Landfill heat conduction prediction error correction method and system based on deep learning

Abstract

The application discloses a landfill heat conduction prediction error correction method and system based on deep learning, relates to the field of prediction error correction, and aims to solve the prediction deviation problem caused by factors such as differences in layered thermal properties of a traditional model. The method first constructs a time-space two-dimensional data acquisition framework through dynamic matching of a prediction period and an initial monitoring interval. Secondly, a vertical layered one-dimensional heat conduction model is established to predict temperature and calculate the error with the measured value. Gradient boosting decision tree and other algorithms are used to construct an error mapping model with influencing factors and the number of layers as input and the prediction error as output. In actual application, the error model is used to dynamically compensate the initial prediction result of the physical model, significantly improving the temperature prediction accuracy. The scheme ultimately provides reliable data support for the optimization of landfill monitoring strategies.

Description

Technical Field

[0001] This invention belongs to the field of prediction error correction, and more specifically, it relates to a method and system for predicting and correcting landfill heat conduction errors based on deep learning. Background Technology

[0002] Traditional landfill temperature prediction models suffer from several problems: they neglect differences in the thermal properties of different layers, leading to biases in shallow or deep temperature predictions; static monitoring intervals cannot adapt to the data lag caused by dynamic changes in waste degradation; the lack of error correction mechanisms results in long-term drift of model parameters; and existing landfill temperature prediction and assessment rely on empirical parameters and lack data-driven decision support. Summary of the Invention

[0003] In view of the problems in the background technology, the present invention proposes a method and system for correcting the prediction error of landfill heat conduction based on deep learning, so as to overcome the above-mentioned technical problems existing in the existing related technologies.

[0004] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0005] This invention relates to a deep learning-based method for correcting prediction errors in landfill heat conduction, comprising the following steps:

[0006] S1. Set the initial landfill monitoring interval to monitor and collect data on factors affecting the internal temperature prediction error of multiple landfills.

[0007] S2. Using a pre-constructed vertical one-dimensional heat conduction model of the landfill, the temperature data of each layer in multiple landfills in S1 are predicted, and the error data of each layer in the prediction results and the actual measurement data, as well as the corresponding layer number, are calculated.

[0008] S3. Based on the influencing factor data collected in S1 and the error data collected in S2, and the corresponding layer number, construct the final landfill inner layer temperature prediction error mapping model.

[0009] S4. Predict the temperature data of each layer of the landfill to be predicted, and input the corresponding temperature prediction error influencing factors into the mapping model in S3 for mapping.

[0010] S5. Collect short-term verification data of the current landfill, and use the mapping results in S4 to compensate for the error of the prediction results in S4; then determine whether the temperature prediction accuracy meets the requirements based on the compensation results; if it meets the requirements, output the prediction results; if it does not meet the requirements, execute S6.

[0011] S6. Adjust the initial landfill monitoring interval described in S1 multiple times and repeat S1, S2, S3, S4 and S5 to construct the final adjusted and compensated inner layer temperature prediction mapping model; adjust the landfill monitoring interval multiple times and input it into the final adjusted and compensated inner layer temperature prediction mapping model for mapping; output the final temperature prediction result based on the mapping result.

[0012] Preferably, step S1 includes the following steps:

[0013] S11. Define several types of factors that affect the error in predicting the internal temperature of a landfill and construct a one-dimensional heat conduction model in the landfill that considers vertical stratification. This will yield a set of factors that affect the internal temperature prediction error and a one-dimensional vertical heat conduction model of the landfill.

[0014] S12. Select several existing landfills and their corresponding monitoring cycles to obtain a set of existing landfills and a set of existing monitoring cycles; then set the monitoring interval for the existing initial landfills.

[0015] Based on the existing initial landfill monitoring interval, existing monitoring cycle set, and internal temperature prediction error influencing factor type set, the data of internal temperature prediction error influencing factors corresponding to each landfill in the existing landfill set are monitored and collected to obtain the existing internal temperature prediction error influencing factor dataset.

[0016] By systematically identifying and monitoring the multidimensional influencing factors of temperature prediction errors within landfills, and establishing a one-dimensional heat conduction model based on vertical stratification, more accurate prediction and control of landfill thermodynamic behavior were achieved. The one-dimensional heat conduction model based on vertical stratification effectively addressed the differences in thermal properties between the waste layer and the cover layer, as well as the spatiotemporal heterogeneity of degradation heat generation, through analytical solution methods. This significantly improved the spatiotemporal resolution and reliability of landfill temperature field prediction, providing a scientific basis for waste degradation control, heat recovery, and environmental risk assessment.

[0017] Preferably, constructing a one-dimensional heat conduction model considering vertical stratification in the landfill in S11 includes the following steps:

[0018] S111. Construct the one-dimensional heat transport equation for the i-th layer of the landfill; as follows.

[0019] ;

[0020] In the formula: Let t be the temperature in the i-th layer, and t be the time. Let be the thermal conductivity coefficient in the z-direction of the i-th layer; Let be the rate of heat change caused by the degradation of municipal solid waste per unit volume in the i-th layer, i.e., the degradation heat generation rate. When the i-th layer is an intermediate cover layer... =0; Let be the specific heat capacity of the i-th layer; Let n be the density of the i-th layer, n be the maximum number of layers, and K represent the K-th layer. The parameter represents the thermal response sensitivity or heating efficiency. Represents the vertical coordinate or depth from a reference plane; , This indicates the depth of a certain reference plane downwards. Located inside the i-th layer;

[0021] S112. Consider the two temperature conditions at the top of the landfill and set the top boundary conditions, the bottom boundary conditions, and the conditions for liquid phase, gas phase, and heat continuity between adjacent layers.

[0022] S113. Based on the two boundary conditions and continuity conditions set in S112, the analytical solution of the one-dimensional heat transport equation in the i-th layer of the landfill in S111 is solved by superposition method, separation of variables and orthogonal expansion method.

[0023] By constructing a one-dimensional heat conduction model considering vertical stratification and employing analytical methods, efficient prediction of the landfill temperature field was achieved. First, the established stratified heat conduction equations, by introducing differentiated thermophysical parameters and degradation heat generation rates between layers, reflect the differences in heat conduction characteristics between the waste layer and the cover layer. Second, by setting two top boundary conditions and a bottom isothermal boundary, the model flexibly adapts to the climatic impacts of different landfill scenarios. The analytical solutions are obtained using the superposition method, the separation of variables method, and the orthogonal expansion method, which not only avoids the stability problems of numerical methods but also ensures computational efficiency through the orthogonality of characteristic functions and series expansion.

[0024] Preferably, step S2 includes the following steps:

[0025] S21. Based on the initial landfill monitoring interval and the existing monitoring cycle set, collect temperature data of each layer in each landfill in the existing landfill set to obtain the existing landfill inner layer temperature dataset; then use the landfill vertical one-dimensional heat conduction model to predict the temperature data of each layer in each landfill in the existing landfill set to obtain the existing landfill inner layer temperature prediction dataset.

[0026] S22. Based on the existing landfill inner layer temperature dataset and the existing landfill inner layer temperature prediction dataset, calculate the average prediction error of the temperature of each layer in each landfill in the existing landfill set and the corresponding number of layers, to obtain the existing landfill inner layer temperature prediction error dataset and the landfill prediction error collection layer set.

[0027] Through a systematic data acquisition and model validation process, the accuracy of landfill temperature prediction was quantitatively evaluated and optimized. A vertically layered one-dimensional heat conduction model was used for temperature prediction. By comparing measured and predicted data, the average prediction error of each layer's temperature could be quantified, revealing the deviation pattern of the model under specific working conditions. Finally, data support was provided for the subsequent construction of a deep learning-based method for correcting landfill heat conduction prediction errors and a prediction error mapping model for the system model.

[0028] Preferably, step S3 includes the following steps:

[0029] S31. Based on the existing landfill inner layer temperature prediction error dataset, the landfill prediction error collection layer set, and the existing inner temperature prediction error influencing factor dataset, construct a mapping model between the landfill prediction error collection layer, the inner temperature prediction error influencing factor data, and the landfill inner layer temperature prediction error data to obtain the final landfill inner layer temperature prediction error mapping model.

[0030] By constructing a mapping model, the accurate source tracing and dynamic correction of landfill temperature prediction errors were achieved. The model supports dynamic parameter inversion and automatically compensates for the prediction errors of the original model by real-time monitoring data, thereby significantly reducing long-term prediction errors.

[0031] Preferably, the final landfill inner layer temperature prediction error mapping model in S31 adopts the GBDT gradient boosting decision tree model;

[0032] The GBDT gradient boosting decision tree model supports the fusion of multi-source heterogeneous data and is robust to missing and outlier values, making it suitable for the characteristics of field engineering data. When the amount of data is limited or rapid deployment is required, GBDT is easier to tune and has high computational efficiency.

[0033] Preferably, step S4 includes the following steps:

[0034] S41. Set the current landfill to be predicted and preset the prediction period to obtain the current initial preset prediction period;

[0035] S42. Based on the current initial preset prediction cycle, initial landfill monitoring interval and internal temperature prediction error influencing factor type set, collect the internal temperature prediction error influencing factor data corresponding to the current landfill to be predicted, and obtain the current internal temperature prediction error influencing factor dataset.

[0036] Then, the vertical one-dimensional heat conduction model of the landfill is used to predict the temperature data of each layer of the current landfill to be predicted, so as to obtain the temperature prediction dataset of the inner layer of the current landfill.

[0037] S43. The current internal temperature prediction error influencing factor dataset is combined with the number of each layer of the current landfill to be predicted and input into the final landfill internal temperature prediction error mapping model for mapping to obtain the current landfill internal temperature prediction error dataset.

[0038] By integrating multi-stage data and model collaboration, the accuracy and reliability of landfill temperature prediction were improved. When using a vertically layered one-dimensional heat conduction model for temperature prediction, the model analyzes the interaction effects of thermal property parameters and boundary conditions of each layer through orthogonal expansion. The collaborative application of its prediction results and error mapping model provides a dynamic correction basis for prediction reliability assessment.

[0039] Preferably, step S5 includes the following steps:

[0040] S51. The current landfill inner layer temperature prediction dataset is compensated using the current landfill inner layer temperature prediction error dataset to obtain the current compensated inner layer temperature prediction dataset.

[0041] S52. Collect current short-term verification data of the landfill. Based on the temperature prediction accuracy requirements, evaluate the current compensated inner layer temperature prediction dataset to obtain an initial evaluation result. If the initial evaluation result shows that the temperature prediction accuracy meets the requirements, output the prediction result; otherwise, execute S6. Through the dual mechanism of prediction error compensation and temperature prediction accuracy evaluation, the scientific decision-making and reliability improvement of landfill temperature prediction are realized. First, the temperature prediction compensation based on the error dataset adopts dynamic correction. By quantifying the error distribution of each layer, the predicted temperature is calibrated layer by layer, so that the mean square error between the compensated dataset and the measured data is significantly reduced.

[0042] Preferably, step S6 includes the following steps:

[0043] S61. Set the cumulative number of data adjustment repetitions; adjust the initial landfill monitoring interval in S12 repeatedly according to the cumulative number of data adjustment repetitions. After each adjustment, repeat S12, S21, S22, S31, S41, S42, S43 and S51 and record the adjusted landfill monitoring interval to obtain the current adjusted and compensated inner layer temperature prediction dataset and the current landfill monitoring interval adjustment dataset.

[0044] S62. Based on the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset, construct a mapping model between the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset to obtain the final adjusted and compensated inner layer temperature prediction mapping model.

[0045] S63. Adjust the current landfill monitoring interval adjustment data multiple times. After each adjustment, input the adjusted current landfill monitoring interval adjustment data into the final adjusted and compensated inner layer temperature prediction mapping model for mapping to obtain the current model verification and compensated inner layer temperature prediction dataset. Continue until the current model verification and compensated inner layer temperature prediction dataset shows that it meets the temperature prediction accuracy requirements, and output the current model verification and compensated inner layer temperature prediction dataset.

[0046] By verifying and adjusting the system, and by coupling the verification results with professional evaluation standards, the system has a low error rate. Through the triple guarantee of data, model and verification, a robust and economical technical path is provided for the energy utilization of landfills.

[0047] The deep learning-based landfill heat conduction prediction error correction system includes a data acquisition module for factors affecting the prediction error of the existing landfill inner layer, a data acquisition module for the prediction error of the existing landfill inner layer, a landfill inner layer temperature prediction error mapping model construction module, a current landfill inner layer temperature prediction error mapping module, a current landfill temperature prediction initial determination module, and a current landfill temperature prediction final determination module.

[0048] The present invention has the following beneficial effects:

[0049] 1. This invention achieves a significant improvement in the accuracy of landfill temperature prediction and optimization of system reliability through multi-stage data fusion and model collaboration. First, a time-space dual-dimensional data acquisition framework is constructed by dynamically matching the preset prediction period with the initial monitoring interval. Second, when using a vertically layered one-dimensional heat conduction model for temperature prediction, the model analyzes the interaction effects of each layer's thermophysical parameters and boundary conditions through orthogonal expansion, and its prediction results are synergistically applied with the error mapping model. The error mapping model identifies key influencing factors through feature importance analysis, greatly reducing the temperature prediction error throughout the entire prediction period. Finally, this solution provides reliable data support for optimizing landfill monitoring strategies.

[0050] 2. In this invention, the layered heat conduction equation, by introducing differentiated thermophysical parameters and degradation heat generation rates between layers, reflects the differences in heat conduction characteristics between the waste layer and the cover layer. Secondly, by setting two top boundary conditions and a bottom isothermal boundary, it flexibly adapts to the climatic influences of different landfill scenarios. Furthermore, by employing the superposition method, the separation of variables method, and the orthogonal expansion method to solve the analytical solution, it not only avoids the stability problem of numerical methods but also ensures computational efficiency through the orthogonality of characteristic functions and series expansion.

[0051] 3. In this invention, the initial monitoring interval is iteratively optimized by adjusting the number of repetitions based on accumulated data. Combined with the full-process data acquisition and model correction from S12 to S51, the spatial resolution of the compensated temperature prediction dataset is greatly improved. At the same time, by recording the mapping relationship between the adjustment interval and the prediction results, an adaptive balance model of monitoring frequency and prediction accuracy is constructed. The design of the maximum number of repetitions is verified, and the coupling analysis of the verification results and professional evaluation standards significantly reduces the system's misjudgment rate. Through the triple guarantee of data, model, and verification, a robust and economical technical path is provided for the accurate prediction and long-term monitoring of landfill thermal behavior.

[0052] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0053] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The accompanying drawings in the description are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating the deep learning-based method for correcting prediction errors in landfill heat conduction according to the present invention.

[0055] Figure 2 This is a comparison chart of temperature variation patterns in the Wuxi landfill according to the present invention;

[0056] Figure 3 This is a comparison chart of the actual measured temperature at a depth of 7m in the landfill in Zone B of the Michigan landfill and the model calculation results.

[0057] Figure 4 This is a comparison chart of the actual measured temperature at a depth of 10m in the landfill area B of the Michigan landfill and the model calculation results. Detailed Implementation

[0058] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0059] Example 1

[0060] Please see Figure 1 This embodiment describes a deep learning-based method for correcting prediction errors in landfill heat conduction, comprising the following steps:

[0061] S1. Set the initial landfill monitoring interval to monitor and collect data on factors affecting the internal temperature prediction error of multiple landfills.

[0062] S1 includes the following steps:

[0063] S11. Define several types of factors influencing the error magnitude of landfill internal temperature prediction and construct a one-dimensional heat conduction model considering vertical stratification in the landfill. This yields a set of factors influencing internal temperature prediction error and a one-dimensional vertical heat conduction model of the landfill. The influencing factors include atmospheric temperature, waste moisture content, organic matter content, and cover layer condition. The model aims to accurately characterize the differences in heat conduction behavior between the waste layer and the cover layer due to differences in thermal properties and degradation heat generation rates.

[0064] The process of constructing a one-dimensional heat conduction model for the landfill considering vertical stratification in S11 is as follows:

[0065] The one-dimensional heat transport equation in the i-th layer of the landfill is:

[0066] (4.3.1)

[0068] In equation (4.3.1): T i Let λ be the temperature (K) in the i-th layer; z,i is the thermal conductivity (W / m / k) in the z-direction of the i-th layer. Let J be the rate of heat change per unit volume of municipal solid waste degradation in the i-th layer, i.e., the degradation heat production rate (J / m³). 3 / s), when the i-th layer is an intermediate overlay layer =0; Represents the vertical coordinate or depth from a reference plane; , This indicates the depth of a certain reference plane downwards. Located inside the i-th layer

[0069] (4.3.2)

[0070] In equation (4.3.2): The specific heat capacity of the i-th layer (J / m³) 3 / K); The density of the i-th layer (kg / m³) 3 ); The parameter represents the thermal response sensitivity or heating efficiency.

[0071] Consider two temperature conditions at the top:

[0072] Working Case 1: Considering the heat exchange between the landfill topsoil and the atmosphere, if there is a cover layer on top of the landfill, the top boundary conditions of the thermal model are:

[0073] (4.3.3)

[0074] In equation (4.3.3): ht c The heat exchange coefficient between the topsoil of the landfill and the outside air (J / m²) 2 / s / K), T air (t) represents the atmospheric temperature (K).

[0075] Condition 2: Heat exchange between the landfill top and the atmosphere is not considered. If there is no overburden layer on top of the landfill, the waste soil is in direct contact with the atmosphere. In this case, it is generally assumed that the temperature at the top of the landfill is consistent with the temperature of the atmosphere. The top boundary condition of the thermal model is:

[0076] (4.3.4)

[0077] In equation (4.3.4): T air (t) represents the atmospheric temperature (K).

[0078] The temperature at the bottom of a landfill is generally considered to be relatively constant, and the boundary conditions for the lower boundary of the thermal model are as follows:

[0079] (4.3.5)

[0080] In equation (4.3.5): T bot This refers to the temperature at the bottom of the landfill.

[0081] The following conditions must be met for continuity of liquid, gas, and heat between adjacent layers:

[0082] (4.3.6)

[0083] (4.3.7)

[0084] The initial conditions are:

[0085] (4.3.8)

[0086] In equation (4.3.8): T 0,i (z) represents the initial temperature (K) in the i-th layer;

[0087] The analytical solution of a one-dimensional heat conduction model in a landfill is obtained by using the superposition method, separation of variables, and orthogonal expansion method.

[0088] For T iTo solve the (z,t) nonlinear boundary condition, the superposition method is first used to solve for T. i (z,t) boundary homogenization, let T i The solution for (z,t) is:

[0089] (4.3.9)

[0090] w i (z,t) satisfies the following governing equations:

[0091] (4.3.10)

[0092] w i (z,t) and T i (z,t) satisfy the same non-homogeneous boundary conditions:

[0093] (4.3.11)

[0094] (4.3.12)

[0095] (4.3.13)

[0096] w i (z,t) and T(z,t) satisfy the same continuity condition:

[0097] (4.3.14)

[0098] (4.3.15)

[0099] Get w i The solution for (z,t) is:

[0100] (4.3.16)

[0101] When the boundary conditions are (4.3.11) and (4.3.13):

[0102] (4.3.17)

[0103] When the boundary conditions are (4.3.12) and (4.3.13):

[0104] (4.3.18)

[0105] v i (z,t) satisfies the following governing equations:

[0106] (4.3.19)

[0107] In equation (4.3.19):

[0108] (4.3.20)

[0109] v i (z,t) satisfies T i The homogeneous boundary conditions corresponding to (z,t) are:

[0110] (4.3.21)

[0111] (4.3.22)

[0112] (4.3.23)

[0113] v i (z,t) satisfies T i (z,t) have the same continuity condition:

[0114] (4.3.24)

[0115] (4.3.25)

[0116] v i The initial conditions for (z,t) are:

[0117] (4.3.26)

[0118] Solve v using the orthogonal expansion method i (z,t) is used to obtain v i The characteristic function of (z,t) is let f i (z,t) satisfies the homogeneous form of the governing equation (4.3.19):

[0119] (4.3.27)

[0120] f i The boundary conditions and continuity conditions of (z,t) are related to v i (z,t) are the same:

[0121] (4.3.28)

[0122] (4.3.29)

[0123] (4.3.30)

[0124] (4.3.31)

[0125] (4.3.32)

[0126] Solve f using the method of separation of variables i (z,t), let:

[0127] (4.3.33)

[0128] Substituting equation (4.3.33) into equation (4.3.27), we obtain the following two ordinary differential equations with respect to z and t, respectively:

[0129] (4.3.34)

[0130] (4.3.35)

[0131] In equations (4.3.34) and (4.3.35): ξ i is the separation constant.

[0132] Solve the ordinary differential equations (4.3.34) and (4.3.35), f i The solution to (z,t) can be expressed in the following series form:

[0133] (4.3.36)

[0134] In equation (4.3.36):

[0135] (4.3.37)

[0136] (4.3.38)

[0137] ξ i,j (j=1,2,3,…) are the eigenvalues.

[0138] Substituting equation (4.3.36) into equation (4.3.28), we get:

[0139] (4.3.39)

[0140] Without loss of generality, C can be made 1,j =1. Similarly, when the top boundary is (4.3.29):

[0141] (4.3.40)

[0142] Substituting equation (4.3.36) into equations (4.3.31) and (4.3.32), we get:

[0143]

[0144] (4.3.41) (4.3.42)

[0145] (4.3.43)

[0146] (4.3.44)

[0147] Substituting equation (4.3.36) into equation (4.3.30), we get:

[0148] (4.3.45)

[0149] C can be obtained from equations (4.3.41) and (4.3.42). i,j D i,j With C i+1,j D i+1,j The recurrence relation between (i=1, 2,…, n-1) is as follows:

[0150] (4.3.46)

[0151] In equation (4.3.46):

[0152] (4.3.47)

[0153] Based on equation (4.3.47), we obtain C. 1,j D 1,j With C n,j D n,j Relationship:

[0154] (4.3.48)

[0155] In equation (4.3.48):

[0156] (4.3.49)

[0157] Substituting equations (4.3.39), (4.3.43), and (4.3.48) into equation (4.3.45), we obtain the equation regarding ξ. 1,j (or ξ) i,j The transcendental equations of (i=2,…,n) are given by equations whose solutions are greater than 0 and are called eigenvalues.

[0158] Using fi The characteristic function obtained from (z,t), v i The solution to (z,t) can be set as:

[0159] (4.3.50)

[0160] The characteristic function (4.3.38) satisfies the following orthogonality relation:

[0161] (4.3.51)

[0162] q v,i The orthogonal expansion of (z,t) with respect to the characteristic function is:

[0163] (4.3.52)

[0164] In equation (4.3.52):

[0165] (4.3.53)

[0166] Substituting equations (4.3.50) and (4.3.52) into equation (4.3.19) and simplifying, we obtain the following ordinary differential equation:

[0167] (4.3.54)

[0168] The solution to the ordinary differential equation (4.3.54) is:

[0169] (4.3.55)

[0170] Substituting equation (4.3.50) into the initial condition equation (4.3.26), and utilizing the orthogonality of the characteristic function equation (4.3.51), we obtain μ. j for:

[0171] (4.3.56)

[0172] Get T i The solution for (z,t) is:

[0173] (4.3.57)

[0174] In equation (4.3.57), v i (z,t) and w i The expressions for (z,t) are given in equations (4.3.50) and (4.3.16), respectively;

[0175] S12. Select several existing landfills and their corresponding monitoring cycles to obtain a set of existing landfills and a set of existing monitoring cycles; then set the monitoring interval for the existing initial landfills.

[0176] Based on the existing initial landfill monitoring interval, existing monitoring cycle set, and internal temperature prediction error influencing factor type set, the data of internal temperature prediction error influencing factors corresponding to each landfill in the existing landfill set are monitored and collected to obtain the existing internal temperature prediction error influencing factor dataset.

[0177] By systematically identifying and monitoring the multidimensional influencing factors of landfill internal temperature prediction errors, and establishing a one-dimensional heat conduction model based on vertical stratification, more accurate prediction and control of landfill thermodynamic behavior were achieved. Specifically, firstly, the constructed set of influencing factors for internal temperature prediction errors includes atmospheric temperature, waste moisture content, organic matter content, and cover layer status, achieving quantitative compensation for the dynamic boundary conditions and material non-uniformity neglected by traditional models. Secondly, the one-dimensional heat conduction model based on vertical stratification effectively addresses the differences in thermal properties between the waste layer and the cover layer, as well as the spatiotemporal heterogeneity of degradation heat generation, through analytical solution methods. Its verification results are consistent with measured data from Wuxi and Michigan landfills. Figure 2 , Figure 3 , Figure 4 The model's engineering applicability under simplified conditions was demonstrated. Finally, through standardized monitoring process data-driven model optimization, the spatiotemporal resolution and reliability of landfill temperature field prediction were significantly improved, laying a data foundation for the subsequent construction of a high-precision error correction model.

[0178] S2. Using a pre-constructed vertical one-dimensional heat conduction model of the landfill, the temperature data of each layer in multiple landfills in S1 are predicted, and the error data of each layer in the prediction results and the actual measurement data, as well as the corresponding layer number, are calculated.

[0179] S2 includes the following steps:

[0180] S21. Based on the initial landfill monitoring interval and the existing monitoring cycle set, collect temperature data of each layer in each landfill in the existing landfill set to obtain the existing landfill inner layer temperature dataset; then use the landfill vertical one-dimensional heat conduction model to predict the temperature data of each layer in each landfill in the existing landfill set to obtain the existing landfill inner layer temperature prediction dataset.

[0181] S22. Based on the existing landfill inner layer temperature dataset and the existing landfill inner layer temperature prediction dataset, calculate the average prediction error of the temperature of each layer in each landfill in the existing landfill set and the corresponding number of layers, to obtain the existing landfill inner layer temperature prediction error dataset and the landfill prediction error collection layer set.

[0182] Through a systematic data acquisition and model validation process, the accuracy of landfill temperature prediction was quantitatively evaluated and optimized. First, by collecting measured temperature data from each layer of the landfill at preset monitoring intervals, a multi-dimensional temperature dataset covering different climatic conditions and landfill structures was constructed, providing benchmark data in a real-world scenario for model validation. Second, a vertically layered one-dimensional heat conduction model was used for temperature prediction. By comparing measured and predicted data, the average prediction error of each layer's temperature could be quantified, revealing the model's deviation patterns under specific operating conditions. Finally, data support was provided for the subsequent construction of a deep learning-based landfill heat conduction prediction error correction method and a prediction error mapping model for the system model.

[0183] S3. Based on the influencing factor data collected in S1 and the error data collected in S2, and the corresponding layer number, construct the final landfill inner layer temperature prediction error mapping model.

[0184] S3 includes the following steps:

[0185] S31. Based on the existing landfill inner layer temperature prediction error dataset, the landfill prediction error collection layer set, and the existing inner temperature prediction error influencing factor dataset, construct a mapping model between the landfill prediction error collection layer, the inner temperature prediction error influencing factor data, and the landfill inner layer temperature prediction error data to obtain the final landfill inner layer temperature prediction error mapping model.

[0186] The final landfill inner layer temperature prediction error mapping model described in S31 adopts the GBDT gradient boosting decision tree model.

[0187] S31 includes the following steps:

[0188] S311. Construct an initial landfill inner layer temperature prediction error mapping model and set a first training data ratio, such as 8:2 or 7:3, which can be adjusted adaptively according to the actual training situation; divide the existing landfill inner layer temperature prediction error dataset, landfill prediction error collection layer set and existing inner temperature prediction error influencing factor dataset according to the first training data ratio to obtain the first training dataset and the first test dataset.

[0189] S312. Set a first training error threshold, such as 10%~15%, which can be adjusted adaptively according to the actual training situation; input the first training dataset into the initial landfill inner layer temperature prediction error mapping model for training; during the training process, if the training error is less than the first training error threshold, stop training and obtain the trained landfill inner layer temperature prediction error mapping model; otherwise, continue training until the training error is less than the first training error threshold.

[0190] S313. Set a first test accuracy threshold, such as 90%~95%, which can be adjusted adaptively according to the actual test situation; input the first test dataset into the trained landfill inner layer temperature prediction error mapping model for testing; after the test is completed, obtain the first test accuracy data; if the first test accuracy data is greater than or equal to the first test accuracy threshold, obtain the final landfill inner layer temperature prediction error mapping model; otherwise, return to S312 to continue training the trained landfill inner layer temperature prediction error mapping model and repeat S313 until the first test accuracy data is greater than or equal to the first test accuracy threshold.

[0191] The structure of the initial landfill inner layer temperature prediction error mapping model can be seen in Table 1 below:

[0192] Table 1

[0193] Model name Model type Model structure Initial landfill inner-layer temperature prediction error mapping model GBDT gradient boosting decision tree model Base learner: CART regression tree, maximum depth of tree set to 5-8 layers to balance overfitting risk; number of iterations: 500-1000 rounds, early stopping mechanism (earlystopping) to terminate when validation set loss does not decrease for 10 consecutive rounds; learning rate: 0.05-0.1, adaptive adjustment strategy used; feature importance evaluation: built-in gain calculation, automatic identification of key influencing factors such as number of collected layers;

[0194] The GBDT gradient boosting decision tree model naturally supports the fusion of multi-source heterogeneous data and is robust to missing and outlier values, making it suitable for the characteristics of engineering field data. When the amount of data is limited (<100,000 samples) or rapid deployment is required, GBDT is easier to tune parameters and has high computational efficiency.

[0195] By constructing a multi-source data-driven error mapping model based on GBDT, accurate quantification and effective compensation of the prediction error of the physical model were achieved. Specifically, the model takes the data on influencing factors such as the number of layers, atmospheric temperature, and waste moisture content collected in the aforementioned steps as input, and the average prediction error of the corresponding layer calculated in S22 as output. Through training, it learns complex nonlinear relationships that were not considered by the physical model. For example, its application at a depth of 13 meters in the Michigan landfill shows that the error mapping model has significantly better fitting accuracy for the prediction error. In practical applications (S4-S5), inputting the data of the current site to be predicted into this trained GBDT model can yield a high-precision error estimate, which can then dynamically compensate for the initial physical prediction results. Tests have shown that this compensation strategy can significantly reduce the long-term average absolute error of temperature prediction. Furthermore, by analyzing the feature importance ranking provided by the GBDT model, key factors that have a significant impact on prediction error, such as the moisture content at a specific depth, can be identified, thus providing a data-driven decision-making basis for optimizing sensor deployment and focusing on key monitoring parameters.

[0196] S4. Predict the temperature data of each layer of the landfill to be predicted, and input the corresponding temperature prediction error influencing factors into the mapping model in S3 for mapping.

[0197] S4 includes the following steps:

[0198] S41. Set the current landfill to be predicted and preset the prediction period to obtain the current initial preset prediction period;

[0199] S42. Based on the current initial preset prediction cycle, initial landfill monitoring interval and internal temperature prediction error influencing factor type set, collect the internal temperature prediction error influencing factor data corresponding to the current landfill to be predicted, and obtain the current internal temperature prediction error influencing factor dataset.

[0200] Then, the vertical one-dimensional heat conduction model of the landfill is used to predict the temperature data of each layer of the current landfill to be predicted, so as to obtain the temperature prediction dataset of the inner layer of the current landfill.

[0201] S43. The current internal temperature prediction error influencing factor dataset is combined with the number of each layer of the current landfill to be predicted and input into the final landfill internal temperature prediction error mapping model for mapping to obtain the current landfill internal temperature prediction error dataset.

[0202] Through multi-stage data fusion and model collaboration, the accuracy and reliability of landfill temperature prediction were improved. First, a time-space dual-dimensional data acquisition framework was constructed by dynamically matching the preset prediction cycle with the initial monitoring interval, such as daily or weekly sampling. Second, when using a vertically layered one-dimensional heat conduction model for temperature prediction, the model analyzes the interaction effects of each layer's thermal properties and boundary conditions using orthogonal expansion, and its prediction results are collaboratively applied with the error mapping model. Finally, this scheme provides reliable data support for optimizing landfill monitoring strategies.

[0203] S5. Collect short-term verification data of the current landfill, and use the mapping results in S4 to compensate for the error of the prediction results in S4; then determine whether the temperature prediction accuracy requirements are met based on the compensation results; if they are met, output the prediction results; if they are not met, execute S6.

[0204] S5 includes the following steps:

[0205] S51. The current landfill inner layer temperature prediction dataset is compensated using the current landfill inner layer temperature prediction error dataset to obtain the current compensated inner layer temperature prediction dataset.

[0206] S52. Collect short-term verification data of the current landfill, evaluate the current compensated inner layer temperature prediction dataset according to the temperature prediction accuracy requirements, and obtain the initial evaluation result; if the initial evaluation result shows that the temperature prediction accuracy meets the requirements, output the prediction result; otherwise, execute S6.

[0207] Through a dual mechanism of prediction error compensation and temperature prediction accuracy assessment, scientific decision-making and reliability improvement of landfill temperature prediction are achieved. First, temperature prediction compensation based on the error dataset employs dynamic correction. By quantifying the error distribution at each layer, the predicted temperature is calibrated layer by layer, significantly reducing the mean square error between the compensated dataset and the measured data. Second, short-term validation data from the current landfill, such as temperature data corresponding to various time intervals within two months, is collected to assess the accuracy of temperature prediction. This integrates industry standards and engineering experience, ensuring that the assessment results are both technically feasible and economically reasonable. The compensation-assessment-decision closed-loop feedback mechanism of this scheme can dynamically identify the reliability of landfill temperature prediction, providing data-driven decision-making basis for landfill thermal management. S6: The initial landfill monitoring interval described in S1 is adjusted multiple times, and S1, S2, S3, S4, and S5 are repeated to construct the final adjusted and compensated inner layer temperature prediction mapping model. The landfill monitoring interval is adjusted multiple times and input into the final adjusted and compensated inner layer temperature prediction mapping model for mapping. The temperature prediction result is output based on the mapping result. S6 includes the following steps:

[0208] S61. Set the cumulative number of data adjustment repetitions; adjust the initial landfill monitoring interval in S12 repeatedly according to the cumulative number of data adjustment repetitions. After each adjustment, repeat S12, S21, S22, S31, S41, S42, S43 and S51 and record the adjusted landfill monitoring interval to obtain the current adjusted and compensated inner layer temperature prediction dataset and the current landfill monitoring interval adjustment dataset.

[0209] S62. Based on the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset, construct a mapping model between the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset to obtain the final adjusted and compensated inner layer temperature prediction mapping model.

[0210] S62 includes the following steps:

[0211] S621. Construct an initial adjusted and compensated inner layer temperature prediction mapping model and set a second training data ratio, such as 8:2 or 7:3, which can be adaptively adjusted according to the actual training situation; divide the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset according to the second training data ratio to obtain the second training dataset and the second test dataset.

[0212] S622. Set a second training error threshold, such as 10%~15%, which can be adjusted adaptively according to the actual training situation; input the second training dataset into the initial adjusted and compensated inner layer temperature prediction mapping model for training; during the training process, if the training error is less than the second training error threshold, stop training and obtain the trained adjusted and compensated inner layer temperature prediction mapping model; otherwise, continue training until the training error is less than the second training error threshold.

[0213] S623. Set a second test accuracy threshold, such as 90%~95%, which can be adjusted adaptively according to the actual test situation; input the second test dataset into the trained adjusted and compensated inner layer temperature prediction mapping model for testing; after the test is completed, obtain the second test accuracy data; if the second test accuracy data is greater than or equal to the second test accuracy threshold, obtain the final adjusted and compensated inner layer temperature prediction mapping model; otherwise, return to S622 to continue training the trained adjusted and compensated inner layer temperature prediction mapping model and repeat S623 until the second test accuracy data is greater than or equal to the second test accuracy threshold;

[0214] The structure of the inner layer temperature prediction mapping model after the initial adjustment and compensation can be seen in Table 2 below:

[0215] Table 2

[0216] Model name Model type Model structure Initial adjustment compensation after inner-layer temperature prediction mapping model LSTM-Transformer hybrid neural network Temporal and spatial feature extraction module: LSTM layer: 128 units, bidirectional structure, capturing time series dependence of monitoring interval adjustment; 1D convolution layer: 64 filters, kernel size 3, step 1, ReLU activation function, extracting inter-layer spatial patterns; max pooling layer: pooling window 2, step 2, reducing dimension; feature fusion module: Transformer encoder: 4 attention mechanisms, hidden layer dimension 256, feedforward network dimension 512; position encoding: using sine function to encode time sequence position information; fully connected module: dense layer 1: 256 neurons, ReLU activation, Dropout rate 0.3; dense layer 2: 128 neurons, LeakyReLU (a=0.1); output layer: linear activation, single neuron output prediction compensation value; parameter settings: batch size: 32-64; optimizer: Adam (initial learning rate 0.001; loss function: Huber loss δ=1.0;

[0217] S63. Adjust the current landfill monitoring interval adjustment data multiple times. After each adjustment, input the adjusted current landfill monitoring interval adjustment data into the final adjusted and compensated inner layer temperature prediction mapping model for mapping to obtain the current model verification and compensated inner layer temperature prediction dataset. Continue until the current model verification and compensated inner layer temperature prediction dataset shows that it meets the temperature prediction accuracy requirements, and output the current model verification and compensated inner layer temperature prediction dataset.

[0218] Through a closed-loop mechanism of dynamic monitoring interval optimization and multi-round model validation, continuous optimization of landfill temperature prediction accuracy and reliable decision-making are achieved. First, the initial monitoring interval is iteratively optimized by adjusting the number of repetitions based on accumulated data. Combined with full-process data acquisition and model correction from S12 to S51, the spatial resolution of the compensated temperature prediction dataset is improved. Simultaneously, by recording the mapping relationship between the adjustment interval and the prediction results, an adaptive balance model of monitoring frequency and prediction accuracy is constructed. Second, the design of the maximum number of validations, such as 100 times, far exceeding the adjustment repetition count of 10, reduces the system's misjudgment rate through coupled analysis of the validation results and professional evaluation standards. It supports real-time triggering of termination conditions, such as determining infeasibility if a single validation fails to meet the standard. The final mapping model (S62) can further generate optimal monitoring strategy recommendations, improving the overall stability of the prediction results. Finally, this solution, through the triple guarantee of data, model, and validation, provides a robust and economical technical path for accurate landfill temperature prediction and efficient monitoring management.

[0219] For example, a temperature monitoring experiment was conducted in the newly filled municipal solid waste layer at the Taohuashan landfill in Wuxi City. The newly filled municipal solid waste layer in the test area was approximately 8.5 m thick, below which was an older layer of waste with a landfill age of approximately 10-20 years. The leachate level in the landfill was high, approximately 4 m below the top of the newly filled municipal solid waste layer. The monitoring experiment used temperature sensors in monitoring wells to measure temperature changes at different depths within the newly filled municipal solid waste layer after landfilling.

[0220] During the first 400 days of the observation experiment, no leachate was drained or injected, nor was any air pumping or injection performed on the newly filled municipal solid waste layer in the experimental area, and the influence of gas-liquid transport on heat transport was not considered. The model and its solution of this invention were applied to the temperature calculation in the Wuxi landfill to analyze the temperature variation law inside the landfill and compare it with the measured temperature data of the Wuxi landfill.

[0221] The thermal conductivity of the unsaturated waste layer at the Wuxi landfill is 8640 J / day / m / K, and the specific heat capacity is 557.4 J / kg / K; the thermal conductivity of the saturated waste layer is 17280 J / day / m / K, and the specific heat capacity is 1376.2 J / kg / K; the waste density is 700 kg / m³. 3 Considering the landfilling process of newly filled municipal solid waste layers, the heat generation function of the newly filled municipal solid waste layers in the Wuxi landfill is taken as:

[0222] (4.3.58)

[0223] The old waste layers in the Wuxi landfill are all over 10 years old and have basically completed degradation; therefore, it is assumed that there is no degradation heat generation in the old waste layers. The boundary conditions at the top of the landfill are obtained by using the measured surface temperature with a cosine function, and the fitted function is as follows:

[0224] (4.3.59)

[0225] The bottom of the old waste layer is kept at a constant temperature of 20℃. The initial temperature of both the old and new waste layers is set at 20℃.

[0226] The calculation results of the model of this invention can reflect the temperature changes in the newly filled municipal solid waste layer in Wuxi landfill well, which proves the rationality and practicality of the one-dimensional heat conduction model considering vertical stratification in landfill and the solution.

[0227] For example, long-term temperature monitoring was conducted at the Michigan landfill in Michigan, USA. Area B of the Michigan landfill primarily contains municipal solid waste, with a total depth of 33 meters. Temperature sensors were used to measure temperature changes at different depths within the waste volume five years after landfilling. The sensors were vertically buried at a depth of 13 meters. Since the measured temperatures in Area B were obtained five years after landfilling, it was assumed that the landfill was in a stable state at this time, and the influence of gas-liquid transport on heat transport was not considered. The model and its solution were applied to the calculation of heat transport in the waste layer and foundation soil layer of the Michigan landfill, analyzing the temperature variation patterns within the landfill and comparing them with the measured temperature data of the Michigan landfill.

[0228] The thermal conductivity of the waste layer at the Michigan landfill is 86,400 J / day / m / K, and the waste density is 1,000 kg / m³. 3 The specific heat capacity of waste is 2000 kJ / m³. 3 / K; the foundation soil layer is 105 m thick, the thermal conductivity of the foundation soil layer is 216000 J / day / m / K, and the garbage density is 2091.8 kg / m³. 3 The specific heat capacity of waste is 2800 kJ / m³. 3 / K; The heat production function for the Michigan landfill is taken as:

[0229] (4.3.60)

[0230] There is no degradation heat generation in the foundation soil layer. The boundary conditions at the top of the landfill are consistent with the air temperature, and the change in air temperature over time is represented by a sine function, taking into account seasonal variations.

[0231] (4.3.61)

[0232] In equation (4.3.61): Tm and A T The temperatures were set at 12.3℃ and 16.6℃ respectively, with t0 set to 110 days. A constant temperature of 12.3℃ was maintained at the bottom of the foundation soil layer. Continuity conditions were applied between the waste layer and the foundation soil layer. The initial temperature of both the waste layer and the foundation soil layer was set to 12.3℃.

[0233] The calculation results of the model in this invention can reflect the temperature changes in the Michigan landfill well, which proves the rationality and practicality of the one-dimensional heat conduction model that considers vertical stratification in the landfill and the solution.

[0234] Through multi-stage data fusion and model collaboration, the accuracy and reliability of landfill temperature prediction were improved. First, by dynamically matching the preset prediction cycle with the initial monitoring interval (e.g., daily or weekly sampling), a time-space dual-dimensional data acquisition framework was constructed, covering the typical fluctuation cycle of waste degradation heat generation. Second, when using a vertically layered one-dimensional heat conduction model for temperature prediction, the model analyzes the interaction effects of each layer's thermophysical parameters and boundary conditions using orthogonal expansion. The collaborative application of its prediction results and the error mapping model quantifies the deviation contribution rate of the degradation heat generation function and heat exchange coefficient. Finally, this solution provides reliable data support for optimizing landfill monitoring strategies, and its modular design supports seamless integration with multi-field coupled models, providing full-chain technical support for landfill temperature prediction.

[0235] Example 2

[0236] This embodiment discloses a deep learning-based landfill heat conduction prediction error correction system. The system can implement the method of the above embodiment, including a data acquisition module for influencing factors of existing landfill internal temperature prediction error, a data acquisition module for existing landfill internal layer prediction error, a landfill internal layer temperature prediction error mapping model construction module, a current landfill internal layer temperature prediction error mapping module, a current landfill temperature prediction initial determination module, and a current landfill temperature prediction final determination module.

[0237] The existing landfill internal temperature prediction error influencing factor data acquisition module sets an initial landfill monitoring interval to monitor and collect data on internal temperature prediction error influencing factors for multiple landfills.

[0238] The existing landfill inner layer prediction error data acquisition module, in conjunction with the pre-constructed vertical one-dimensional heat conduction model of the landfill, predicts the temperature data of each layer in multiple landfills in S1, and calculates the error data of each layer and the corresponding layer number in the prediction results and the actual measurement data.

[0239] The landfill inner layer temperature prediction error mapping model construction module constructs the final landfill inner layer temperature prediction error mapping model based on the influencing factor data collected by S1, the error data collected by S2, and the corresponding layer number.

[0240] The current landfill inner layer temperature prediction error mapping module uses the landfill vertical one-dimensional heat conduction model in S2 to predict the temperature data of each layer of the current landfill to be predicted, and inputs the corresponding temperature prediction error influencing factor data into the mapping model in S3 for mapping.

[0241] The current landfill temperature prediction initial determination module uses the mapping result in S4 to perform error compensation on the prediction result in S4; then it determines whether the temperature prediction accuracy requirement is met based on the compensation result; if it is met, the prediction result is output; if it is not met, S6 is executed.

[0242] The current landfill temperature prediction final determination module adjusts the initial landfill monitoring interval in S1 multiple times and repeats S1, S2, S3, S4 and S5 to obtain multiple compensated prediction data and corresponding landfill monitoring intervals, and constructs a final adjusted and compensated inner layer temperature prediction mapping model; then, it adjusts the landfill monitoring interval multiple times and inputs the data of the influencing factors of the current landfill's inner temperature prediction error into the final adjusted and compensated inner layer temperature prediction mapping model for mapping; and outputs the prediction result.

[0243] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0244] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for correcting prediction errors of landfill heat conduction based on deep learning, characterized in that, Includes the following steps: S1. Set the initial landfill monitoring interval to monitor and collect data on factors affecting the internal temperature prediction error of multiple landfills. Specifically, it includes: S11. Define several types of factors that affect the error in predicting the internal temperature of a landfill and construct a one-dimensional heat conduction model in the landfill that considers vertical stratification. This will yield a set of factors that affect the internal temperature prediction error and a one-dimensional vertical heat conduction model of the landfill. S12. Select several existing landfills and their corresponding monitoring cycles to obtain an existing landfill set and an existing monitoring cycle set; then set an existing initial landfill monitoring interval; based on the existing initial landfill monitoring interval, the existing monitoring cycle set, and the set of internal temperature prediction error influencing factors, monitor and collect the data of internal temperature prediction error influencing factors corresponding to each landfill in the existing landfill set to obtain an existing internal temperature prediction error influencing factor dataset; Constructing a one-dimensional heat conduction model for a landfill that considers vertical stratification includes the following steps: S111. Construct the one-dimensional heat transport equation for the i-th layer of the landfill; as follows. ； In the formula: Let t be the temperature in the i-th layer, and t be the time. Let be the thermal conductivity coefficient in the z-direction of the i-th layer; Let be the rate of heat change caused by the degradation of municipal solid waste per unit volume in the i-th layer, i.e., the degradation heat generation rate. When the i-th layer is an intermediate cover layer... =0; Let be the specific heat capacity of the i-th layer; Let n be the density of the i-th layer, n be the maximum number of layers, and K represent the K-th layer. The parameter represents the thermal response sensitivity or heating efficiency. Represents the vertical coordinate or depth from a reference plane; , This indicates the depth of a certain reference plane downwards. Located inside the i-th layer; S112. Consider the two temperature conditions at the top of the landfill and set the top boundary conditions, the bottom boundary conditions, and the conditions for liquid phase, gas phase, and heat continuity between adjacent layers. S113. Based on the two boundary conditions and continuity conditions set in S112, the analytical solution of the one-dimensional heat transport equation in the i-th layer of the landfill in S111 is solved by superposition method, separation of variables and orthogonal expansion method. S2. Using a pre-constructed vertical one-dimensional heat conduction model of the landfill, the temperature data of each layer in multiple landfills in S1 are predicted, and the error data of each layer in the prediction results and the actual measurement data, as well as the corresponding layer number, are calculated. S3. Based on the influencing factor data collected in S1 and the error data collected in S2, and the corresponding layer number, construct the final landfill inner layer temperature prediction error mapping model. S4. Predict the temperature data of each layer of the landfill to be predicted, and input the corresponding temperature prediction error influencing factors into the mapping model in S3 for mapping. S5. Use the mapping results in S4 to perform error compensation on the prediction results in S4; collect short-term verification data of the current landfill, and then determine whether the temperature prediction accuracy meets the requirements based on the compensation results; if it does, output the prediction results; otherwise, proceed to S6. S6. Adjust the initial landfill monitoring interval described in S1 multiple times and repeat S1, S2, S3, S4 and S5 to construct the final adjusted and compensated inner layer temperature prediction mapping model; adjust the landfill monitoring interval multiple times and input it into the final adjusted and compensated inner layer temperature prediction mapping model for mapping; output the final temperature prediction result based on the mapping result.

2. The method for correcting landfill heat conduction prediction errors based on deep learning according to claim 1, characterized in that, S2 includes the following steps: S21. Based on the initial landfill monitoring interval and the existing monitoring cycle set, collect temperature data of each layer in each landfill in the existing landfill set to obtain the existing landfill inner layer temperature dataset; then use the landfill vertical one-dimensional heat conduction model to predict the temperature data of each layer in each landfill in the existing landfill set to obtain the existing landfill inner layer temperature prediction dataset. S22. Based on the existing landfill inner layer temperature dataset and the existing landfill inner layer temperature prediction dataset, calculate the average prediction error of the temperature of each layer in each landfill in the existing landfill set and the corresponding number of layers, to obtain the existing landfill inner layer temperature prediction error dataset and the landfill prediction error collection layer set.

3. The method for correcting landfill heat conduction prediction errors based on deep learning according to claim 2, characterized in that, S3 includes the following steps: S31. Based on the existing landfill inner layer temperature prediction error dataset, the landfill prediction error collection layer set, and the existing inner temperature prediction error influencing factor dataset, construct a mapping model between the landfill prediction error collection layer set, the inner temperature prediction error influencing factor data, and the landfill inner layer temperature prediction error data to obtain the final landfill inner layer temperature prediction error mapping model.

4. The method for correcting landfill heat conduction prediction errors based on deep learning according to claim 3, characterized in that: The error mapping model for predicting the inner layer temperature of the final landfill described in S31 adopts the GBDT gradient boosting decision tree model.

5. The method for correcting landfill heat conduction prediction errors based on deep learning according to claim 4, characterized in that, S4 includes the following steps: S41. Set the current landfill to be predicted and preset the prediction period to obtain the current initial preset prediction period; S42. Based on the current initial preset prediction cycle, initial landfill monitoring interval and internal temperature prediction error influencing factor type set, collect the internal temperature prediction error influencing factor data corresponding to the current landfill to be predicted, and obtain the current internal temperature prediction error influencing factor dataset. Then, the vertical one-dimensional heat conduction model of the landfill is used to predict the temperature data of each layer of the current landfill to be predicted, so as to obtain the temperature prediction dataset of the inner layer of the current landfill. S43. The current internal temperature prediction error influencing factor dataset is combined with the number of each layer of the current landfill to be predicted and input into the final landfill internal temperature prediction error mapping model for mapping to obtain the current landfill internal temperature prediction error dataset.

6. The method for correcting landfill heat conduction prediction errors based on deep learning according to claim 5, characterized in that, S5 includes the following steps: S51. The current landfill inner layer temperature prediction dataset is compensated using the current landfill inner layer temperature prediction error dataset to obtain the current compensated inner layer temperature prediction dataset. S52. Collect short-term verification data of the current landfill, evaluate the current compensated inner layer temperature prediction dataset according to the temperature prediction accuracy requirements, and obtain the initial evaluation result; if the initial evaluation result shows that the temperature prediction accuracy requirements are met, output the temperature prediction result; otherwise, execute S6.

7. The method for correcting landfill heat conduction prediction errors based on deep learning according to claim 6, characterized in that, S6 includes the following steps: S61. Set the cumulative number of data adjustment repetitions; adjust the initial landfill monitoring interval in S12 repeatedly according to the cumulative number of data adjustment repetitions. After each adjustment, repeat S12, S21, S22, S31, S41, S42, S43 and S51 and record the adjusted landfill monitoring interval to obtain the current adjusted and compensated inner layer temperature prediction dataset and the current landfill monitoring interval adjustment dataset. S62. Based on the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset, construct a mapping model between the current landfill monitoring interval adjustment dataset and the current adjusted and compensated inner layer temperature prediction dataset to obtain the final adjusted and compensated inner layer temperature prediction mapping model. S63. Adjust the current landfill monitoring interval adjustment data multiple times. After each adjustment, input the adjusted current landfill monitoring interval adjustment data into the final adjusted and compensated inner layer temperature prediction mapping model for mapping to obtain the current model verification and compensated inner layer temperature prediction dataset. Continue until the current model verification and compensated inner layer temperature prediction dataset shows that it meets the temperature prediction accuracy requirements, and output the current model verification and compensated inner layer temperature prediction dataset.

8. A system for implementing the deep learning-based landfill heat conduction prediction error correction method as described in any one of claims 1-7, characterized in that: It includes a data acquisition module for factors affecting the temperature prediction error in existing landfills, a data acquisition module for the prediction error in the inner layer of existing landfills, a model construction module for the prediction error mapping of the inner layer of landfills, a mapping module for the prediction error of the inner layer of current landfills, an initial determination module for the temperature prediction of current landfills, and a final determination module for the temperature prediction of current landfills.

Images