Regional temperature field prediction system based on big data analysis
By constructing and refining a temperature field prediction model through pilot testing in a small area, and combining meteorological factors and heat conduction effects, the problems of high cost and insufficient accuracy of traditional large-area temperature field monitoring have been solved, achieving efficient and accurate temperature field prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- PLA OF CHINA AIR FORCE EARLY WARNING ACADEMY LEIDA SERGEANT SCHOOL
- Filing Date
- 2026-03-31
- Publication Date
- 2026-05-26
AI Technical Summary
Traditional large-area temperature field monitoring is costly and lacks real-time performance. Furthermore, existing prediction methods ignore the correlation between small and large areas, resulting in insufficient accuracy of prediction results that are difficult to meet the requirements of practical applications.
Based on the prediction and promotion approach of small-area pilot projects, and combining meteorological elements with measured data, a temperature field prediction model is constructed through a process of modeling, comparison, and correction. Taking into account the heat conduction effect and error correction, the accurate prediction of the temperature field in large areas is achieved.
It achieves efficient and accurate large-area temperature field prediction, reduces monitoring costs, improves real-time performance and prediction accuracy, and ensures that the prediction results conform to the actual temperature distribution pattern.
Smart Images
Figure CN122087771A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of meteorological forecasting, specifically relating to a regional temperature field forecasting system based on big data analysis. Background Technology
[0002] Due to various factors such as topography and underlying surface conditions, the spatial distribution of temperature fields in large regions exhibits significant heterogeneity, posing numerous challenges to the accurate acquisition of temperature fields across the entire region.
[0003] Traditional large-area temperature field monitoring often adopts the method of full-area on-site measurement. This method not only requires the deployment of a large number of monitoring devices, but also requires a lot of manpower, material resources and time costs. The monitoring cycle is long and the real-time performance is poor, making it difficult to meet the dynamic needs of production for temperature field data. Moreover, the feasibility of full-area on-site measurement is low in large contiguous areas.
[0004] Currently, most existing temperature field prediction methods ignore the correlation between small and large areas, directly modeling and predicting the entire large area. This leads to significant discrepancies between the predicted results and the actual temperature field, making it difficult to meet the accuracy requirements of practical applications. Therefore, how to overcome the limitations of traditional monitoring and prediction technologies and find an efficient, accurate, and low-cost method for acquiring large-area temperature fields has become a pressing technical challenge in the field of environmental monitoring. Summary of the Invention
[0005] This invention provides a regional temperature field prediction system based on big data analysis. Based on a prediction generalization approach using small-area pilot projects, and combining relevant meteorological elements with measured data, it achieves accurate prediction of large-area temperature fields through a process of modeling, comparison, and correction. This effectively solves the problems in the background technology.
[0006] This invention provides a regional temperature field prediction system based on big data analysis, comprising the following steps:
[0007] S1: Select a small pilot area within the large area and divide the pilot area into grids;
[0008] S2: Collect meteorological data for the pilot area, including the average daily temperature of the pilot area. Total solar radiation in pilot areas Wind speed in pilot areas relative humidity in the pilot area Simultaneously, the actual temperature of the pilot area was collected. , Indicates the first )Grid at time The measured temperature;
[0009] S3: Preprocess the data collected in S2;
[0010] S4: Based on the meteorological data in S3, construct a temperature field prediction model:
[0011] ,in Indicates the first )Grid at time The predicted temperature The regression coefficients to be solved are... This is random error;
[0012] S5: Solve for the regression coefficients;
[0013] S6: Divide the area into large-area grids and acquire large-area meteorological data, including the large-area daily average temperature. Total solar radiation over a large area Large area wind speed and the relative humidity of the large area , For the large region Grid;
[0014] S7: Based on the large-area meteorological data obtained in S6 and the temperature field prediction model constructed in S4, the large-area temperature field prediction results are obtained:
[0015]
[0016] in, This is a correction term for heat conduction.
[0017] As a further optimization of the present invention, a small meteorological station is set up in the center of the pilot area to obtain meteorological data of the pilot area.
[0018] As a further optimization of the present invention, the meteorological data acquisition frequency in the pilot area of S2 is once per hour, and the acquisition cycle is... Day, total acquisition Group time series data.
[0019] As a further optimization of the present invention, the preprocessing in S3 includes:
[0020] Missing values are handled by using linear interpolation to fill in missing data in the time series.
[0021] Outlier handling: Outliers are removed using the 3σ criterion.
[0022] And normalization processing, normalizing the meteorological data to... Intervals, eliminating the influence of dimensions.
[0023] As a further optimization of the present invention, the solution of the regression coefficients in step S5 includes the following steps:
[0024] First, minimize the sum of squared errors:
[0025]
[0026] Then, the temperature field prediction model is transformed into matrix form. ,in:
[0027] , ,
[0028] Finally, the regression coefficients for each pilot area grid are solved using the least squares formula:
[0029]
[0030] As a further optimization of the present invention, a heat conduction correction step considering spatial heat conduction is also included, as follows:
[0031] Construct the heat conduction equation for the temperature field in the pilot area:
[0032]
[0033] in, Soil thermal conductivity; For soil heat generation, , For soil absorption rate, Temperature acquisition depth; , For temperature at , Spatial gradient in direction;
[0034] The heat conduction equation is discretized using the finite difference method. For the ... The discretization formula for the mesh is:
[0035]
[0036] After processing, the mesh temperature after thermal conductivity correction is obtained:
[0037]
[0038] in, For the first Initial time-series predicted temperature of four adjacent grid cells .
[0039] As a further optimization of the present invention, an error correction step considering error accuracy is also included, as follows:
[0040] First, select the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination. As an error indicator:
[0041]
[0042]
[0043]
[0044] in, , which is the average value of the actual temperature field;
[0045] Preset precision threshold: If the preset accuracy threshold is met simultaneously, no accuracy correction is needed; if any one of them is not met, the regression coefficient is corrected. At the same time, adjust the soil thermal conductivity The steps are as follows:
[0046] S'1: Calculate the gradient of the regression coefficients ;
[0047] S'2: Update regression coefficients , The learning rate is set to 0.001-0.01.
[0048] S'3: Adjusting soil thermal conductivity , This is a correction factor;
[0049] S'4: Repeat S'1-S'3 until the error meets the preset threshold.
[0050] As a further optimization of the present invention, a preset precision threshold is provided. The values are 1.5℃, 1.0℃, and 0.85, respectively.
[0051] As a further optimization of the present invention, the acquired large-area meteorological data is the regional mean, and a meteorological correction coefficient is introduced. ,in The average daily temperature in the pilot area;
[0052] For the total radiation of a large area For large-area wind speeds For relative humidity .
[0053] This invention provides a regional temperature field prediction system based on big data analysis. By employing a method of small-area pilot modeling, comparing and correcting predictions with actual measurements, and then extending the system to a large area, it effectively addresses the technical pain points of traditional large-area temperature field monitoring, such as high cost, poor real-time performance, and insufficient prediction accuracy. This invention combines meteorological factors and spatial heat conduction effects to construct a composite prediction model. Through multiple rounds of iterative parameter correction and optimization, it significantly improves the accuracy of temperature field prediction, ensuring that the prediction results closely match the actual temperature distribution patterns. This invention eliminates the need for full-area field measurements; it achieves efficient prediction of large-area temperature fields solely through pilot data modeling combined with meteorological corrections. This significantly reduces the manpower and material resources required for monitoring and improves the real-time performance and feasibility of temperature field data acquisition. Attached Figure Description
[0054] Figure 1 This is a flowchart of this embodiment. Detailed Implementation
[0055] like Figure 1 As shown, this embodiment selects a small area as a pilot area from a large region. The pilot area and the large region are highly homogeneous in terms of topography, soil texture, crop cover, and farming methods. The meteorological data of the pilot area can characterize the spatial distribution pattern of meteorological elements in the large region during the same period.
[0056] The solution in this embodiment is applicable to contiguous areas with relatively flat terrain, such as plains and gentle slopes, but not to fragmented areas with significant terrain differences, such as mountains and valleys.
[0057] The pilot area is defined by a rectangle, indicating its geographical boundaries. , area .
[0058] The pilot area will be divided into A uniform spatial grid, with grid side length... , Then the center coordinates of each grid are , Each small grid serves as a temperature prediction unit, facilitating subsequent spatial discretization of the temperature field.
[0059] Then, meteorological data was collected from the pilot area. Specifically, small automatic weather stations were deployed in the central area of the pilot area. The collected meteorological data included the daily average temperature of the pilot area. Total solar radiation in pilot areas Wind speed in pilot areas relative humidity in the pilot area .
[0060] The meteorological data collection frequency in the pilot area is once per hour, and the forecast period is [missing information]. Day, total acquisition Group time series data.
[0061] Simultaneously, soil temperature sensors are installed at the center of each small grid, buried at a depth of 10cm, to measure the surface temperature of the pilot area and simultaneously collect the actual temperature of the pilot area. , Indicates the first )Grid at time The actual measured temperature.
[0062] Considering the presence of missing and anomalies in the collected pilot area data, this embodiment also performs preprocessing on the collected pilot area data to ensure the accuracy of subsequent modeling, specifically including missing value processing, outlier processing, and normalization processing.
[0063] For missing values, linear interpolation is used to fill in missing data in the time series; for outliers, the 3σ criterion is used to remove outliers; and for normalization, meteorological data is normalized to the [0,1] interval to eliminate the influence of dimensions.
[0064] Then, based on the processed meteorological data from the pilot area, a temperature field prediction model was constructed, which adopted a multiple linear regression model:
[0065] ,in Indicates the first )Grid at time The predicted temperature; The regression coefficients to be solved; For random error, It follows a normal distribution.
[0066] To solve for the regression coefficients, the least squares method is used, with the core objective being to minimize the sum of squared errors.
[0067]
[0068] Transform the regression model into matrix form. ,in:
[0069] , ,
[0070] Finally, the regression coefficients for each grid cell are solved using the least squares formula:
[0071]
[0072] The regression coefficients for each grid are calculated to complete the time-series temperature prediction for a single grid.
[0073] Single-grid time-series prediction does not consider the heat conduction effect between grids, resulting in a discontinuous spatial distribution of the temperature field. Therefore, it needs to be corrected by incorporating the heat conduction equation. The heat conduction equation for the soil surface (10cm) temperature field is as follows:
[0074]
[0075] in, The specific value is the soil thermal conductivity, which is determined based on the soil type in the pilot area. This is a soil heat generation term, mainly derived from solar radiation. , For soil absorption rate, Temperature acquisition depth; , For temperature at , Spatial gradient in direction.
[0076] Then, the heat conduction equation is discretized using the finite difference method. The discretization formula for the mesh is:
[0077]
[0078] After processing, the mesh temperature after thermal conductivity correction is obtained:
[0079]
[0080] in, For the first Initial time-series predicted temperature of four adjacent grid cells The boundary mesh uses extrapolation to supplement boundary values.
[0081] Considering the possibility of discrepancies between the predicted results and the actual data, this embodiment also provides an error correction step for improving the accuracy of the error, as follows:
[0082] First, the core error metrics include root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination. The core error metrics used to evaluate prediction accuracy include:
[0083] The root mean square error (RMSE) reflects the overall deviation between the predicted and actual values; the smaller the RMSE, the higher the accuracy.
[0084]
[0085] The mean absolute error (MAE) reflects the average deviation of the predicted values and is more resistant to outlier interference.
[0086]
[0087] The coefficient of determination This reflects the model's goodness of fit. The closer the result is to 1, the better the fit.
[0088]
[0089] in, , which is the average value of the actual temperature field.
[0090] Then, a precision threshold is preset. Specifically, in this embodiment, the preset precision threshold is: In other embodiments, the preset accuracy threshold can also be set to other values.
[0091] If the preset accuracy threshold is met simultaneously, no accuracy correction is needed; if any one of them is not met, it indicates that the model accuracy is insufficient and further correction is required. Specifically, the gradient of the regression coefficients should be calculated first. , Indicates the first The next iteration; then update the regression coefficients: , The learning rate is set to 0.001-0.01 to avoid iteration oscillations; then the soil thermal conductivity is adjusted. , The correction factor is adjusted according to the direction of the error. If the predicted values are generally too high, then... If it is too low, then Finally, repeat these steps, iteratively correcting until all three error indicators meet the preset thresholds, to obtain the optimized final prediction model and the temperature field prediction data for the pilot area. .
[0092] For large areas, grid partitioning is also necessary, dividing the large area into... One grid, The center coordinates of each large grid are , , .
[0093] Meteorological data for a large regional grid, including the average daily temperature of the large region, was obtained through methods such as regional meteorological stations and remote sensing inversion. Total solar radiation over a large area Large area wind speed and the relative humidity of the large area .
[0094] Considering that the acquired regional meteorological data are regional averages, a meteorological correction factor is introduced. ,in The average daily temperature in the pilot area.
[0095] Similarly, for the total radiation of a large area For large-area wind speeds For relative humidity .
[0096] Finally, the acquired regional meteorological data is fed into the constructed temperature field prediction model to obtain the predicted temperature field data for the large region:
[0097]
[0098] in, The thermal conductivity correction term is calculated using the same method as in the pilot area, employing the regional soil thermal conductivity. Considering that random errors have been offset in the pilot model iterations, and that the heat conduction correction term is the spatial correction term needed for large-area temperature field prediction, the heat conduction correction term is adopted.
[0099] It should be understood that the descriptions of directions or positional relationships such as up, down, left, right, front, back, top, bottom, tail, horizontal and vertical in this application are all based on the accompanying drawings in the specification and are only used to express the technical solution more clearly and simplify the description, rather than indicating or implying that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the scope of protection of this application.
[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.
Claims
1. A regional temperature field prediction system based on big data analysis, characterized in that, Includes the following steps: S1: Select a small pilot area within the large area and divide the pilot area into grids; S2: Collect meteorological data for the pilot area, including the average daily temperature of the pilot area. Total solar radiation in pilot areas Wind speed in pilot areas relative humidity in the pilot area Simultaneously, the actual temperature of the pilot area was collected. , Indicates the first )Grid at time The measured temperature; S3: Preprocess the data collected in S2; S4: Based on the meteorological data in S3, construct a temperature field prediction model: ,in Indicates the first )Grid at time The predicted temperature The regression coefficients to be solved are... This is random error; S5: Solve for the regression coefficients; S6: Divide the area into large-area grids and acquire large-area meteorological data, including the large-area daily average temperature. Total solar radiation over a large area Large area wind speed and the relative humidity of the large area , For the large region Grid; S7: Based on the large-area meteorological data obtained in S6 and the temperature field prediction model constructed in S4, the large-area temperature field prediction results are obtained: in, This is a correction term for heat conduction.
2. The regional temperature field prediction system based on big data analysis according to claim 1, characterized in that, Small weather stations were set up in the pilot area to obtain meteorological data for the pilot area.
3. The regional temperature field prediction system based on big data analysis according to claim 1, characterized in that, In S2, the meteorological data collection frequency for the pilot area is once per hour, with a collection cycle of... Day, total acquisition Group time series data.
4. The regional temperature field prediction system based on big data analysis according to claim 1, characterized in that, Preprocessing in S3 includes: Missing values are handled by using linear interpolation to fill in missing data in the time series. Outlier handling: Outliers are removed using the 3σ criterion. And normalization processing, normalizing the meteorological data to... Intervals, eliminating the influence of dimensions.
5. The regional temperature field prediction system based on big data analysis according to claim 1, characterized in that, Step S5, solving for the regression coefficients, includes the following steps: First, minimize the sum of squared errors: Then, the temperature field prediction model is transformed into matrix form. ,in: ， ， Finally, the regression coefficients for each pilot area grid are solved using the least squares formula:
6. The regional temperature field prediction system based on big data analysis according to claim 5, characterized in that, It also includes a heat conduction correction step that takes into account space heat conduction, as follows: Construct the heat conduction equation for the temperature field in the pilot area: in, Soil thermal conductivity; For soil heat generation, , For soil absorption rate, Temperature acquisition depth; , For temperature at , Spatial gradient in direction; The heat conduction equation is discretized using the finite difference method. For the ... The discretization formula for the mesh is: After processing, the mesh temperature after thermal conductivity correction is obtained: in, For the first Initial time-series predicted temperature of four adjacent grid cells .
7. A regional temperature field prediction system based on big data analysis according to claim 6, characterized in that, It also includes error correction steps that take into account the accuracy of the error, as follows: First, select the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination. As an error indicator: in, , which is the average value of the actual temperature field; Preset precision threshold: If the preset accuracy threshold is met simultaneously, no accuracy correction is needed; if any one of them is not met, the regression coefficient is corrected. At the same time, adjust the soil thermal conductivity The steps are as follows: S'1: Calculate the gradient of the regression coefficients ; S'2: Update regression coefficients , The learning rate is set to 0.001-0.
01. S'3: Adjusting soil thermal conductivity , This is a correction factor; S'4: Repeat S'1-S'3 until the error meets the preset threshold.
8. A regional temperature field prediction system based on big data analysis according to claim 7, characterized in that, Preset accuracy threshold The values are 1.5℃, 1.0℃, and 0.85, respectively.
9. A regional temperature field prediction system based on big data analysis according to claim 1, characterized in that, The acquired regional meteorological data is the regional mean, and a meteorological correction factor is introduced. ,in The average daily temperature in the pilot area; For the total radiation of a large area For large-area wind speeds For relative humidity .