that is, the average solar radiation intensity of the day.

[0021] Furthermore, the method for constructing a general data set based on the daily average temperature and daily temperature difference for historical meteorological data and solar thermal efficiency is as follows:

[0022] Set the daily average temperature interval to 5°C and calculate the number of daily average temperature intervals:

[0023]

[0024] Where, T qm Indicates the number of temperature intervals;

[0025] Set the daily temperature difference interval to 2°C and calculate the number of daily temperature difference intervals:

[0026]

[0027] Where, ΔT max Indicates the maximum daily temperature difference in historical data, ΔT min Indicates the minimum daily temperature difference in historical data;

[0028] Define the boundaries of the daily mean temperature interval:

[0029] T bin =T min +5i

[0030] Where, T bin represents the boundary of the i-th daily average temperature interval, and each interval is closed on the left and open on the right, where i = 1, 2, ..., T qm -1;

[0031] Define the boundaries of the daily temperature range:

[0032] ΔT bin =ΔT min +2j

[0033] Where, ΔT in represents the boundary of the ith daily temperature difference interval, and each interval is closed on the left and open on the right, where j = 1, 2, ..., Twa -1;

[0034] For each daily average temperature in the historical meteorological data, determine the daily average temperature interval number to which it belongs:

[0035]

[0036] Where C T Indicates the daily average temperature interval number;

[0037] For each daily temperature difference in the historical meteorological data, determine the daily temperature difference interval number:

[0038]

[0039] Where C ΔT Indicates the daily temperature range number;

[0040] Number the daily average temperature interval C T and daily temperature range number C ΔT Combined into a two-dimensional classification mark (C T , C ΔT ), according to the combined classification mark (C T , C ΔT ) Group the historical meteorological data and the corresponding light and heat efficiency η to form T qm ×T wa A general dataset.

[0041] Furthermore, the daily average temperature, daily solar radiation intensity and sunshine duration of each general data set were extracted as input variables, and the light and heat efficiency was used as the output variable to construct a multivariate nonlinear regression model:

[0042] For each common data set, extract the daily average temperature T a , daily solar radiation intensity G and sunshine duration D as input variables, generating square terms: T a 2 , G 2 、D 2 , interaction term: T a G, T a ·D, G·D;

[0043] Construct the input feature matrix:

[0044] X=[T a , G, D, T a 2 , G 2 、D 2 , T a G, T a ·D, G·D]

[0045] Where X represents the input feature matrix of the general dataset;

[0046] Normalize the input features of this general dataset:

[0047]

[0048] In the formula, x represents the input feature of the general dataset, μ represents the mean of the input feature of the general dataset, σ represents the standard deviation of the input feature of the general dataset, X std Represents the input feature matrix of the general data set after normalization;

[0049] Construct a second-order polynomial regression model by inputting the feature matrix of the standardized general data set:

[0050] η new =β0+β1T a +β2G+β3D+β4T a 2 +β5G 2 +β6D 2 +β7T a G+β8T a ·D+β9G·D+∈

[0051] In the formula, β0-β9 represent the regression coefficients to be trained, ∈ represents the error term, and η new represents the predicted value of photothermal efficiency of the second-order polynomial regression model;

[0052] The least squares method is used to calculate the regression coefficients, with the goal of minimizing the residual sum of squares, according to the formula:

[0053]

[0054] In the formula, β=[β0, β1+β2+β3+β4+β5+β6+β7+β8+β9] T , x fg represents the g-th input feature of the f-th sample, where g = 1, 2, ..., 9, respectively representing T a , G, D, T a 2 , G 2 、D 2 , T a G, T a ·D, G·D, f = 1, 2, ..., n0, where n0 represents the number of samples in the general data set.

[0055] Furthermore, the method for calculating the sunshine duration based on astronomical formulas is:

[0056] Construct the solar declination calculation formula:

[0057]

[0058] Where δ represents the solar declination, N ts Indicates the number of days in a year;

[0059] Construct the hour angle calculation formula:

[0060]

[0061] Where, Indicates the local latitude;

[0062] Calculate sunshine duration based on solar declination and hour angle:

[0063]

[0064] Furthermore, the method of dividing the fixed area is:

[0065] Taking the logical center point of the solar thermal collection system as the origin, the area extends 10 km in all directions from the origin to form a fixed area of ​​20 km × 20 km.

[0066] Furthermore, the method of constructing regional datasets by combining historical meteorological data, solar thermal efficiency, and the tilt angle of the solar thermal system is as follows:

[0067] Pre-quantity threshold N th and the collection days threshold T for the cumulative light and heat efficiency of a single solar thermal collection system th , and N th >0,T th >1, count the number of solar thermal systems N under the same terrain in a single grid area actual , the number of days and tilt angles of each solar thermal collection system to collect light and heat efficiency, when a grid area meets N actual ≥N th and T actual ≥T th When the solar thermal system is used, the historical meteorological data, solar thermal efficiency and tilt angle are used to construct regional datasets.

[0068] Furthermore, the method of constructing a random forest model to predict the photothermal efficiency and calculate the root mean square error is as follows:

[0069] For each regional dataset, the daily average temperature, daily solar radiation intensity, sunshine duration, and inclination angle were standardized as input variables to construct a random forest model. The standardization formula was as follows:

[0070]

[0071] In the formula, μ′ represents the mean of the input variable of the regional dataset, σ′ represents the standard deviation of the input variable of the regional dataset, and X std′ Represents the input variable after the regional dataset is standardized;

[0072] The regional dataset was split into training and test sets with a ratio of 7:3. The number of trees was set to 100, the maximum depth was 6, the minimum number of samples for split nodes was 5, the minimum number of samples for leaf nodes was 3, the feature subset size was 2, self-sampling and out-of-bag validation were enabled, the random seed was 42, and the predicted light-to-heat efficiency was output.

[0073] For each sample, calculate the difference between the actual photothermal efficiency and the predicted photothermal efficiency:

[0074]

[0075] Where η k represents the actual light-heat efficiency of the kth sample, represents the predicted photothermal efficiency output by the random forest model;

[0076] Calculate the root mean square error:

[0077]

[0078] Where RMSE stands for root mean square error.

[0079] Furthermore, when the number of times that the statistical root mean square error in the preset time window is higher than the preset root mean square error threshold exceeds the preset number threshold, the method of switching to the multivariate nonlinear regression model to predict the photothermal efficiency is:

[0080] Set time window T sj =7 days, each time the time window is rolled over in units of natural days, and the root mean square error threshold is RMSE th , the number of times threshold T cs =3, when in time window T sj =7, the root mean square error is higher than the root mean square error threshold RMSE th When the number of times reaches the threshold, the multivariate nonlinear regression model is returned to predict the photothermal efficiency.

[0081] In addition, a solar thermal efficiency prediction system for a solar thermal collection system is provided. The system is used to execute the above-mentioned solar thermal efficiency prediction method for a solar thermal collection system, comprising:

[0082] A general data set construction module is used to collect historical meteorological data of the location of the solar thermal system and the solar thermal efficiency of the solar thermal system. The meteorological data includes daily average temperature, daily temperature range, solar radiation intensity and sunshine duration. A general data set is constructed based on the daily average temperature and daily temperature range and the historical meteorological data and solar thermal efficiency.

[0083] The regression model prediction module is used to extract the daily average temperature, daily solar radiation intensity, and sunshine duration of each common data set as input variables, and use the photothermal efficiency as the output variable to construct a multiple nonlinear regression model. The sunshine duration is calculated using an astronomical formula, and the daily average temperature, solar radiation intensity, and sunshine duration are input into the multiple nonlinear regression model to predict the photothermal efficiency.

[0084] The regional dataset construction module is used to divide a fixed area and classify the terrain where the solar thermal system is located. When the light and heat efficiency collected under the same terrain in a fixed area reaches the preset collection day threshold and quantity threshold, the historical meteorological data, light and heat efficiency, and the tilt angle of the solar thermal system are used to construct a regional dataset by region.

[0085] A random forest model building module is used to extract historical meteorological data, solar thermal efficiency, and tilt angle from a regional dataset, build a random forest model to predict solar thermal efficiency, calculate the root mean square error (RMS), set a RMS error threshold, and switch to a multivariate nonlinear regression model to predict solar thermal efficiency when the number of times the RMS error exceeds the preset RMS error threshold within a preset time window exceeds the preset threshold.

[0086] The model dynamic switching module is used to trigger the switching of the multivariate nonlinear regression model. When the cumulative number of days reaches the collection day threshold, the random forest model is re-enabled for prediction.

[0087] Compared with the prior art, the present invention has the following beneficial effects:

[0088] The present invention constructs a universal data set based on the daily average temperature and daily temperature difference according to the daily average temperature, daily temperature difference, solar radiation intensity, sunshine duration and photothermal efficiency, effectively capturing the nonlinear relationship between photothermal efficiency and meteorological data, and improving the prediction accuracy through a multivariate nonlinear regression model; the present invention also generates regional data sets according to different terrains, constructs a random forest model based on each regional data set, and intelligently switches between the two models, which is more targeted and reduces the risk of long-term prediction deviations. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 Schematic diagram of the overall method flow of the present invention;

[0090] Figure 2 This is a fitting diagram of daily average temperature and light and heat efficiency of the present invention;

[0091] Figure 3 This is a fitting diagram of the light-heat efficiency at high temperature of the present invention;

[0092] Figure 4 Switch schematic diagram for the model;

[0093] Figure 5 Schematic diagram of model accuracy;

[0094] Figure 6 It is a schematic diagram of the overall system module of the present invention. DETAILED DESCRIPTION

[0095] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0096] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0097] Example:

[0098] See also Figures 1 to 5 , the present invention provides a technical solution:

[0099] The method for predicting the photothermal efficiency of a solar thermal collection system includes the following specific steps:

[0100] Step 1: Collect historical meteorological data and the solar thermal efficiency of the solar thermal system at the location of the solar thermal system. The meteorological data includes daily average temperature, daily temperature range, solar radiation intensity, and sunshine duration. Construct a common dataset based on the historical meteorological data and solar thermal efficiency based on the daily average temperature and daily temperature range.

[0101] Optical efficiency reflects the collector's ability to convert solar energy into thermal energy under ideal conditions and is one of the important data for evaluating photothermal efficiency. Optical efficiency is obtained by referring to the standard product specifications of the solar thermal collection system and collecting historical meteorological data at the location of the solar thermal collection system.

[0102] Due to limitations in monitoring equipment or costs in some areas, it is impossible to obtain actual photothermal efficiency data for solar thermal systems, resulting in insufficient data for model training. Therefore, to address the lack of measured photothermal efficiency data in some areas, a physical model of the solar collector is combined with meteorological data to calculate the photothermal efficiency instead. Photothermal efficiency is directly related to temperature, radiation meteorological conditions, and collector performance parameters such as optical efficiency and heat loss coefficient. A mathematical relationship can be established through thermodynamic principles, providing theoretical support for the alternative calculation and a low-cost evaluation method for small and medium-sized solar projects with limited resources.

[0103] The specific method of replacing the photothermal efficiency by calculating the physical model of the solar collector in combination with meteorological data is as follows:

[0104] Calculate the average collector temperature based on the daily maximum and minimum temperatures:

[0105]

[0106] Where, T m Indicates the average collector temperature, T max Indicates the maximum daily temperature, T min Indicates the daily minimum temperature, taking into account the daily average temperature and the fluctuation range of daily temperature to more accurately reflect the actual operating temperature of the collector in a day;

[0107] Refer to the standard product specifications of the solar thermal system to obtain the heat loss coefficient and construct a mathematical calculation model for the photothermal efficiency:

[0108]

[0109] Where η0 represents the optical efficiency, a1 and a2 represent the heat loss coefficient, and T a represents the daily average temperature, G represents the daily solar radiation intensity, that is, the average solar radiation intensity of the day, and the heat loss coefficient describes the degree to which the collector dissipates heat to the surrounding environment under different working conditions. This is a simplified version of the efficiency model based on the solar collector test standard, which is used to test the collector efficiency.

[0110] Heat loss will increase at high temperatures. The effects of temperature and temperature difference on photothermal efficiency are nonlinear. Data can be discretized by interval division to avoid ignoring local features. The meteorological conditions within the same temperature and temperature difference intervals are similar, which can reduce the interference of outliers on the model. The daily average temperature interval is set to 5°C, and the number of daily average temperature intervals is calculated:

[0111]

[0112] Where, T qmRepresents the number of temperature intervals. The heat loss of solar collectors is closely related to the temperature difference. The complex temperature effect is converted into a quantifiable interval classification. The 5°C interval can effectively distinguish the changes in photothermal efficiency under different temperature gradients. The degree of heat loss within this interval is relatively consistent. If the interval is too small, the sample size in each interval will be insufficient, resulting in insufficient training of the multivariate nonlinear regression model. If the interval is too large, the feature differences cannot be reflected, which reduces the generalization ability of the model.

[0113] Set the daily temperature difference interval to 2°C and calculate the number of daily temperature difference intervals:

[0114]

[0115] Where, ΔT max Indicates the maximum daily temperature difference in historical data, ΔT min Indicates the minimum daily temperature difference in historical data. The 2°C interval ensures that there is enough data in each interval for analysis and that the classification is detailed enough.

[0116] Define the boundaries of the daily mean temperature interval:

[0117] T bin =T min +5i

[0118] Where, T bin represents the boundary of the i-th daily average temperature interval, and each interval is closed on the left and open on the right, where i = 1, 2, ..., T qm -1;

[0119] Define the boundaries of the daily temperature range:

[0120] ΔT bin =ΔT min +2j

[0121] Where, ΔT in represents the boundary of the ith daily temperature difference interval, and each interval is closed on the left and open on the right, where j = 1, 2, ..., T wa -1;

[0122] For each daily average temperature in the historical meteorological data, determine the daily average temperature interval number to which it belongs:

[0123]

[0124] Where C T Indicates the daily average temperature interval number, and the number starts from 0;

[0125] For each daily temperature difference in the historical meteorological data, determine the daily temperature difference interval number:

[0126]

[0127] Where C ΔT Indicates the daily temperature difference interval number, and the number starts from 0;

[0128] Number the daily average temperature interval C T and daily temperature range number C ΔT Combined into a two-dimensional classification mark (C T , C ΔT ), according to the combined classification mark (C T , C ΔT ) Group the historical meteorological data and the corresponding light and heat efficiency η to form T qm ×T wa The universal data set cross-groups historical meteorological data according to two key variables: daily average temperature and daily temperature difference, and discretizes complex meteorological conditions into quantifiable feature labels. It can accurately capture the performance differences of solar thermal systems under different climatic conditions. For example, under the same daily average temperature, the photothermal efficiency of high temperature difference (such as in desert areas) and low temperature difference (such as in coastal areas) will vary significantly due to different heat loss patterns.

[0129] Step 2: Extract daily average temperature, daily solar radiation intensity, and sunshine duration from each common data set as input variables, and use photothermal efficiency as the output variable to construct a multiple nonlinear regression model. Calculate the sunshine duration using an astronomical formula, and input the daily average temperature, solar radiation intensity, and sunshine duration into the multiple nonlinear regression model to predict photothermal efficiency.

[0130] From the heat loss characteristics, we can see that when the temperature difference is large, the radiation loss and the temperature difference are nonlinear. The quadratic term can better describe the influence of radiation loss. From a large number of experiments, we can know that the linear model will underestimate the efficiency growth rate in the low temperature area and overestimate the efficiency maintenance ability in the high temperature area. Especially when the temperature is above 30℃, the measured value of the photothermal efficiency will decrease. For each general data set, the daily average temperature T is extracted. a , daily solar radiation intensity G and sunshine duration D as input variables, generating square terms: T a 2 , G 2 、D 2 , to capture the nonlinear relationship between the variables themselves, the interaction term: T a G, T a ·D, G·D, to capture the synergistic effects among variables;

[0131] Construct the input feature matrix:

[0132] X=[T a , G, D, T a 2 , G 2 、D 2 , Ta G, T a ·D, G·D]

[0133] Where X represents the input feature matrix of the general dataset;

[0134] The units of the input variables are different. Directly inputting the model will cause certain features to dominate the training process. The input features of this common dataset are standardized:

[0135]

[0136] In the formula, x represents the input feature of the general dataset, μ represents the mean of the input feature of the general dataset, σ represents the standard deviation of the input feature of the general dataset, X std Represents the input feature matrix of the general data set after normalization;

[0137] The relationship between solar thermal efficiency and meteorological data is nonlinear. The second-order polynomial model can capture this nonlinear trend. The second-order polynomial regression model is constructed by inputting the feature matrix of the standardized general data set:

[0138] η new =β0+β1T a +β2G+β3D+β4T a 2 +β5G 2 +β6D 2 +β7T a G+β8T a ·D+β9G·D+∈

[0139] In the formula, β0-β9 represent the regression coefficients to be trained, ∈ represents the error term, and η new represents the predicted value of photothermal efficiency of the second-order polynomial regression model;

[0140] The least squares method is used to calculate the regression coefficients, with the goal of minimizing the residual sum of squares, according to the formula:

[0141]

[0142] In the formula, β=[β0, β1+β2+β3+β4+β5+β6+β7+β8+β9] T , x fg represents the g-th input feature of the f-th sample, where g = 1, 2, ..., 9, respectively representing T a , G, D, T a 2 , G 2 、D 2 , T a G, T aD, G·D, f = 1, 2, …, n0, where n0 represents the number of samples in the general dataset;

[0143] Construct the solar declination calculation formula:

[0144]

[0145] Where δ represents the solar declination, N ts Indicates the number of days in a year;

[0146] Construct the hour angle calculation formula:

[0147]

[0148] Where, Indicates the local latitude;

[0149] Calculate sunshine duration based on solar declination and hour angle:

[0150]

[0151] For the date on which the light and thermal efficiency is to be predicted, the daily average temperature and daily solar radiation intensity are obtained from the weather forecast. The sunshine duration is calculated according to the solar declination and hour angle. The classification mark is determined by the daily average temperature and daily temperature difference. The input feature matrix is ​​constructed. The daily average temperature, daily solar radiation intensity and sunshine duration are standardized according to the same standardization method. The input is input into the completed second-order polynomial regression model to obtain the predicted light and thermal efficiency. Table 1 shows that under the condition that the daily average temperature and daily temperature difference are in the same classification mark, for the daily average temperature in the range of 20℃-25℃, the fixed solar radiation intensity is 805W / m 2 , under the working conditions of 9.85 hours of sunshine, 40 sets of photothermal efficiency data of a certain solar thermal collection system were collected. At the same time, the daily average temperature, solar radiation intensity and sunshine duration were standardized according to the above formulas. The standardized data were summarized into the following data table for verification of the photothermal efficiency prediction model of the solar thermal collection system.

[0152]

[0153]

[0154] Table 1 Daily average temperature fitting data table

[0155] like Figure 2 As shown in the figure, under the conditions of fixed solar radiation intensity and sunshine duration, the photothermal efficiency and the daily average temperature are fitted into a quadratic function relationship. As the daily average temperature gradually increases, the photothermal efficiency is generally distributed around the fitting curve, which proves that this model can more accurately predict the photothermal efficiency of the solar thermal collection system.

[0156] Table 3 shows that under the conditions that the daily average temperature and the daily temperature difference are in the same classification, for the daily average temperature in the range of 35℃-40℃, the fixed solar radiation intensity is 870W / m 2 , under the working condition of 8.5 hours of sunshine, 40 sets of light and heat efficiency data of a certain solar thermal collection system were collected.

[0157]

[0158]

[0159] Table 2 Photothermal efficiency performance at high temperature

[0160] like Figure 3 As shown in the figure, the photothermal efficiency has a significant quadratic function relationship with the daily average temperature. The photothermal efficiency first increases with the increase of the daily average temperature, and then slowly decreases when the daily average temperature reaches 37°C, indicating that the heat loss of this solar thermal collection system increases at high temperatures, resulting in a decrease in efficiency. From the analysis of industrial cases, it can be seen that in the solar thermal power generation system, the higher the temperature of the superheated steam, the greater the heat loss from the heat exchanger to the environment, and it needs to be optimized through insulation design and high-efficiency heat collection structure.

[0161] Step 3: Divide a fixed area and classify the terrain where the solar thermal system is located. When the solar thermal efficiency collected under the same terrain in the fixed area reaches the preset collection day threshold and quantity threshold, the historical meteorological data, solar thermal efficiency, and the tilt angle of the solar thermal system are used to construct a regional dataset.

[0162] GIS software is used to obtain topographic data for the location of the solar thermal collection system. The terrain type is a landform classification that describes the external form and structure of the terrain, such as plains, mountains, hills, etc. GIS software, such as Tiandi Map software, can be used to input addresses or coordinates to identify terrain such as valleys and plains, and the geographical location of the solar thermal collection system for which the solar thermal efficiency is to be predicted is found. The logical center point of the solar thermal collection system is used as the origin, and a fixed area of ​​20 km × 20 km is formed from the origin in all directions. The meteorological variation within this area is small, thereby meeting the meteorological consistency requirement. For example, meteorological bureau data shows that the spatial variation of daily solar radiation intensity in plain areas is generally less than 5%, and the terrain characteristics within a 20 km grid are relatively small. Even in different terrains, input variables such as solar radiation correction coefficient and heat loss compensation coefficient can be dynamically adjusted based on the terrain type, or solar thermal collection systems with smaller terrain differences can be directly merged into data sets of other terrains. The solar thermal collection systems are of the same brand and specifications, thereby ensuring the consistency of various parameters.

[0163] Pre-quantity threshold N thand the cumulative collection day threshold T of the solar thermal system's light and heat efficiency th , where N th >0,T th >1 and is a positive integer. For example, for large-scale power stations, the number threshold N can be preset. th =50, T th =365, small-scale distributed system, pre-threshold N th =15, T th =300, the number of solar thermal systems N actual , count the days when each solar thermal system collects light and heat efficiency, and the light and heat efficiency is not calculated by combining the physical model with meteorological data, avoiding the error caused by theoretical assumptions, ignoring the effects of collector aging, dirt, etc., verifying the system performance through real operation data, and improving the model's adaptability to complex real-world scenarios such as extreme weather. When a grid area meets N actual ≥N th and T actual ≥T th When determining the optimal solar thermal system, a regional data set will be constructed using the historical meteorological data, solar thermal efficiency, and tilt angle of the solar thermal system that meet the conditions, thus forming the basis for transitioning from empirical judgment to data judgment.

[0164] Step 4: Extract historical meteorological data, solar thermal efficiency, and tilt angle from the regional dataset, build a random forest model to predict solar thermal efficiency, and calculate the root mean square error (RMS). Set a RMS error threshold. When the number of times the RMS error exceeds the preset RMS error threshold within a preset time window exceeds the preset threshold, switch to a multivariate nonlinear regression model to predict solar thermal efficiency.

[0165] For each regional data set, the daily average temperature T a , daily solar radiation intensity G, sunshine duration D, and solar thermal system tilt angle S are standardized as input variables. This operation eliminates the dimensional differences of the variables, allowing the model to more efficiently learn the relationship between features and construct a random forest model. The integration characteristics of random forests reduce the overfitting risk of a single decision tree and are particularly suitable for nonlinear relationships, providing a basis for optimizing the tilt angle. The formula for standardization is:

[0166]

[0167] In the formula, μ′ represents the mean of the input variable of the regional dataset, σ′ represents the standard deviation of the input variable of the regional dataset, and X std′ Represents the input variable after the regional dataset is standardized;

[0168] The regional dataset is divided into a training set and a test set with a ratio of 7:3. The training set is used for model parameter learning, and the test set is used to verify the model's generalization ability and avoid overfitting. The 7:3 ratio balances the amount of data and verification accuracy, taking into account both the adequacy of model training and the reliability of testing, so that the model can capture the potential mapping relationship between input variables and photothermal efficiency from the data. The number of trees is set to 100. The overfitting risk of a single tree is reduced by integrating multiple decision trees, improving model stability and prediction accuracy. The maximum depth is 6, which is suitable for feature interactions of medium complexity. The minimum number of samples for split nodes is 5, and the minimum number of samples for leaf nodes is 3 to prevent the model from overfitting noisy data during training. The feature subset size is 2. Self-service sampling and out-of-bag validation are enabled. Training subsets are generated by random sampling with replacement to increase tree diversity and reduce the model's dependence on specific samples. Unselected samples are used to evaluate the model without the need for an additional validation set. The random seed is 42, and the random process is fixed to ensure experimental repeatability. The predicted photothermal efficiency is output.

[0169] For each sample, calculate the difference between the actual photothermal efficiency and the predicted photothermal efficiency:

[0170]

[0171] Where η k represents the actual light-heat efficiency of the kth sample, represents the predicted photothermal efficiency output by the random forest model;

[0172] Calculate the root mean square error:

[0173]

[0174] Where RMSE stands for root mean square error, which can accurately reflect the overall deviation between the predicted value and the true value. The smaller the value, the closer the predicted value is to the true value, and the higher the model accuracy. It provides a data source for model evaluation and switching. Set the time window T sj =7 days. A 7-day window can completely cover a natural week. This is a common practice. The daily root mean square error is calculated within the time window. The time window is updated on a rolling basis in natural days. This allows you to promptly detect whether the prediction accuracy has decreased, thus avoiding the problem of long-term prediction inaccuracy. The root mean square error threshold is set to RMSE. th , and 1>RMSE th >0, for example, RMSE thThe error distribution statistics based on historical data are close to the normal fluctuation range of most solar thermal systems in stable operation. In 85% of the prediction results, RMSE is ≤ 0.03. For the prediction results with RMSE> 0.05, most of them are caused by sensor failure, terrain statistics errors or seasonal changes. At this time, the multivariate nonlinear regression model will perform more stably. Comparing the RMSE> RMSE in the current time window, the th The number of times, set the number threshold T cs =3, when in time window T si =7, the number of times the root mean square error is higher than the root mean square error threshold reaches the number threshold, and the multivariate nonlinear regression model is returned to predict the photothermal efficiency. The solar thermal system may cause RMSE>RMSE due to sudden weather such as heavy rain, sandstorms or temporary equipment failure. th The set time window and number threshold allow for two errors exceeding the limit within 7 days to avoid the assumption that the model is inaccurate due to accidental factors.

[0175] Step 5: After triggering the switch to the multivariate nonlinear regression model, when the cumulative number of days reaches the collection day threshold, the random forest model is re-enabled for prediction;

[0176] During the prediction period using the multivariate nonlinear regression model, new meteorological data and light-thermal efficiency data are continuously collected. When the time for the random forest model to update with the newly collected data reaches the set time, that is, the cumulative light-thermal efficiency collection days threshold T of the solar thermal system, the new meteorological data and light-thermal efficiency data are continuously collected. th , the random forest model is enabled again for prediction. At this time, the collection days threshold T thReset to 3. This means that after the three-day multivariate nonlinear regression model is used, the random forest model is used for prediction. The efficiency of solar thermal systems is significantly affected by short-term fluctuations in weather conditions. As a non-time-series machine learning model, the random forest model is sensitive to situations where short-term fluctuations may have a significant impact. Therefore, after the three-day multivariate nonlinear regression model is used, the random forest model is switched back to the random forest model, and the date the random forest model is reactivated is set to the first day of the new time window. If the random forest model's prediction error within the time window frequently exceeds the root mean square error threshold, it indicates that the model may not be able to adapt well to data changes, such as significant changes in weather conditions or shifts in the data distribution. At this time, switching to the multivariate nonlinear regression model, whose different data fitting methods can further improve prediction accuracy, is recommended. The random forest model generally has stronger generalization capabilities and the ability to capture complex relationships. After a period of time of updating with new data, it will adapt to the changes in the data, so re-enabling it can give full play to its advantages and continue to make more accurate predictions. Table 3 shows that, with natural days as the unit, it slides back one day every day to form a continuous 7-day window. By counting the number of times the root mean square error exceeds the root mean square error threshold within the time window, that is, the number of times the limit is exceeded within the window, it is determined whether the model needs to be switched and summarized to form a data table. When switching from the multivariate nonlinear regression model to the random forest model for predicting the photothermal efficiency of the solar thermal system, although the time window is less than 7 days, when the number of times the root mean square error exceeds the root mean square error threshold reaches the number threshold, it still switches back to the multivariate nonlinear regression model to avoid excessive prediction deviation.

[0177]

[0178]

[0179] Table 3 Model switching data table

[0180] like Figure 4 As shown in the figure, this figure shows the change of the root mean square error of the forest model and the multiple nonlinear regression model with the date serial number. The horizontal axis is the date serial number and the vertical axis is the root mean square error. Whenever the root mean square error of the forest model is higher than the root mean square error threshold 3 times in a time window, the multiple nonlinear regression model is used for prediction, which effectively controls the error. The root mean square error of the forest model gradually decreases and stabilizes below the threshold. There is no need to trigger the operation of switching models. This shows that the forest model may improve the prediction accuracy and maintain stable performance through parameter adjustment or data adaptability optimization in the later stage.

[0181] like Figure 5As shown in the figure, a line graph showing the change of the root mean square error over the time window is shown. Within a time window, the blue line represents that the root mean square error exceeds the root mean square error threshold line at least 3 times, indicating that the current model prediction error is large, and the model needs to switch from the random forest model to the multivariate nonlinear regression model to predict the photothermal efficiency. The black line represents that the error of the predicted photothermal efficiency is small, indicating that the predicted photothermal efficiency of the solar thermal collection system is highly reliable.

[0182] See also Figure 6 The present invention further provides a solar thermal efficiency prediction system for a solar thermal collection system, wherein the system is used to execute the above-mentioned solar thermal efficiency prediction method for a solar thermal collection system, comprising:

[0183] A general data set construction module is used to collect historical meteorological data of the location of the solar thermal system and the solar thermal efficiency of the solar thermal system. The meteorological data includes daily average temperature, daily temperature range, solar radiation intensity and sunshine duration. A general data set is constructed based on the daily average temperature and daily temperature range and the historical meteorological data and solar thermal efficiency.

[0184] The regression model prediction module is used to extract the daily average temperature, daily solar radiation intensity, and sunshine duration of each common data set as input variables, and use the photothermal efficiency as the output variable to construct a multiple nonlinear regression model. The sunshine duration is calculated using an astronomical formula, and the daily average temperature, solar radiation intensity, and sunshine duration are input into the multiple nonlinear regression model to predict the photothermal efficiency.

[0185] The regional dataset construction module is used to divide a fixed area and classify the terrain where the solar thermal system is located. When the light and heat efficiency collected under the same terrain in a fixed area reaches the preset collection day threshold and quantity threshold, the historical meteorological data, light and heat efficiency, and the tilt angle of the solar thermal system are used to construct a regional dataset by region.

[0186] A random forest model building module is used to extract historical meteorological data, solar thermal efficiency, and tilt angle from a regional dataset, build a random forest model to predict solar thermal efficiency, calculate the root mean square error (RMS), set a RMS error threshold, and switch to a multivariate nonlinear regression model to predict solar thermal efficiency when the number of times the RMS error exceeds the preset RMS error threshold within a preset time window exceeds the preset threshold.

[0187] The model dynamic switching module is used to trigger the switching of the multivariate nonlinear regression model. When the cumulative number of days reaches the collection day threshold, the random forest model is re-enabled for prediction.

[0188] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0189] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0190] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0191] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for predicting the photothermal efficiency of a solar thermal collection system, characterized in that: The specific steps include: Step 1: Collect historical meteorological data and the solar thermal efficiency of the solar thermal system at the location of the solar thermal system. The meteorological data includes daily average temperature, daily temperature range, solar radiation intensity, and sunshine duration. Construct a common dataset based on the historical meteorological data and solar thermal efficiency based on the daily average temperature and daily temperature range. Step 2: Extract daily average temperature, daily solar radiation intensity, and sunshine duration from each common data set as input variables, and use photothermal efficiency as the output variable to construct a multiple nonlinear regression model. Calculate the sunshine duration using an astronomical formula, and input the daily average temperature, solar radiation intensity, and sunshine duration into the multiple nonlinear regression model to predict photothermal efficiency. Step 3: Divide a fixed area and classify the terrain where the solar thermal system is located. When the solar thermal efficiency collected under the same terrain in the fixed area reaches the preset collection day threshold and quantity threshold, the historical meteorological data, solar thermal efficiency, and the tilt angle of the solar thermal system are used to construct a regional dataset. Step 4: Extract historical meteorological data, solar thermal efficiency, and tilt angle from the regional dataset, build a random forest model to predict solar thermal efficiency, and calculate the root mean square error (RMS). Set a RMS error threshold. When the number of times the RMS error exceeds the preset RMS error threshold within a preset time window exceeds the preset threshold, switch to a multivariate nonlinear regression model to predict solar thermal efficiency. Step 5: After triggering the switch of the multivariate nonlinear regression model, when the cumulative number of days reaches the collection day threshold, the random forest model is re-enabled for prediction.

2. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 1, wherein: Collect historical meteorological data and the solar thermal efficiency of the solar thermal system in the location of the solar thermal system. By consulting the standard product specifications of the solar thermal system, the optical efficiency and heat loss coefficient are obtained. In order to solve the problem of lack of measured data on solar thermal efficiency in some areas, the solar thermal efficiency is replaced by calculation using the physical model of the solar thermal collector combined with meteorological data. Calculate the average collector temperature: Where, T m Indicates the average collector temperature, T max Indicates the maximum daily temperature, T min Indicates the daily minimum temperature; Constructing a mathematical calculation model for photothermal efficiency: Where η0 represents the optical efficiency, a1 and a2 represent the heat loss coefficient, and T a represents the daily average temperature, and G represents the daily solar radiation intensity, that is, the average solar radiation intensity of the day.

3. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 2, wherein: The method for constructing a general data set based on the daily average temperature and daily temperature difference for historical meteorological data and light and heat efficiency is as follows: Set the daily average temperature interval to 5°C and calculate the number of daily average temperature intervals: Where, T qm Indicates the number of temperature intervals; Set the daily temperature difference interval to 2°C and calculate the number of daily temperature difference intervals: Where, ΔT max Indicates the maximum daily temperature difference in historical data, ΔT min Indicates the minimum daily temperature difference in historical data; Define the boundaries of the daily mean temperature interval: T bin =T min +5i Where, T bin represents the boundary of the i-th daily average temperature interval, and each interval is closed on the left and open on the right, where i = 1, 2, ..., T qm -1; Define the boundaries of the daily temperature range: ΔT bin =ΔT min +2j Where Δt bin represents the boundary of the ith daily temperature difference interval, and each interval is closed on the left and open on the right, where j = 1, 2, ..., t wa -1; For each daily average temperature in the historical meteorological data, determine the daily average temperature interval number to which it belongs: Where C T Indicates the daily average temperature interval number; For each daily temperature difference in the historical meteorological data, determine the daily temperature difference interval number: Where C ΔT Indicates the daily temperature range number; The daily average temperature interval is numbered C T and daily temperature range number C ΔT Combined into a two-dimensional classification mark (C T , C ΔT ), according to the combined classification mark (C T , C ΔT ) Group the historical meteorological data and the corresponding light and heat efficiency η to form T qm ×T wa A general dataset.

4. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 3, wherein: The method of extracting the daily average temperature, daily solar radiation intensity and sunshine duration of each general data set as input variables and the light and heat efficiency as the output variable to construct a multivariate nonlinear regression model is as follows: For each common data set, extract the daily average temperature T a , daily solar radiation intensity G and sunshine duration D as input variables, generating square terms: T a 2 , G 2 、D 2 , interaction term: T a G, T a ·D, G·D; Construct the input feature matrix: X=[T a ,G,D,T a 2 ,G 2 、D 2 ,T a ·G,T a ·D,G·D] Where X represents the input feature matrix of the general dataset; Normalize the input features of this general dataset: In the formula, x represents the input feature of the general dataset, μ represents the mean of the input feature of the general dataset, σ represents the standard deviation of the input feature of the general dataset, X std Represents the input feature matrix of the general data set after normalization; Construct a second-order polynomial regression model by inputting the feature matrix of the standardized general data set: or new =β0+β1T a +β2G+β3D+β4T a 2 +β5G 2 +β6D 2 +β7T a ·G+β8T a ·D+β9G·D+∈ In the formula, β0-β9 represent the regression coefficients to be trained, ∈ represents the error term, and η new represents the predicted value of photothermal efficiency of the second-order polynomial regression model; The least squares method is used to calculate the regression coefficients, with the goal of minimizing the residual sum of squares, according to the formula: In the formula, β=[β0, β1+β2+β3+β4+β5+β6+β7+β8+β9] T , x fg represents the g-th input feature of the f-th sample, where g = 1, 2, ..., 9, respectively representing T a , G, D, T a 2 , G 2 、D 2 , T a G, T a ·D, G·D, f = 1, 2, ..., n0, where n0 represents the number of samples in the general data set.

5. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 1, wherein: The method for calculating the duration of sunshine based on astronomical formulas is: Construct the solar declination calculation formula: Where δ represents the solar declination, N ts Indicates the number of days in a year; Construct the hour angle calculation formula: Where, Indicates the local latitude; Calculate sunshine duration based on solar declination and hour angle:

6. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 1, wherein: The method of dividing the fixed area is: Taking the logical center point of the solar thermal collection system as the origin, the area extends 10 km in all directions from the origin to form a fixed area of ​​20 km × 20 km.

7. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 1, wherein: The method of constructing regional datasets by combining historical meteorological data, solar thermal efficiency, and the tilt angle of the solar thermal collection system is as follows: Pre-quantity threshold N th and the collection days threshold T for the cumulative light and heat efficiency of a single solar thermal collection system th , and N th >0,T th >1, count the number of solar thermal systems N under the same terrain in a single grid area actual , the number of days and tilt angles of each solar thermal collection system to collect light and heat efficiency, when a grid area meets N actual ≥N th and T actual ≥T th When the solar thermal system is used, the historical meteorological data, solar thermal efficiency and tilt angle are used to construct regional datasets.

8. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 1, wherein: The method for constructing a random forest model to predict photothermal efficiency and calculate the root mean square error is: For each regional dataset, the daily average temperature, daily solar radiation intensity, sunshine duration, and inclination angle were standardized as input variables to construct a random forest model. The standardization formula was as follows: In the formula, μ′ represents the mean of the input variable of the regional dataset, σ′ represents the standard deviation of the input variable of the regional dataset, and X std′ Represents the input variable after the regional dataset is standardized; The regional dataset was split into training and test sets with a ratio of 7:

3. The number of trees was set to 100, the maximum depth was 6, the minimum number of samples for split nodes was 5, the minimum number of samples for leaf nodes was 3, the feature subset size was 2, self-sampling and out-of-bag validation were enabled, the random seed was 42, and the predicted light-to-heat efficiency was output. For each sample, calculate the difference between the actual photothermal efficiency and the predicted photothermal efficiency: Where η k represents the actual light-heat efficiency of the kth sample, represents the predicted photothermal efficiency output by the random forest model; Calculate the root mean square error: Where RMSE stands for root mean square error.

9. The method for predicting the photothermal efficiency of a solar thermal collection system according to claim 8, wherein: When the number of times that the statistical root mean square error in the preset time window exceeds the preset root mean square error threshold exceeds the preset number threshold, the method of switching to the multivariate nonlinear regression model to predict the light thermal efficiency is: Set time window T sj =7 days, each time the time window is rolled over in units of natural days, and the root mean square error threshold is RMSE th , the number of times threshold T cs =3, when in time window T sj =7, the root mean square error is higher than the root mean square error threshold RMSE th When the number of times reaches the threshold, the multivariate nonlinear regression model is returned to predict the photothermal efficiency.

10. Solar thermal efficiency prediction system, characterized by: The system is used to execute the method for predicting the photothermal efficiency of a solar thermal collection system according to any one of claims 1 to 9, comprising: A general data set construction module is used to collect historical meteorological data of the location of the solar thermal system and the solar thermal efficiency of the solar thermal system. The meteorological data includes daily average temperature, daily temperature range, solar radiation intensity and sunshine duration. A general data set is constructed based on the daily average temperature and daily temperature range and the historical meteorological data and solar thermal efficiency. The regression model prediction module is used to extract the daily average temperature, daily solar radiation intensity, and sunshine duration of each common data set as input variables, and use the photothermal efficiency as the output variable to construct a multiple nonlinear regression model. The sunshine duration is calculated using an astronomical formula, and the daily average temperature, solar radiation intensity, and sunshine duration are input into the multiple nonlinear regression model to predict the photothermal efficiency. The regional dataset construction module is used to divide a fixed area and classify the terrain where the solar thermal system is located. When the light and heat efficiency collected under the same terrain in a fixed area reaches the preset collection day threshold and quantity threshold, the historical meteorological data, light and heat efficiency, and the tilt angle of the solar thermal system are used to construct a regional dataset by region. A random forest model building module is used to extract historical meteorological data, solar thermal efficiency, and tilt angle from a regional dataset, build a random forest model to predict solar thermal efficiency, calculate the root mean square error (RMS), set a RMS error threshold, and switch to a multivariate nonlinear regression model to predict solar thermal efficiency when the number of times the RMS error exceeds the preset RMS error threshold within a preset time window exceeds the preset threshold. The model dynamic switching module is used to trigger the switching of the multivariate nonlinear regression model. When the cumulative number of days reaches the collection day threshold, the random forest model is re-enabled for prediction.

Images