A method and system for predicting ecological restoration in mining areas based on carbon emission time series data

Abstract

This application provides a method and system for predicting ecological restoration in mining areas based on carbon emission time-series data, belonging to the field of carbon emission prediction technology. The method includes: acquiring historical carbon emission time-series data; using a dynamic matching model to employ differentiated prediction time granularity at different stages of ecological restoration, ensuring the prediction frequency aligns with the restoration demand cycle; using a simulated annealing algorithm to adaptively optimize the prediction step size, where the simulated annealing algorithm sets an initial temperature corresponding to the high-frequency prediction demand in the early stages of vegetation restoration, and gradually lowers the temperature to extend the prediction step size as the restoration process progresses; and designing a step-size evolution algorithm to allow the prediction time granularity to autonomously evolve with the restoration process, initially setting a high-frequency prediction mode with gene encoding corresponding to a weekly step size. This invention can improve the prediction accuracy of carbon emissions and resource utilization efficiency, providing scientific decision support for ecological restoration in mining areas.

Description

Technical Field

[0001] This invention relates to the field of carbon emission prediction technology, and in particular to a method and system for predicting ecological restoration in mining areas based on carbon emission time-series data. Background Technology

[0002] In the field of mine ecological restoration, research on how to effectively predict and manage carbon emissions is of paramount importance. This area is directly related to environmental protection and sustainable resource utilization, and is a crucial link in achieving green development goals. During the restoration process, the dynamic changes in carbon emissions not only affect the restoration effect of the ecosystem but also have a profound impact on regional environmental quality and climate change. Therefore, accurate prediction methods are urgently needed to guide restoration work.

[0003] However, existing methods often struggle to adapt to the actual needs of different stages in carbon emission prediction during mining area ecological restoration. Many methods lack flexibility in setting prediction frequencies, failing to adjust for changes in environmental conditions throughout the restoration process. This results in overly frequent predictions at some stages, wasting computational resources, while prediction intervals are too long at other stages, missing critical changes and affecting the timeliness and accuracy of decision-making. This mismatch significantly reduces the scientific rigor and efficiency of restoration efforts.

[0004] A deeper technical challenge lies in the unresolved issue of coordinating the changing characteristics of carbon emissions during the ecological restoration of mining areas with the predicted time intervals. The changing characteristics of carbon emissions exhibit significant differences at different restoration stages. For example, in the early stages of soil improvement, carbon flux fluctuations are very drastic, requiring shorter time intervals to capture these rapid changes; while in the stage where plant communities gradually stabilize, the changing trend tends to be gentler, and excessively short prediction intervals become redundant. It is precisely because the prediction time intervals have not been dynamically adjusted according to these stage-specific characteristics that the prediction results are either inaccurate or inefficient in resource utilization.

[0005] Specifically, in practical operations, if the prediction interval is too long during the soil remediation stage, it may fail to detect the surge in carbon emissions caused by soil disturbance or organic matter decomposition in time, thus missing the optimal intervention opportunity. Conversely, if the prediction interval is too short during the community succession stage, frequent calculations will increase unnecessary costs and affect the overall efficiency of the remediation plan. This contradiction between the time interval and the characteristics of the remediation stage runs through the entire process of ecological restoration in mining areas.

[0006] Therefore, how to dynamically adjust the prediction time interval based on the carbon emission change characteristics at different stages of ecological restoration in mining areas, in order to balance accuracy and resource efficiency, has become a key issue that urgently needs to be addressed. Summary of the Invention

[0007] This invention provides a method for predicting ecological restoration in mining areas based on carbon emission time-series data, aiming to improve the accuracy of carbon emission prediction and resource utilization efficiency, and provide scientific decision support for ecological restoration in mining areas.

[0008] In a first aspect, the present invention provides a method for predicting ecological restoration in mining areas based on carbon emission time-series data, mainly comprising: Historical carbon emission time-series data are obtained, which reflects the characteristics of carbon flux fluctuations during ecological restoration. The fluctuation characteristics of the historical carbon emission time series data were analyzed, and a dynamic matching model between the prediction step size and the ecological restoration stage was established. The ecological restoration stage includes the soil improvement period, the pioneer plant establishment period, and the community succession period. According to the dynamic matching model, differentiated prediction time granularity is adopted at different stages of ecological restoration to coordinate the prediction frequency with the restoration demand cycle. The simulated annealing algorithm is used to adaptively optimize the prediction step size. The simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction needs in the early stage of vegetation restoration. As the restoration process progresses, the temperature is gradually reduced to extend the prediction step size. The step size evolution algorithm is designed to make the prediction time granularity evolve autonomously with the restoration process. In the initial stage, the gene encoding is set to correspond to the high-frequency prediction mode with a weekly step size.

[0009] Furthermore, the simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction requirements in the early stage of vegetation restoration, including: the simulated annealing algorithm evaluates the comprehensive objective function of carbon emission prediction error and computational cost at the current step size in each iteration; the simulated annealing algorithm sets a smaller step size neighborhood search range during the soil improvement period to capture rapidly changing carbon flux, and expands the neighborhood during the stable period of community succession to allow for larger step size jumps.

[0010] Furthermore, the simulated annealing algorithm sets a smaller step size neighborhood search range during the soil improvement period to capture rapidly changing carbon fluxes, including: the simulated annealing algorithm accepts step size changes within the step size neighborhood search range through a probability acceptance mechanism during the annealing process; the simulated annealing algorithm generates the predicted step size according to the probability acceptance mechanism, so that the predicted time granularity follows the dynamic evolution of the carbon cycle characteristics of the mining area ecosystem.

[0011] Furthermore, the design step size evolution algorithm enables the prediction time granularity to evolve autonomously with the restoration process, including: the step size evolution algorithm evaluates the synchronicity between the carbon emission prediction accuracy and the changes in ecological indicators under different step sizes through a fitness function.

[0012] Furthermore, the step-size evolutionary algorithm evaluates the synchronicity between the carbon emission prediction accuracy and the changes in ecological indicators under different step sizes through a fitness function, including: the step-size evolutionary algorithm adopts a small step-size mutation strategy to capture the rapid carbon sink formation process in the early stage of vegetation restoration; the step-size evolutionary algorithm gradually evolves into predictive genotypes with monthly or quarterly step sizes as the soil microbial community stabilizes; so that the step-size evolutionary trajectory is consistent with the carbon balance succession stage of the mining area ecosystem.

[0013] Furthermore, the analysis of the fluctuation characteristics of the historical carbon emission time series data includes: determining the carbon flux change sensitivity of the ecological restoration stage based on the fluctuation characteristics; and the dynamic matching model associating the prediction step size with the ecological restoration stage based on the carbon flux change sensitivity.

[0014] Furthermore, based on the dynamic matching model, different prediction time granularities are adopted at different stages of ecological restoration, including: dense prediction with a weekly or monthly step size in the early stage of vegetation restoration; and prediction with a quarterly or annual step size in the period of ecosystem stability.

[0015] Furthermore, the simulated annealing algorithm accepts step size changes within the step size neighborhood search range through a probability acceptance mechanism during the annealing process, including: the probability acceptance mechanism calculates the probability value of accepting the step size change based on the current temperature; if the probability value meets a preset condition, the step size change is accepted to generate the predicted step size for the next iteration.

[0016] Secondly, the present invention provides a mining area ecological restoration prediction system based on carbon emission time-series data, comprising: The acquisition module is used to acquire historical carbon emission time-series data, which is used to reflect the fluctuation characteristics of carbon flux during ecological restoration. A module is established to analyze the fluctuation characteristics of the historical carbon emission time series data and to establish a dynamic matching model between the prediction step size and the ecological restoration stage, which includes the soil improvement period, the pioneer plant establishment period and the community succession period. The prediction module is used to apply differentiated prediction time granularity to different stages of ecological restoration based on the dynamic matching model, so that the prediction frequency is coordinated with the restoration demand cycle. The optimization module is used to adaptively optimize the prediction step size using a simulated annealing algorithm. The simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction requirements in the early stage of vegetation restoration, and gradually lowers the temperature to extend the prediction step size as the restoration process progresses. The design module is used to design a step-size evolutionary algorithm so that the prediction time granularity evolves autonomously with the repair process. In the initial stage, a high-frequency prediction mode with a weekly step size corresponding to the gene encoding is set.

[0017] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention constructs a dynamic matching model between the prediction step size and the ecological restoration stage, combining simulated annealing and step-size evolution algorithms. During the soil improvement period, it uses high-frequency, small-step predictions to capture rapidly changing carbon fluxes. During community succession, it gradually extends the step size to adapt to a stable trend. Simultaneously, it utilizes a probabilistic acceptance mechanism and fitness function to avoid local optima, ensuring that the prediction granularity dynamically evolves with the carbon cycle characteristics. Ultimately, this invention achieves a high degree of consistency between the prediction time granularity and the carbon balance succession stage of the mining area ecosystem, improving the accuracy of carbon emission prediction and resource utilization efficiency, and providing scientific decision support for ecological restoration in mining areas. Attached Figure Description

[0018] Figure 1 A flowchart of a mining area ecological restoration prediction method based on carbon emission time series data provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the functional modules of a mining area ecological restoration prediction method based on carbon emission time series data, provided in an embodiment of the present invention.

[0020] Figure 3 This is a functional module block diagram of a mining area ecological restoration prediction system based on carbon emission time series data, provided in an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0022] like Figures 1-2 As shown in the figure, the method for predicting ecological restoration in mining areas based on carbon emission time-series data provided by this invention may specifically include the following steps: S100: Obtain historical carbon emission time series data for the mining area.

[0023] The historical carbon emission time series data reflects the fluctuation characteristics of carbon flux during ecological restoration.

[0024] Multi-year continuous carbon emission time-series data were obtained from a mining area ecological restoration monitoring database. This data includes daily or weekly carbon flux measurements, reflecting the carbon emission fluctuation characteristics at different stages of ecological restoration. The acquired carbon emission time-series data underwent preprocessing, including outlier removal and missing value imputation, to form a continuous and complete carbon emission sequence, preserving the true fluctuation characteristics of carbon flux during ecological restoration. The preprocessed carbon emission sequence was then decomposed into trend components, seasonal components, and random fluctuation components to extract carbon flux fluctuation characteristics reflecting changes at different stages of ecological restoration. The fluctuation amplitude and period length of the separated trend and seasonal components were calculated to determine the fluctuation characteristics of the carbon emission time-series data at different restoration stages, which is used for dynamic matching of subsequent prediction step sizes.

[0025] For example, long-term continuous carbon emission time-series data can be obtained from a mining area ecological restoration monitoring database. Such databases typically integrate satellite remote sensing, ground sensors, and historical records to form a comprehensive carbon flux archive, which is beneficial for capturing the fluctuations in the entire process of ecological restoration from soil improvement to vegetation stabilization. Doing so can ensure the reliability of the data source and avoid subsequent analysis bias.

[0026] In one possible implementation, assuming the mine remediation project covers a 10-year cycle, the data includes daily carbon flux measurements, such as the gradual change of carbon dioxide emissions from high to low. Once acquired, the data is directly imported into an analysis platform, which is beneficial for revealing the characteristics of the transformation from high emissions in the early stage to carbon sinks in the later stage.

[0027] Specifically, this acquisition method extracts data for a specified time period through a query interface, ensuring the integrity of the time series and thus providing a solid foundation for extracting fluctuation characteristics, which is beneficial to improving prediction accuracy. For example, the acquired carbon emission time series data is preprocessed, including removing outliers (such as extreme readings caused by sudden equipment failures), eliminating these values ​​using a median filtering method, filling in missing values, and using linear interpolation to connect consecutive data points to form a continuous and complete carbon emission sequence. This preserves the true fluctuation characteristics of carbon flux during ecological restoration and avoids noise interference with the analysis results.

[0028] In one possible implementation, for missing weekly data in a copper mining sequence, preprocessing reveals clearer seasonal peaks and troughs, which improves the accuracy of subsequent decomposition. Specifically, preprocessing ensures a smooth and continuous sequence, thereby enhancing data usability and helping to capture subtle changes during the remediation phase, such as the rapid decline in carbon emissions during the pioneer plant period.

[0029] For example, the preprocessed carbon emission sequence is decomposed into time series components. This decomposition uses a classic additive model to separate the sequence into trend components, representing the long-term direction of carbon emission decline, seasonal components to capture annual cycles (such as high emissions in summer), and random fluctuation components to isolate unpredictable noise. This extracts the carbon flux fluctuation characteristics that reflect changes in the ecological restoration stage. This decomposes complex signals into manageable parts, which is beneficial for identifying unique patterns in different stages.

[0030] In one possible implementation, for a sequence, the trend component shows a steep decline from the early stages of repair to a plateau later, while the seasonal component highlights the impact of the rainy season, which is helpful for understanding the dynamics of the carbon cycle.

[0031] Specifically, this decomposition process first estimates the trend using a moving average, then subtracts it to obtain the seasonality, leaving the random component. This ensures comprehensive feature extraction, which is beneficial for the targeted nature of subsequent calculations. For example, for the separated trend and seasonal components, fluctuation amplitudes (e.g., standard deviation to measure the degree of variability) and period lengths (e.g., autocorrelation function to determine the repetition interval) are calculated to determine the fluctuation characteristics of carbon emission time-series data at different remediation stages. This is used for dynamic matching of subsequent prediction step sizes, thus quantifying the features to guide the adjustment of prediction frequency and facilitating alignment with ecological needs.

[0032] In one possible implementation, the magnitude calculation of the trend component shows that high volatility in the initial stage corresponds to weekly forecasts, while the cycle length of the seasonal component indicates that the annual pattern is suitable for quarterly steps, which is beneficial for optimizing resource allocation. Specifically, this calculation uses statistical tools to process the component data, ensuring that features are aligned with the remediation phase, which is beneficial for achieving efficient carbon emission forecasting.

[0033] S200: Analyze the fluctuation characteristics of the historical carbon emission time series data and establish a dynamic matching model between the prediction step size and the ecological restoration stage.

[0034] The ecological restoration stages include the soil improvement period, the pioneer plant establishment period, and the community succession period. Fluctuation characteristics are extracted from historical carbon emission time-series data. For different stages of ecological restoration in the copper mining area, carbon emission variation patterns are obtained for the soil improvement period, the pioneer plant establishment period, and the community succession period. The fluctuation amplitude and periodic characteristics of each stage are determined, forming a stage-specific fluctuation dataset. For this stage-specific fluctuation dataset, a matching rule between the prediction step size and the ecological restoration stage is constructed. A shorter prediction time granularity is set during the soil improvement period to capture rapidly changing carbon fluxes. The prediction time granularity is appropriately extended during the pioneer plant establishment period to adapt to the changing rhythm of the initial vegetation recovery. During the community succession period, the prediction time granularity is further adjusted to a longer time granularity to reflect the stable trend of the ecosystem, generating a stage-matched step size scheme. Based on the phase-matching step size scheme, the prediction time granularity for each ecological restoration stage is dynamically adjusted. Carbon emission data is collected intensively with a short time granularity during the soil improvement stage, gradually transitioning to a medium time granularity during the pioneer plant planting stage, and employing a sparser collection method with a longer time granularity during the community succession stage, thus forming a dynamically adjusted prediction step size configuration. Using this dynamically adjusted prediction step size configuration, combined with the fluctuation characteristics of historical carbon emission time-series data, differentiated prediction frequencies are implemented at different ecological restoration stages to ensure that the prediction step size is coordinated with the actual needs of the soil improvement stage, pioneer plant planting stage, and community succession stage, completing the construction of a dynamic matching model between the prediction step size and the ecological restoration stage.

[0035] For example, the process of extracting fluctuation features from historical carbon emission time series data involves time-domain analysis of the data sequence to identify the fluctuation patterns of carbon emissions. These fluctuation features include amplitude (i.e., the difference between peak and trough emissions) and periodicity (i.e., the repetition interval of changes). This extraction can provide basic data support for subsequent matching models, making predictions more consistent with actual ecological changes, thereby improving the accuracy of restoration decisions.

[0036] In one possible approach, carbon emission variation patterns are obtained for different stages of ecological restoration in copper mining areas. For example, during the soil improvement period, carbon emission variation patterns exhibit high-frequency fluctuations because the addition of organic matter to the soil leads to increased microbial activity in the short term, causing emissions to rise sharply and then fall back, forming large peaks. During the pioneer plant establishment period, the variation pattern shifts to a medium frequency because plant roots begin to fix the soil, carbon flux gradually tends to balance, the fluctuation amplitude decreases, and the cycle extends to several weeks. During the community succession period, a low-frequency stable pattern is observed, with slow emission changes and a cycle that can reach several months. This multi-faceted pattern acquisition can support each other, ensuring that the dataset comprehensively reflects the stage differences, which is beneficial to avoiding prediction bias and improving the reliability of ecological monitoring.

[0037] Specifically, when determining the fluctuation amplitude and periodic characteristics of each stage, the amplitude can be quantified by calculating the standard deviation of the sequence. For example, a large standard deviation indicates drastic changes during the soil improvement period, while the periodic characteristics are analyzed by using the autocorrelation function to analyze the repeating patterns. The role of these characteristics in forming a stage fluctuation dataset is to provide a quantitative basis for dynamic matching, making the subsequent rule construction more targeted, thereby optimizing the efficiency of resource allocation in the restoration process.

[0038] In one possible implementation, a matching rule is constructed for the prediction step size and ecological restoration stage for the periodic fluctuation dataset. Specifically, during the soil improvement period, a shorter prediction time granularity, such as daily or weekly, is set to capture rapidly changing carbon fluxes. This helps to adjust restoration measures in a timely manner, such as adding soil stabilizers, to prevent runaway emissions. During the pioneer plant planting period, the granularity is appropriately extended to monthly to adapt to the changing rhythm of vegetation restoration in the early stage, which is beneficial for monitoring the impact of plant growth on carbon sinks. During the community succession period, the granularity is adjusted to quarterly or annual to reflect the stable trend of the ecosystem. The construction of such rules supports the adaptability of the model from multiple directions, ensures that the prediction frequency is coordinated with the stage requirements, and improves the sustainability of the overall restoration effect.

[0039] For example, the predicted time granularity for each stage of ecological restoration can be dynamically adjusted based on a stage-matched step size scheme. For instance, during the soil improvement period, carbon emission data can be collected intensively at a short time granularity, capturing instantaneous changes through high-frequency sampling, which is beneficial for early intervention of pollution sources. During the pioneer plant establishment period, the data collection frequency can be gradually transitioned to a medium time granularity, allowing observation of the gradual impact of plant establishment on emissions. During the community succession period, a sparse collection method with a longer time granularity can be used to reduce unnecessary monitoring costs while maintaining accuracy. The dynamic configuration formed by these adjustments can support each other from both the cost and accuracy perspectives, achieving efficient resource utilization.

[0040] Specifically, by using dynamically adjusted prediction step size configurations combined with the fluctuation characteristics of historical carbon emission time series data, differentiated prediction frequencies are implemented. For example, configurations are applied at different stages to ensure that the prediction step size is coordinated with the actual needs of the soil improvement period, pioneer plant planting period, and community succession period. This approach can improve prediction accuracy and save computational resources, ultimately completing the construction of a dynamic matching model, which is beneficial to the long-term optimization of ecological restoration in copper mining areas.

[0041] S300: Based on the dynamic matching model, differentiated prediction time granularity is adopted at different stages of ecological restoration to coordinate the prediction frequency with the restoration demand cycle.

[0042] By extracting fluctuation characteristics from historical carbon emission time-series data, the ecological restoration stages corresponding to the fluctuation amplitude and frequency are determined. For each ecological restoration stage determined by the fluctuation amplitude and frequency, a correspondence between the prediction step size and the restoration stage is established, forming a differentiated prediction time granularity. When the ecological restoration stage changes, the frequency of the current carbon emission prediction is adjusted according to the established differentiated prediction time granularity to align the prediction frequency with the demand cycle of the restoration stage. Based on the adjusted prediction frequency, the fluctuation characteristics of the latest carbon emission time-series data are continuously acquired to update the ecological restoration stage judgment and re-establish a differentiated prediction time granularity.

[0043] In one possible implementation, the process of obtaining fluctuation characteristics from historical carbon emission time-series data involves performing time-domain analysis on the data sequence, such as quantifying the fluctuation amplitude by calculating the standard deviation of data points and using Fourier transform to identify the frequency of periodic fluctuations. This can accurately map to the ecological restoration stage, because large fluctuation amplitude usually corresponds to highly sensitive changes during the soil improvement period, while high fluctuation frequency indicates rapid carbon flux adjustment during the pioneer plant planting period. The beneficial effect is to improve the accuracy of stage judgment and avoid misjudgments caused by static methods.

[0044] For example, when establishing a correspondence between prediction step size and restoration stage for ecological restoration stages with defined fluctuation amplitude and frequency, the soil improvement period can be mapped to a weekly step size, the pioneer plant planting period to a monthly step size, and the community succession period to a quarterly step size, forming differentiated prediction time granularity. The purpose of doing so is to make the prediction more in line with the actual ecological dynamics and bring about higher resource utilization efficiency. For example, intensive prediction in the initial stage can detect carbon emission anomalies early and support timely intervention.

[0045] In one possible implementation, the process of adjusting the current carbon emission forecast frequency based on the established differentiated forecast time granularity during changes in the ecological restoration phase includes monitoring transition signals, such as switching from weekly to monthly forecasts when carbon emission fluctuations tend to stabilize. This aligns the forecast frequency with the restoration demand cycle, which has the benefit of reducing unnecessary computational load while maintaining forecast sensitivity. For example, when transitioning from the initial to the stable period, adjusting the frequency can optimize data processing and prevent resource waste caused by over-prediction.

[0046] For example, the process of continuously acquiring the latest carbon emission time-series data fluctuation characteristics for the adjusted prediction frequency can update the standard deviation and frequency calculation through real-time data streams, thereby updating the ecological restoration stage judgment and re-forming a differentiated prediction time granularity. The rationale for doing so is to achieve dynamic cyclic optimization, bringing continuous adaptive capabilities. For example, in the restoration of vegetation in mining areas, if new data shows increased fluctuations, it can roll back to a finer granular step size, supporting accurate tracking of carbon balance succession and ensuring the coordination and reliability of the overall restoration prediction.

[0047] In one possible implementation, these processes support each other. For example, the initial fluctuation characteristics are directly input into the correspondence establishment, and subsequent adjustments and updates form a closed loop, working together to coordinate the prediction frequency with the maintenance demand cycle. The beneficial effect is to enhance the robustness of the system and improve the overall effectiveness of ecological maintenance prediction from multiple aspects such as data accuracy and computational efficiency. For example, in actual copper mining applications, high-frequency prediction in the early stage captures rapid changes, mid-term adjustments reduce costs, and late-stage updates maintain stability, thus mutually reinforcing the practical value of prediction.

[0048] For example, from another perspective, the acquisition of fluctuation characteristics can also be combined with satellite remote sensing data to assist in the verification stage, such as amplitude analysis and cross-checking with vegetation indices, to ensure accurate judgment. This can bring more comprehensive ecological insights and support the formation of differentiated granularity. When adjusting the frequency, machine learning such as support vector machine classification stage changes can be incorporated to further optimize coordination. These aspects support each other and jointly improve the effectiveness of the method in the long-term restoration process.

[0049] S400: The simulated annealing algorithm is used to adaptively optimize the prediction step size.

[0050] In this embodiment, the simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction requirements in the early stage of vegetation restoration, and gradually lowers the temperature as the restoration process progresses to extend the prediction step size.

[0051] In the initial stage of ecological restoration in the copper mining area, a relatively high initial temperature value was set to simulate the rapidly changing carbon emission fluctuations, addressing the high-frequency prediction requirements during the vegetation restoration phase. A pre-established mapping relationship between temperature and prediction step size was used to map the initial temperature to a shorter prediction time granularity, ensuring the capture of subtle early carbon flux changes and generating a preliminary step size configuration scheme. Based on this preliminary step size configuration scheme, the temperature parameter was gradually reduced as the restoration progressed. Pre-defined cooling rules were used to adjust the prediction time granularity, transitioning the step size from the initial high-frequency mode to a longer period in the middle stage. Simultaneously, after each cooling, a comprehensive index of prediction error and computational cost at the current step size was calculated, forming the adjusted step size configuration data. Based on the adjusted step size configuration data, in the soil improvement stage, the neighborhood range of step size changes was limited. A small-range search was used to ensure the prediction time granularity adapted to the rapidly changing carbon flux characteristics, generating a refined step size scheme suitable for this stage. For the fine step size scheme, during the stable stage of community succession, the neighborhood range of step size changes is expanded, allowing the prediction time granularity to jump to a longer period. The step size is dynamically adjusted through a probability acceptance mechanism to form the final adaptive step size configuration, so as to match the dynamic needs of ecological restoration prediction in copper mining areas.

[0052] For example, when setting a higher initial temperature value in the early stage of ecological restoration in a copper mining area, this can be achieved through a pre-established mapping relationship between temperature and prediction step size. This mapping relationship uses the temperature parameter as an input variable and corresponds to a shorter prediction time granularity. For example, in the vegetation restoration stage, a higher temperature value corresponds to a weekly step size, thereby simulating the rapidly changing characteristics of carbon emission fluctuations. This approach can capture subtle characteristics of early carbon flux changes, ensuring that the generated initial step size configuration scheme is more in line with the needs of high-frequency prediction and avoiding the omission of key fluctuation data.

[0053] In one possible implementation, the temperature parameters are gradually reduced for the initial step size configuration scheme. The prediction time granularity is adjusted by a preset cooling rule. For example, during the transition from weekly to monthly, the prediction error and computational cost are calculated after each cooling, forming the adjusted step size configuration data. This approach can bring about the beneficial effect of a smooth transition of the step size from a high-frequency mode to a medium-term cycle, making the prediction process more adaptable to the dynamic changes in the repair process and enhancing the overall accuracy and efficiency of the prediction.

[0054] For example, when the adjusted step size configuration data restricts the neighborhood range of step size changes during the soil improvement stage, a small-scale search is used to ensure that the prediction time granularity adapts to the rapidly changing carbon flux characteristics. For example, the neighborhood range is set as the finite variation range of the current step size, and a local optimization search is performed to generate a fine step size scheme suitable for this stage. This can bring the beneficial effect of finely capturing the rapid changes in carbon flux, improve the prediction response capability in sensitive stages, and provide continuous basic data support for the expansion of subsequent stable stages.

[0055] In one possible implementation, when the fine-step-size scheme expands the neighborhood range of step-size changes during the stable stage of community succession, it allows the prediction time granularity to jump to the quarterly level and dynamically adjusts it through a probability acceptance mechanism. For example, the mechanism calculates the probability of accepting a poor step-size based on the current temperature, avoiding local optimum traps and forming a final adaptive step-size configuration. This approach can bring the beneficial effect of matching the step-size with the dynamic needs of ecological restoration, improving the flexibility of long-term prediction and resource utilization efficiency, while connecting with the fine-step scheme in the early stage to ensure the logical continuity of the entire process.

[0056] For example, the combination of initial temperature setting and cooling rules can be examined from multiple perspectives. For instance, high temperatures in the early stages of vegetation restoration correspond to short step sizes, which can support small-scale searches during the soil improvement stage. This is because the preliminary scheme generated in the early stages provides a data foundation, while subsequent cooling lays the foundation for a gradual adjustment path to expand the neighborhood. In this way, multiple directions support each other and form a consistent step size adaptive evolution chain, which is beneficial to the stability of overall carbon emission prediction.

[0057] In one possible implementation, regarding the implementation of the probability acceptance mechanism and the expansion of the neighborhood range, for example, when a large jump is allowed in the stable phase, the mechanism evaluates the comprehensive index and decides to accept, supporting the use of data from the intermediate transition. This can bring the beneficial effect of avoiding the fixed step size mode and connect with the mapping relationship of the initial high-frequency demand, ensuring the integrity and effectiveness of the adaptive optimization process.

[0058] For example, the process of setting a cooling rule is explained. For instance, the rule is defined as geometric cooling, which means that the temperature is adjusted by multiplying by a decay factor in each iteration and then calculating the index. This detailed process can bring the beneficial effect of smooth step transition and support the multi-faceted application of fine-grained scheme generation, such as enhanced adaptability in different repair stages.

[0059] In one possible implementation, the process of calculating the comprehensive index can be described by using prediction error as a deviation measure and computation cost as a resource consumption measure, and then weighting and summing the two to form the index. This approach can achieve a balance between accuracy and efficiency, and when combined with a probabilistic mechanism, it can support the dynamic adjustment of the final configuration, enabling full-chain optimization from the initial stage to stability.

[0060] The simulated annealing algorithm evaluates a combined objective function of carbon emission prediction error and computational cost at the current step size in each iteration.

[0061] In each iteration, the error value between the predicted carbon emission sequence and the actual sequence at the current step size is obtained. The weighted sum of the error value and the preset computational cost coefficient is calculated to form the comprehensive objective function value. The difference between the comprehensive objective function value and the comprehensive objective function value of the previous iteration is judged. If the difference is less than zero, the current step size is directly accepted as the initial step size for the next iteration. If the difference is greater than zero, the acceptance probability is calculated, and a random number is generated and compared with the acceptance probability. If the random number is less than the acceptance probability, the current step size is accepted as the initial step size for the next iteration. A new step size is generated in the neighborhood from the accepted current step size and returned to the prediction sequence calculation, forming a continuously iterative comprehensive objective function value evaluation process.

[0062] In one possible implementation, the simulated annealing algorithm optimizes the prediction step size through an iterative process. First, it obtains the error value between the predicted carbon emission sequence and the actual sequence at the current step size. This error value reflects the accuracy of the prediction. For example, in the early stage of ecological restoration in a copper mining area, the predicted sequence may show large fluctuations in carbon emissions every week. If the step size is too long, it will lead to an amplification of the error. The weighted sum of the calculated error value and the preset calculation cost coefficient forms a comprehensive objective function value, which can balance accuracy and efficiency. This helps to avoid wasting computational resources and ensures that the algorithm captures subtle differences in carbon flux during high-frequency change phases, thereby improving the reliability of the overall restoration prediction.

[0063] It should be noted that the formation process of this comprehensive objective function value is similar to a trade-off multi-objective optimization, where the error value represents the prediction deviation and the computational cost coefficient quantifies the resource consumption. For example, assuming that the prediction sequence is generated based on historical time series data, the weighted sum can prevent excessive computational burden caused by excessively shortening the step size. The beneficial effect is that it enables the algorithm to adapt to the needs of different repair cycles and avoids the trap of local optima.

[0064] For example, during the vegetation restoration period, if the error value is high but the cost is low, the function value may still be low, prompting step size adjustment to improve accuracy. A difference judgment is made between the comprehensive objective function value and the value of the previous iteration. If the difference is less than zero, the current step size is directly accepted as the initial step size for the next iteration. This judgment mechanism mimics the cooling acceptance strategy in annealing, enabling rapid convergence to a better solution. For instance, during the soil improvement period, a negative difference indicates an improved step size; accepting it accelerates the algorithm's evolution towards lower errors. The beneficial effect is enhanced adaptability of the predicted step size, ensuring synchronization with ecological succession.

[0065] It should be noted that this direct acceptance avoids unnecessary randomness and maintains iterative stability. If the difference is greater than zero, the acceptance probability is calculated and a random number is generated for comparison. If the random number is less than the acceptance probability, the current step size is accepted as the initial step size for the next iteration. This process introduces a probability mechanism to escape local optima. For example, during the stable period of community succession, the acceptance probability decays based on the temperature parameter, and acceptance may be possible even if the function value is slightly worse. The beneficial effect is to explore a wider step size space, prevent the algorithm from getting stuck in a fixed pattern, and improve the accuracy of capturing long-term trends in carbon emissions.

[0066] For example, the acceptance probability can be calculated using an exponential function, with random numbers uniformly distributed between zero and a certain range. This simulates the random jumps in metallurgical annealing, resulting in a more robust optimization path. A new step size is generated within the neighborhood of the accepted current step size and returned to the calculation of the predicted sequence, forming a continuously iterative process for evaluating the comprehensive objective function value. This loop ensures the algorithm's dynamic evolution; for instance, a new step size is generated through small perturbations, returning to calculate a new predicted sequence. After multiple iterations, the step size tends to be optimal. The beneficial effect is that the prediction time granularity follows the carbon cycle characteristics of the mining area, achieving a comprehensive minimization of error and cost.

[0067] It should be noted that the iteration of the entire process forms a closed loop. Neighborhood generation is similar to searching the nearby solution space. During the stabilization period, larger jumps are allowed, which enhances the algorithm's adaptability to slow changes and thus supports the accurate assessment of carbon emission predictions.

[0068] The simulated annealing algorithm sets a small step size neighborhood search range during the soil improvement period to capture rapidly changing carbon flux, and expands the neighborhood during the stable period of community succession to allow for larger step size jumps.

[0069] During the soil improvement period, a smaller step size neighborhood is set to capture short-term fluctuations due to the rapid changes in carbon flux. Historical carbon emission time-series data is segmented to obtain the rate of change for each segment, and this rate is used as the basis for adjusting the step size neighborhood, ensuring that the prediction granularity adapts to the dynamic evolution characteristics of the soil improvement period. A mapping relationship for step size adjustment is constructed based on the rate of change obtained from the soil improvement period. During the stable community succession period, the step size neighborhood is gradually expanded according to the slowing trend of the rate of change, allowing for larger-scale adjustments to the prediction step size to cover long-term changes after ecosystem stability. Based on the expanded step size neighborhood, and combined with the stable carbon flux characteristics during the stable community succession period, the adjusted step size is validated successively. The corresponding prediction error distribution is extracted from historical data to ensure that the adjusted step size balances prediction accuracy and computational cost, adapting to the prediction needs of the stable period. Based on the validated step size adjustment results, they were applied to the prediction process of ecological restoration in copper mining areas. The carbon flux changes and step size matching were continuously tracked to ensure that the prediction time granularity was consistent with the carbon cycle characteristics of the ecological restoration stage, thus completing the dynamic optimization of the prediction of ecological restoration in copper mining areas.

[0070] In one possible implementation, a smaller step size neighborhood is set during the soil improvement period to account for the rapid changes in carbon flux. Here, the step size neighborhood refers to the range of values ​​that the current step size can be adjusted in the simulated annealing algorithm. For example, by dividing historical carbon emission time series data into multiple time segments, the difference in carbon emissions for each segment is calculated and divided by the time interval to obtain the rate of change. This rate of change is used as the basis to narrow the neighborhood, thereby capturing short-term fluctuations. This approach can lead to more accurate predictive adaptability because adjusting rapidly changing carbon flux over a large range would ignore subtle dynamics and increase prediction bias.

[0071] For example, a mapping relationship for step size adjustment is constructed from the rate of change obtained during the soil improvement period. This mapping relationship is achieved by establishing a correspondence function between the rate of change and the step size range. For example, a high rate of change corresponds to a small range, and a low rate of change corresponds to a large range. During the stable period of community succession, the step size neighborhood is gradually expanded according to the slowing trend of the rate of change, allowing for larger-scale adjustments to cover long-term change patterns. This approach can improve computational efficiency because frequent small-step adjustments are not required during the stable period, avoiding unnecessary resource consumption while maintaining the comprehensiveness of the prediction.

[0072] In one possible implementation, the adjusted step size is validated successively by combining the stable carbon flux characteristics during the stable period of community succession with the expanded neighborhood of the step size. For example, the difference distribution between the predicted and actual values ​​is extracted from historical data to form an error distribution map. Then, the mean error and computation time under different step sizes are compared to ensure a balance between prediction accuracy and computational cost. This approach can bring more reliable adaptability because the validation process confirms the applicability of the step size during the stable period and reduces the accumulation of errors caused by over-optimization.

[0073] For example, the validated step size adjustment results are applied to the prediction process of ecological restoration in copper mining areas. The matching between carbon flux changes and step size is continuously tracked. For example, the matching degree is checked every quarter in actual restoration. If the matching is not good, the step size is finely adjusted to ensure that the prediction time granularity is consistent with the carbon cycle characteristics of the ecological restoration stage. This can bring about dynamic optimization because continuous tracking makes the prediction model follow the ecological evolution, avoiding the prediction lag caused by static step size, and completing the dynamic optimization of the prediction of ecological restoration in copper mining areas.

[0074] The simulated annealing algorithm avoids getting trapped in a fixed step size mode of local optima through the probability acceptance mechanism in the annealing process, so that the prediction time granularity follows the dynamic evolution of the carbon cycle characteristics of the mining area ecosystem.

[0075] Carbon emission temporal fluctuation characteristics were obtained from historical data on ecological restoration in copper mining areas. A set of stage-specific fluctuation characteristics was constructed for carbon flux changes at different restoration stages, such as soil improvement and community succession. These characteristics were divided into high-frequency and low-frequency ranges for dynamic adjustment of subsequent prediction step sizes. An initial prediction step size range was set for each of the high-frequency and low-frequency ranges. A shorter step size was used in the high-frequency range to capture rapidly changing carbon flux data, while the step size range was widened in the low-frequency range to accommodate stable carbon cycle characteristics, forming a preliminary step size matching scheme. Error data of the current prediction step size was extracted from the preliminary step size matching scheme. Combined with a comprehensive evaluation index of carbon emission prediction accuracy and computational cost, the step size was fine-tuned through a probabilistic acceptance mechanism. If the current step size caused the error to exceed a preset threshold, a new step size value was searched within the neighborhood to generate an optimized step size configuration. For the optimized step size configuration, the evolution of carbon cycle characteristics during the ecological restoration of the copper mining area is continuously monitored. The real-time carbon flux data is compared with the optimized step size, and the prediction time granularity is dynamically adjusted to ensure that the prediction step size is always consistent with the carbon cycle characteristics of the mining area's ecosystem.

[0076] For example, when obtaining the time-series fluctuation characteristics of carbon emissions from historical data on ecological restoration in copper mining areas, these characteristics can be extracted by collecting monitoring records from the past few years. For instance, carbon flux may exhibit dramatic daily or weekly fluctuations during the soil improvement period, while tending to level off during the community succession period. This method helps to construct a set of phased fluctuation characteristics, dividing them into high-frequency change intervals, such as the rapid rise phase of the initial restoration, and low-frequency change intervals, such as the stable equilibrium phase in the later stage. This provides a data basis for the dynamic adjustment of the subsequent prediction step size, which is beneficial to improving the adaptability of the prediction because it ensures that the feature set directly reflects the actual changes in the restoration process and avoids the bias caused by static step size.

[0077] In one possible implementation, when setting the initial prediction step size range for the fluctuation characteristics of the high-frequency and low-frequency variation ranges, for example, a one-week step size can be used in the high-frequency range to capture rapid carbon flux data, which can respond promptly to the formation of carbon sinks in the early stage of vegetation restoration. In the low-frequency range, the step size can be widened to one quarter to adapt to stable carbon cycle characteristics, forming a preliminary step size matching scheme. This setting is beneficial to balancing prediction accuracy and resource consumption because it matches the step size range with the ecological stage, reduces unnecessary computational burden, and improves the accuracy of tracking carbon emission trends.

[0078] For example, when extracting the error data of the current prediction step size from the initial step size matching scheme and combining it with the comprehensive evaluation index of carbon emission prediction accuracy and computational cost, the process of fine-tuning the step size through the probability acceptance mechanism involves calculating the error value. If the error exceeds a preset threshold, a new step size is searched in the neighborhood. For example, if the initial step size is 1 month but the error is high, the mechanism accepts a slightly longer step size, such as 1.5 months, with a certain probability to test optimization. This fine-tuning is beneficial to avoid local optima because it introduces randomness to allow for the exploration of a wider range of step size options, thereby generating an optimized step size configuration and improving the robustness of the overall prediction.

[0079] In one possible implementation, when continuously monitoring the evolution of carbon cycle characteristics during ecological restoration in copper mining areas with an optimized step size configuration, the process of comparing real-time carbon flux data with the optimized step size can include comparing daily data input with the step size value. If a deviation is found, the prediction time granularity is dynamically adjusted, for example, switching from a weekly step size to a monthly step size to match the characteristics of the stable period. This helps ensure that the prediction step size is always consistent with the carbon cycle characteristics of the mining area's ecosystem because it achieves real-time evolution response, enhances the long-term reliability of restoration predictions, and supports sustainable ecological management.

[0080] S500: For the evolution of ecological restoration in mining areas from the initial stage to the stable stage, a step-size evolutionary algorithm is designed to make the prediction time granularity evolve autonomously with the restoration process. In the initial stage, a high-frequency prediction mode with a weekly step size corresponding to gene encoding is set.

[0081] For the initial stage of ecological restoration in copper mining areas, gene coding sequences are pre-established, with each gene locus corresponding to a prediction step size option. Weekly step sizes correspond to high-frequency predicted genotypes. Initial fluctuation characteristics are obtained from historical carbon emission time-series data to initialize the gene coding of multiple individuals in the population. Weekly step-size prediction sequences for each individual are obtained from the initialized population. High-frequency prediction is performed on the carbon emission time-series data, and the deviation between the predicted values ​​and actual carbon sink formation data is calculated. This deviation is converted into fitness values, and gene coding of individuals with high fitness is retained within the population. Crossover and mutation operations are performed on the retained individual gene coding. During mutation, new genotypes are preferentially generated within the weekly step-size neighborhood, forming the next generation of population gene coding. High-frequency prediction is then performed on these new individuals, and fitness values ​​are calculated. This process of crossover, mutation, and fitness evaluation is repeated until the prediction step size corresponding to the dominant genotype in the population stably converges to the weekly high-frequency prediction mode, achieving autonomous evolution of the prediction time granularity in the initial stage of restoration.

[0082] For example, when establishing gene coding sequences in the early stage of ecological restoration in copper mining areas, each gene locus corresponds to a prediction step size option, where the weekly step size corresponds to the high-frequency prediction genotype. This gene coding sequence is a representation method in genetic algorithms, used to simulate the biological evolution process to optimize problem solutions. After obtaining the initial fluctuation characteristics from historical carbon emission time series data, it is used to initialize the gene coding of multiple individuals in the population. This can bring higher predictive fitness because the initial fluctuation characteristics are often drastic. The initialization by the genetic algorithm can quickly generate diverse step size options, thereby improving the ability to capture the rapid carbon sink formation process.

[0083] In one possible implementation, the gene coding sequence is assumed to be a binary string, such as 000 representing a daily step size and 001 representing a weekly step size. Fluctuation features such as the peak change rate of carbon emissions are extracted from historical data to set the diversity of the initial population. This helps the algorithm avoid premature convergence in subsequent iterations and obtain a better step size configuration. The beneficial effect is that it improves the response speed of the prediction model to the dynamics of the early stage of ecological restoration.

[0084] In one possible implementation, after obtaining the weekly step-size prediction sequence for each individual from the initial population, high-frequency prediction is performed on the carbon emission time-series data. This high-frequency prediction refers to estimating carbon emissions at shorter time intervals, such as once a week. When calculating the degree of deviation between the predicted value and the actual carbon sink formation data, the degree of deviation is converted into a fitness value. The fitness value is an evaluation index in the genetic algorithm, used to measure the quality of individuals. It is usually calculated by the mean square error formula, such as the square average of the predicted value minus the actual value. Then, the genes of individuals with high fitness are retained in the population. This can bring the beneficial effect of selection pressure, because retaining high-fit individuals can simulate the natural selection mechanism and ensure that the algorithm evolves towards a more accurate step size.

[0085] In one possible implementation, for an individual genotype corresponding to a weekly step size, a prediction sequence is obtained, such as a weekly carbon emission prediction sequence, and the deviation is compared with the actual carbon sink data, such as the carbon absorption caused by vegetation growth. If the deviation is small, the fitness is high and it is retained, which helps to gradually optimize the prediction accuracy and synchronize with ecological evolution.

[0086] For example, when performing crossover and mutation operations on the preserved individual gene codes, crossover refers to exchanging two individual gene loci to produce new offspring, while mutation preferentially generates new genotypes within the weekly step size neighborhood. This neighborhood refers to the similar range of step size options, such as from weekly mutation to ten-day mutation. After forming the next generation population gene codes, high-frequency prediction and fitness values ​​are continued to be performed on the new individuals. This approach can bring beneficial effects of diversity and exploration, because crossover increases global search capability while mutation provides local adjustment, avoiding the algorithm from getting stuck in local optima and maintaining the match between step size and repair process.

[0087] In one possible implementation, for two retained individuals (such as genotypes 001 and 010), a new genotype 011 is generated by a single-point crossover. The weekly neighborhood mutation is slightly adjusted to 002, corresponding to a slightly longer step size. Then, the fitness of the new individual is calculated by predicting the carbon emission sequence. This helps the algorithm capture rapid changes in the early stages and gradually evolve and grow.

[0088] In one possible implementation, the crossover mutation and fitness evaluation process is repeated until the prediction step size corresponding to the dominant genotype in the population is stably converged to the weekly high-frequency prediction pattern. This convergence means that the genotypes of most individuals tend to be similar to the step size option, realizing the autonomous evolution of the prediction time granularity in the early stage of repair. This can bring the beneficial effect of optimizing stability because the repeated iteration simulates the evolutionary process, ensuring that the step size adapts to ecological changes rather than being a fixed setting.

[0089] In one possible implementation, if 80% of the individual genotypes converge to the weekly pattern after multiple generations of repetition, the iteration stops. This helps to form an efficient predictive framework in the early stages of copper mine restoration and lays the foundation for subsequent stable evolution.

[0090] The step-size evolutionary algorithm uses a fitness function to evaluate the synchronicity between the carbon emission prediction accuracy and changes in ecological indicators under different step sizes.

[0091] Carbon emission time-series records and ecological indicator change data were obtained from historical data on ecological restoration in the copper mining area. Preliminary divisions were made for prediction step lengths at different time granularities. A weekly step length was applied in the early stages of vegetation restoration to capture rapidly changing carbon sink characteristics. Simultaneously, the fluctuations of ecological indicators within the corresponding time periods were recorded, generating an initial step length and indicator correlation dataset. For this initial step length and indicator correlation dataset, a carbon emission error assessment system was constructed under the prediction step length. By comparing the deviations between the prediction results and actual carbon emission data for different step lengths, and considering the fluctuations of ecological indicators, a quantitative standard for the synchronicity between step length and indicator changes was determined, forming a step length synchronicity assessment result. Step length intervals with smaller errors were extracted from the step length synchronicity assessment results. The periodic characteristics of ecological indicator changes were further analyzed in these intervals, and the prediction step length was adjusted to match the frequency of indicator changes, forming an optimized step length configuration scheme. For the optimized step length configuration scheme, the prediction step length was gradually extended to the monthly or quarterly level during the soil microbial community stabilization stage. Simultaneously, the matching degree between the carbon emission prediction accuracy and ecological indicator changes was continuously monitored to ensure that the step length adjustment was consistent with the carbon balance succession stage of ecological restoration in the copper mining area.

[0092] In one possible implementation, when obtaining carbon emission time-series records and ecological indicator change data from historical data on ecological restoration in copper mining areas, vegetation cover and soil organic matter content can be collected as ecological indicators first. This information can be obtained by combining satellite remote sensing imagery with data from ground monitoring stations. For example, in the early stages of vegetation restoration, carbon emission fluctuations can be recorded weekly to capture rapid carbon sink formation. This ensures high data timeliness and is beneficial to the accuracy of subsequent step size division. Furthermore, preliminary division of prediction step sizes for different time granularities can be made, and weekly step sizes can be applied to the early stages of vegetation restoration to capture rapidly changing carbon sink characteristics. At the same time, the fluctuation of ecological indicators within the corresponding time period can be recorded, thereby generating an initial step size and indicator association dataset. This association dataset can reflect the dynamic relationship between carbon emissions and ecological restoration in the short term, which is beneficial to improving the adaptability of the prediction model.

[0093] In one possible implementation, when constructing a carbon emission error assessment system under a prediction step size for an initial step size and index-related dataset, the root mean square error can be used to compare the deviation between the prediction results and the actual carbon emission data corresponding to different step sizes. For example, the prediction error under a weekly step size can be compared with that under a monthly step size. Combined with the fluctuation of ecological indicators such as vegetation indices, a synchronicity quantification standard can be determined. This forms a step size synchronicity assessment result, which can reveal the impact of step size selection on prediction accuracy and help avoid the waste of computational resources caused by excessively fine step sizes. In addition, considering multiple aspects, such as using a small step size during the soil improvement period to better match rapid carbon flux changes, this contrasts with the long step size during the community succession period, ensuring the comprehensiveness of the assessment results.

[0094] In one possible implementation, after extracting step size intervals with smaller errors from the step size synchronization assessment results, the periodic characteristics of ecological indicator changes are further analyzed for these intervals. For example, the seasonal cycle of vegetation restoration can be identified through Fourier transform. Then, the prediction step size is adjusted to match the frequency of indicator changes, forming an optimized step size configuration scheme. This scheme can make the prediction more in line with the actual ecological process and is beneficial to the reliability of long-term restoration planning. Supported from another direction, such as the low error interval extracted during the pioneer plant planting period, can guide the fine-tuning of the step size. This is consistent with the analysis during the stable period, forming a consistent optimization logic.

[0095] In one possible implementation, for the optimized step size configuration scheme, when the prediction step size is gradually extended to the monthly or quarterly level during the stable stage of soil microbial community, the adjustment effect can be verified by continuously monitoring the matching degree between the carbon emission prediction accuracy and the changes in ecological indicators. For example, the matching degree can be assessed quarterly during the stable period to ensure that the step size adjustment is consistent with the carbon balance succession stage of ecological restoration in copper mining areas. This can dynamically adapt to the restoration process, which is beneficial to reduce prediction bias and optimize resource allocation. Further, from the perspective of carbon sink stability, this extension can capture long-term trends, while from the perspective of computational efficiency, it can reduce the burden of high-frequency prediction and form a multi-directional mutual support.

[0096] The step-size evolutionary algorithm uses a small step-size mutation strategy to capture the rapid carbon sink formation process in the early stage of vegetation restoration. As the soil microbial community stabilizes, it gradually evolves into a predicted genotype with a monthly or quarterly step size, so that the step-size evolutionary trajectory is consistent with the carbon balance succession stage of the mining area ecosystem.

[0097] In the early stages of vegetation restoration, a weekly time-series data acquisition scheme was constructed to address the rapid carbon sink formation process in the copper mining area. Real-time data on soil carbon flux and vegetation growth rate were obtained from environmental monitoring points, forming an initial dynamic carbon sink dataset. For this initial dataset, a small-step mutation approach was used to segment the time-series data, determining the fluctuation range of carbon sink changes within each time period and generating corresponding short-term predictive gene codes to reflect the rapidly changing carbon sink characteristics. During the gradual stabilization phase of the soil microbial community, stable trend characteristics of carbon sink changes were extracted from the short-term predictive gene codes. The prediction time granularity was adjusted, gradually transitioning to monthly predictive gene codes to generate a prediction time framework adapted to medium-term ecological succession. For this medium-term ecological succession prediction time framework, the long-term trend of the carbon balance succession stage was further analyzed, expanding the prediction time granularity to the quarterly level, forming a step-size evolutionary trajectory matching the carbon balance succession stage of the mining area's ecosystem.

[0098] In one possible approach, a weekly time-series data collection scheme is constructed to target the rapid carbon sink formation process in copper mining areas during the initial stage of vegetation restoration. Real-time change data on soil carbon flux and vegetation growth rate are obtained from environmental monitoring points to form an initial dynamic carbon sink dataset. This approach can capture the rapid accumulation of carbon sinks in a timely manner, which is beneficial to improving the sensitivity of predictions. For example, when vegetation in the mining area just begins to grow, weekly data collection can monitor the change in carbon flux from low to high values, thereby providing accurate basic data for subsequent processing and avoiding missing key change points.

[0099] For example, the initial dynamic carbon sink dataset is segmented using a small-step mutation method to determine the fluctuation range of carbon sink changes within each time period and generate corresponding short-term prediction gene codes that reflect the rapidly changing carbon sink characteristics. This helps to refine the analysis of fluctuations and improve the accuracy of predictions. For example, by dividing the data into multiple segments and calculating the amplitude differences, it is possible to discover the sudden increase in carbon sinks during the rainy season, thereby generating codes to optimize the initial prediction model.

[0100] In one possible implementation, during the gradual stabilization phase of the soil microbial community, stable trend features of carbon sink changes are extracted from short-term predictive gene coding. The prediction time granularity is adjusted, gradually transitioning to predictive gene coding on a monthly basis, generating a prediction time frame adapted to medium-term ecological succession. This approach aligns with the ecological stabilization process, helps reduce computational burden while maintaining accuracy. For example, if microbial activity tends to stabilize after trend extraction, weekly coding can be extended to monthly coding to ensure the framework is synchronized with the succession rhythm.

[0101] For example, by further analyzing the long-term trends of carbon balance succession stages in the prediction timeframe for medium-term ecological succession, the prediction time granularity is extended to the quarterly level, forming a step-size evolutionary trajectory that matches the carbon balance succession stage of the mining area ecosystem. This helps to achieve long-term consistency and is beneficial to the overall ecological restoration planning. For example, by extending the trend analysis to the quarterly level, seasonal carbon balance changes can be tracked, thereby making the trajectory accurately match the succession stage and improving the reliability of the prediction.

[0102] In one possible implementation, the initial dataset formed by the above acquisition scheme is directly used to generate codes in the segmented processing. This connection ensures the continuity of data from acquisition to processing and helps to avoid information loss. For example, the amplitude is calculated immediately after the weekly data is segmented, and the generated codes are extracted for adjustment in the stable phase, forming a chain to support the overall evolution.

[0103] For example, the segmented processing of small step size variation is linked to the subsequent extraction of trends. By using the fluctuation amplitude as a bridge to transfer fast features to stable features, it is beneficial to smooth the transition. For example, amplitude data helps to identify stable points, thereby maintaining the consistency of predictions when adjusting granularity.

[0104] In one possible implementation, the forecast timeframe generated from the monthly encoding is used to extend to quarterly in further analysis. This connection reinforces the evolutionary logic of the trajectory and is beneficial for dynamically adapting to ecological changes. For example, the trend of the medium-term framework is directly input into the long-term analysis to ensure that the step evolution is consistent with the carbon balance succession.

[0105] For example, the entire process forms a closed loop from initial data collection to quarterly expansion, with the output of each stage supporting the next, which is beneficial for comprehensively capturing the carbon sink process. For instance, the initial dataset is processed and adjusted to finally match the succession stage, achieving autonomous evolution of predictions.

[0106] This invention provides a mining area ecological restoration prediction system based on carbon emission time-series data, characterized in that it includes: The acquisition module is used to acquire historical carbon emission time-series data, which is used to reflect the fluctuation characteristics of carbon flux during ecological restoration. A module is established to analyze the fluctuation characteristics of the historical carbon emission time series data and to establish a dynamic matching model between the prediction step size and the ecological restoration stage, which includes the soil improvement period, the pioneer plant establishment period and the community succession period. The prediction module is used to apply differentiated prediction time granularity to different stages of ecological restoration based on the dynamic matching model, so that the prediction frequency is coordinated with the restoration demand cycle. The optimization module is used to adaptively optimize the prediction step size using a simulated annealing algorithm. The simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction requirements in the early stage of vegetation restoration, and gradually lowers the temperature to extend the prediction step size as the restoration process progresses. The design module is used to design a step-size evolutionary algorithm so that the prediction time granularity evolves autonomously with the repair process. In the initial stage, a high-frequency prediction mode with a weekly step size corresponding to the gene encoding is set.

[0107] It should be noted that the modules provided in the embodiments of the present invention have the same implementation principle and technical effects as those in the aforementioned method embodiments. For the sake of brevity, the specific working process of the modules described above can be referred to the corresponding process in the aforementioned method embodiments, and will not be repeated here.

[0108] Based on the embodiments of the present invention described above, and through the above description, those skilled in the art can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for predicting ecological restoration in mining areas based on carbon emission time-series data, characterized in that, include: Historical carbon emission time-series data are obtained, which are used to reflect the fluctuation characteristics of carbon flux during ecological restoration. The fluctuation characteristics of the historical carbon emission time series data were analyzed, and a dynamic matching model between the prediction step size and the ecological restoration stage was established. The ecological restoration stage includes the soil improvement period, the pioneer plant establishment period, and the community succession period. According to the dynamic matching model, differentiated prediction time granularity is adopted at different stages of ecological restoration to coordinate the prediction frequency with the restoration demand cycle. The simulated annealing algorithm is used to adaptively optimize the prediction step size. The simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction needs in the early stage of vegetation restoration, and gradually lowers the temperature to extend the prediction step size as the restoration process progresses. The step-size evolution algorithm is designed to enable the prediction time granularity to evolve autonomously with the repair process. In the initial stage, a high-frequency prediction mode with a weekly step size corresponding to the gene encoding is set.

2. The method as described in claim 1, characterized in that, The simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction requirements in the early stage of vegetation restoration, including: The simulated annealing algorithm evaluates a comprehensive objective function that balances the carbon emission prediction error and computational cost at the current step size in each iteration. The simulated annealing algorithm sets a small step size neighborhood search range during the soil improvement period to capture rapidly changing carbon flux, and expands the neighborhood during the stable period of community succession to allow for larger step size jumps.

3. The method as described in claim 2, characterized in that, The simulated annealing algorithm sets a small step-size neighborhood search range during the soil improvement period to capture rapidly changing carbon fluxes, including: The simulated annealing algorithm accepts step size changes within the step size neighborhood search range through a probability acceptance mechanism during the annealing process; The simulated annealing algorithm generates the prediction step size according to the probability acceptance mechanism, so that the prediction time granularity dynamically evolves with the carbon cycle characteristics of the mining area ecosystem.

4. The method as described in claim 1, characterized in that, The design step size evolution algorithm enables the prediction time granularity to evolve autonomously with the repair process, including: The step-size evolutionary algorithm uses a fitness function to evaluate the synchronicity between the carbon emission prediction accuracy and changes in ecological indicators under different step sizes.

5. The method as described in claim 4, characterized in that, The step-size evolutionary algorithm evaluates the synchronicity between the carbon emission prediction accuracy and changes in ecological indicators under different step sizes through a fitness function, including: The step-size evolutionary algorithm uses a small-step mutation strategy to capture the rapid carbon sink formation process in the early stage of vegetation restoration. The step-size evolutionary algorithm gradually evolves into a predictive genotype with a monthly or quarterly step size as the soil microbial community stabilizes. This ensures that the evolutionary trajectory of the step size is consistent with the carbon balance succession stage of the mining area ecosystem.

6. The method as described in claim 1, characterized in that, The analysis of the fluctuation characteristics of the historical carbon emission time-series data includes: The sensitivity of carbon flux changes during the ecological restoration phase is determined based on the fluctuation characteristics. The dynamic matching model correlates the prediction step size with the ecological restoration stage based on the sensitivity of carbon flux changes.

7. The method as described in claim 1, characterized in that, The step of employing differentiated prediction time granularity at different stages of ecological restoration based on the dynamic matching model includes: In the early stages of vegetation recovery, dense forecasting with weekly or monthly steps is used. Predictions using quarterly or annual steps are used during periods of ecosystem stability.

8. The method as described in claim 3, characterized in that, The simulated annealing algorithm accepts step size changes within the step size neighborhood search range through a probabilistic acceptance mechanism during the annealing process, including: The probability acceptance mechanism calculates the probability value of accepting the step size change based on the current temperature; If the probability value meets the preset conditions, the step size change is accepted to generate the predicted step size for the next iteration.

9. A prediction system for ecological restoration in mining areas based on carbon emission time-series data, characterized in that, include: The acquisition module is used to acquire historical carbon emission time-series data, which is used to reflect the fluctuation characteristics of carbon flux during ecological restoration. A module is established to analyze the fluctuation characteristics of the historical carbon emission time series data and to establish a dynamic matching model between the prediction step size and the ecological restoration stage, which includes the soil improvement period, the pioneer plant establishment period and the community succession period. The prediction module is used to apply differentiated prediction time granularity to different stages of ecological restoration based on the dynamic matching model, so that the prediction frequency is coordinated with the restoration demand cycle. The optimization module is used to adaptively optimize the prediction step size using a simulated annealing algorithm. The simulated annealing algorithm sets the initial temperature to correspond to the high-frequency prediction requirements in the early stage of vegetation restoration, and gradually lowers the temperature to extend the prediction step size as the restoration process progresses. The design module is used to design a step-size evolutionary algorithm so that the prediction time granularity evolves autonomously with the repair process. In the initial stage, a high-frequency prediction mode with a weekly step size corresponding to the gene encoding is set.

Images