Energy and double carbon management system and method for smart park

Abstract

The application discloses a smart park energy and double-carbon management system and method, relates to the technical field of energy management and carbon emission control, and comprises the following steps: collecting real-time operation data and carbon emission monitoring data, and performing unified formatting and denoising processing to generate standardized energy data and carbon emission data; inputting the standardized energy data and carbon emission data into a pre-constructed energy carbon emission model, and mapping various energy consumption and carbon emission factors to generate an energy-carbon mapping matrix; dynamically analyzing energy consumption rules and carbon emission changes at different times according to the energy-carbon mapping matrix to obtain analysis results, identifying peak load intervals and high emission periods, and forming preliminary optimization requirements; and processing the preliminary optimization requirements by using a prediction scheduling algorithm, and generating an energy optimization scheduling scheme through the collaborative operation mode of edge-side rapid prediction and cloud-end global optimization. The application can take into account energy efficiency and carbon emission reduction targets in the scheduling strategy.

Description

Technical Field

[0001] This invention relates to the field of energy management and carbon emission control technology, and in particular to a smart park energy and dual-carbon management system and method. Background Technology

[0002] With the continuous advancement of new energy technologies, smart park construction, and the "dual-carbon" strategy, the coordinated management of energy and carbon emissions has gradually become a research and application hotspot. In smart parks, energy consumption types are diversified, including electricity, heat, cooling, and renewable energy, and the generation process of carbon emissions has a complex mapping relationship with various energy consumption methods. Existing park energy and carbon emission management largely relies on distributed monitoring and centralized analysis, typically by collecting statistical data on electricity load, operating energy consumption, and emissions, combined with a carbon emission factor database for calculation, to assist in improving energy efficiency and controlling total carbon emissions in the park.

[0003] In the application of existing technologies, there are still some shortcomings in the management of energy and carbon emissions in industrial parks. On the one hand, the correlation between energy consumption and carbon emissions is often obtained through linear calculations or fixed factor mapping, which makes it difficult to fully capture dynamic changes, thus limiting the timeliness and precision of optimization strategies. On the other hand, in the process of energy dispatching, existing methods tend to emphasize global planning, while insufficient consideration is given to deviation correction and feedback optimization during actual execution. This can easily lead to inconsistencies between dispatching results and actual operating conditions, affecting the reliability of coordinated energy and carbon management. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a smart park energy and dual-carbon management method to solve the problems of low accuracy in energy and carbon emission mapping and large deviations between scheduling results and actual operation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a smart park energy and dual-carbon management method, which includes,

[0008] Collect real-time operational data and carbon emission monitoring data, and perform unified formatting and noise reduction processing to generate standardized energy data and carbon emission data;

[0009] Standardized energy data and carbon emission data are input into a pre-built energy carbon emission model, and various energy consumption and carbon emission factors are mapped to generate an energy-carbon mapping matrix.

[0010] Based on the energy-carbon mapping matrix, the energy consumption patterns and carbon emission changes at different times are dynamically analyzed to obtain the analysis results, identify peak load intervals and high emission periods, and form preliminary optimization requirements.

[0011] Predictive scheduling algorithms are used to process initial optimization requirements, and energy optimization scheduling schemes are generated through a collaborative operation mode of rapid prediction at the edge and global optimization in the cloud.

[0012] The energy optimization and scheduling scheme is allocated to various energy control links within the park. During the execution process, actual energy and carbon emission data are collected and transmitted back simultaneously, and compared with the energy-carbon mapping matrix to generate execution deviation information.

[0013] Based on the execution deviation information, the parameter thresholds and optimization weights of the predictive scheduling algorithm are adjusted to generate scenario-based energy and carbon emission time-series data. Through multi-objective robust optimization, a corrected energy optimization scheduling scheme is generated. Feasibility verification and operation comparison are performed in conjunction with actual energy and carbon emission data to generate energy and carbon control instructions.

[0014] As a preferred embodiment of the smart park energy and dual-carbon management method of the present invention, the real-time operating data collected includes electricity, heat, cooling capacity and gas consumption data;

[0015] The carbon emission monitoring data includes direct emission data and indirect emission factor data.

[0016] As a preferred embodiment of the smart park energy and dual-carbon management method described in this invention, the specific steps for generating standardized energy data and carbon emission data are as follows:

[0017] The real-time operating data is time-aligned, outlier removed, dimensionless, smoothed, denoised, and filtered to generate standardized energy data.

[0018] Carbon emission monitoring data is processed by time alignment, outlier removal, missing value completion, and data interpolation to generate carbon emission data.

[0019] As a preferred embodiment of the smart park energy and dual-carbon management method described in this invention, the pre-constructed energy carbon emission model is based on a deep neural network architecture. After receiving standardized energy data and carbon emission data, a high-dimensional feature representation of energy consumption and carbon emission factors is obtained through a multi-dimensional feature extraction layer. An abnormal pattern elimination mechanism is introduced in the time series modeling and dynamic analysis process. The nonlinear mapping relationship between energy consumption and carbon emission factors is fitted using a deep learning network and then integrated and normalized to construct the model.

[0020] As a preferred embodiment of the smart park energy and dual-carbon management method of the present invention, the specific steps for generating the energy-carbon mapping matrix are as follows:

[0021] Standardized energy data and carbon emission data are input into the energy carbon emission model to perform preliminary mapping and matching of various energy and carbon emission indicators, generating an original correspondence matrix.

[0022] The original correspondence matrix is ​​dynamically weighted and multi-dimensional features are fused to adjust the correlation strength between energy categories and carbon emission factors, thereby generating a weighted correlation matrix.

[0023] By using multi-scale time series analysis and deep learning prediction methods, the weighted correlation matrix is ​​fitted and optimized to eliminate noise and abnormal patterns and generate an accurate correspondence matrix.

[0024] The precise correspondence matrix is ​​normalized and integrated through multidimensional mapping to generate an energy-carbon mapping matrix.

[0025] As a preferred embodiment of the smart park energy and dual-carbon management method described in this invention, the specific steps for forming preliminary optimization requirements are as follows:

[0026] The energy-carbon mapping matrix is ​​decomposed according to different time scales of day, week and month to generate energy consumption sequence and carbon emission sequence.

[0027] Dynamic pattern recognition and trend analysis are performed on energy consumption and carbon emission sequences to extract periodic fluctuations, abnormal peaks and potential change patterns, and generate time series analysis results.

[0028] Based on the time series analysis results, the energy consumption series and carbon emission series are sorted and peak statistics are performed according to the time scale to identify peak load intervals and high emission periods. The energy consumption series and carbon emission series at the corresponding time scale are then sorted to form preliminary optimization requirements.

[0029] As a preferred embodiment of the smart park energy and dual-carbon management method of the present invention, the specific steps for generating the energy optimization scheduling scheme are as follows:

[0030] Based on the initial optimization requirements, a predictive scheduling algorithm is used to make short-term predictions of energy consumption and carbon emission sequences, generating rapid prediction results at the edge.

[0031] Based on the rapid prediction results from the edge side, the energy consumption and carbon emission sequences for each time period are integrated, and global optimization results are generated through global optimization calculations.

[0032] The global optimization results are compared and adjusted with the preliminary optimization requirements to generate an energy optimization scheduling scheme.

[0033] As a preferred embodiment of the smart park energy and dual-carbon management method of the present invention, the specific steps for generating execution deviation information are as follows:

[0034] The energy optimization scheduling scheme is transformed into control instructions and distributed to various energy control links, generating feedback energy and emission data.

[0035] The feedback energy and emission data are time-synchronized, bias-corrected, and filtered to obtain standardized energy and emission data.

[0036] By comparing standardized energy and emission data with the carbon mapping matrix, the magnitude of the difference between energy and emissions is calculated, and execution deviation information is generated.

[0037] As a preferred embodiment of the smart park energy and dual-carbon management method of the present invention, the specific steps for generating energy and carbon control instructions are as follows:

[0038] Extract a hierarchical set of deviation indicators from the execution deviation information, and recalibrate the parameter thresholds and optimization weights of the predictive scheduling algorithm to generate an updated parameter set;

[0039] The updated parameter set is dynamically reorganized and extended over time to generate scenario-based time-series data on energy and carbon emissions. Multi-objective robust optimization calculations are then performed to obtain the corrected energy optimization scheduling scheme.

[0040] The revised scheduling scheme is tested for feasibility and compared with actual operation. Combined with actual energy and carbon emission data, energy and carbon control instructions are generated.

[0041] Secondly, this invention provides a smart park energy and dual-carbon management system, including,

[0042] The data acquisition module is used to collect real-time operational data and carbon emission monitoring data, and perform unified formatting and noise reduction processing to generate standardized energy data and carbon emission data.

[0043] The matrix mapping module is used to input standardized energy data and carbon emission data into a pre-built energy carbon emission model, calculate the correspondence between various energy consumption and carbon emission factors, and generate an energy-carbon mapping matrix.

[0044] The demand analysis module is used to dynamically analyze the energy consumption patterns and carbon emission changes at different times based on the energy-carbon mapping matrix, obtain analysis results, identify peak load intervals and high emission periods, and form preliminary optimized demand.

[0045] The scheduling generation module is used to process the initial optimization requirements using predictive scheduling algorithms, and generate energy optimization scheduling schemes through a collaborative operation mode of rapid prediction at the edge and global optimization in the cloud.

[0046] The deviation comparison module is used to allocate energy optimization scheduling schemes to various energy control links within the park. During the execution process, it synchronously collects and transmits actual energy and carbon emission data, and then compares them with the energy-carbon mapping matrix to generate execution deviation information.

[0047] The closed-loop optimization module is used to adjust the parameter thresholds and optimization weights of the predictive scheduling algorithm based on the execution deviation information, generate scenario-based energy and carbon emission time-series data, perform multi-objective robust optimization, generate a corrected energy optimization scheduling scheme, and perform feasibility verification and operation comparison with actual energy and carbon emission data to generate energy and carbon control instructions.

[0048] The beneficial effects of this invention are as follows: by inputting standardized energy data and carbon emission data into an energy carbon emission model with a deep neural network architecture, and combining dynamic weighting and multi-scale time series analysis to generate an energy-carbon mapping matrix, a precise nonlinear and multi-dimensional fitting between energy consumption behavior and carbon emission factors is achieved. This effectively bridges the data gap between the energy side and the carbon emission side, providing a high-precision causal relationship basis for energy consumption pattern analysis and optimized scheduling, and significantly improving the accuracy and stability of energy-carbon correspondence modeling, enabling scheduling strategies to take into account both energy efficiency and carbon emission reduction targets. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 A flowchart for energy and dual-carbon management methods in smart parks.

[0051] Figure 2 This is a schematic diagram of the smart park's energy and dual-carbon management system.

[0052] Figure 3 A flowchart for constructing an energy carbon emission model.

[0053] Figure 4 The flowchart for generating the carbon mapping matrix. Detailed Implementation

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0057] Reference Figures 1-4 This is one embodiment of the present invention, which provides a smart park energy and dual-carbon management method, including the following steps:

[0058] S1. Collect real-time operational data and carbon emission monitoring data, and perform unified formatting and noise reduction processing to generate standardized energy data and carbon emission data.

[0059] S1.1 Collect real-time operating data, including electricity, heat, cooling capacity, and gas consumption data.

[0060] Specifically, within the park, electricity, heat, cooling, and gas consumption data are read hourly according to the collection frequency. A unified timestamp is added to each data point, such as using minute-level or hour-level recording to form a continuous time series. The electricity, heat, cooling, and gas consumption data are entered into a unified data record table in the order of collection, ensuring that the field formats of each data point are consistent. The entered data is then checked for completeness, such as checking for missing records or format abnormalities, and anomalies are marked when detected, thus providing an input basis for generating standardized energy data.

[0061] S1.2 Carbon emission monitoring data includes direct emission data and indirect emission factor data.

[0062] Specifically, direct emission data is collected hourly, such as recording emissions from gas combustion or fuel consumption on an hourly basis, and a unified timestamp is added to each direct emission data point. Simultaneously, indirect emission factor data is acquired, such as recording emission factors corresponding to electricity consumption or heat supply, and a unified timestamp is added during acquisition. The direct emission data and indirect emission factor data are aligned according to the timestamps to form one-to-one corresponding record entries. The aligned direct emission data and indirect emission factor data are then entered into a unified data record table according to the field format, and integrity checks are performed, such as detecting missing timestamps or numerical anomalies, and marking any anomalies detected, thus providing an input basis for subsequent carbon emission data generation.

[0063] S1.3 Perform time alignment, outlier removal, unit unification, smoothing, noise reduction, and filtering on real-time operating data to generate standardized energy data.

[0064] Specifically, the real-time operational data, including electricity, heat, cooling, and gas consumption data, is sorted in ascending order by timestamp. When timestamps are discontinuous, interpolation is used to fill in missing time points, achieving time alignment. Outliers are detected field-by-field in the aligned real-time operational data; for example, a three-standard-deviation method is used to identify electricity or heat data exceeding reasonable ranges, and these outliers are removed. The real-time operational data from different sources undergoes unitization processing; for example, electricity data is converted from kilowatt-hours to megawatt-hours, and gas consumption data from cubic meters to standard cubic meters. The real-time operational data is then smoothed and denoised; for example, a moving average method is used to smooth the electricity data, and a Butterworth low-pass filter is used to filter the heat and cooling data. Finally, the electricity, heat, cooling, and gas consumption data, after time alignment, outlier removal, unitization, smoothing, denoising, and filtering, are merged into a unified structure to generate standardized energy data.

[0065] S1.4 Perform time alignment, outlier removal, missing value completion, and data interpolation on the carbon emission monitoring data to generate carbon emission data.

[0066] Specifically, the direct emission data and indirect emission factor data in the carbon emission monitoring data are arranged in ascending order according to timestamps. When there are discontinuous timestamps, missing time points are filled in using linear interpolation to complete time alignment. Anomaly detection is performed on the aligned direct emission data and indirect emission factor data, for example, using box plots to identify and remove data points that exceed the upper and lower quartile ranges. Missing values ​​are filled in the carbon emission monitoring data after anomaly removal, for example, using nearest neighbor completion to fill in missing direct emission data values. Interpolation is performed on the filled carbon emission monitoring data, for example, using cubic spline interpolation to smooth and fill in points for indirect emission factor data. The direct emission data and indirect emission factor data after time alignment, anomaly removal, missing value completion, and data interpolation are merged into a unified format to generate carbon emission data.

[0067] S2. Input standardized energy data and carbon emission data into a pre-built energy carbon emission model, and map various energy consumption and carbon emission factors to generate an energy-carbon mapping matrix.

[0068] S2.1 The pre-built energy carbon emission model is based on a deep neural network architecture. It receives standardized energy data and carbon emission data, obtains high-dimensional feature representations of energy consumption and carbon emission factors through a multi-dimensional feature extraction layer, introduces an anomaly pattern removal mechanism in the time series modeling and dynamic analysis process, uses a deep learning network to fit the nonlinear mapping relationship between energy consumption and carbon emission factors, and integrates and normalizes the data to build the model.

[0069] Specifically, standardized energy data and carbon emission data are subjected to time-series windowing and multi-dimensional feature extraction. This includes calculating time-domain features (mean, variance, peak value, energy, etc.), frequency-domain features (main frequency components extracted through fast Fourier transform), and cumulative statistical features within each time window to form a high-dimensional feature vector set. The aforementioned high-dimensional feature vector set is then subjected to dimensionality reduction and normalization, for example, by using principal component analysis or zero-mean unit variance standardization to reduce dimensionality and unify dimensions. Anomaly pattern removal is performed on the normalized feature vector set, for example, by using isolated forest or box plot rules to identify and remove outlier samples, and the removed features are used as training data.

[0070] The features after anomaly removal are divided into training, validation, and test sets according to time series, and carbon emission data is used as the supervision label. A temporal modeling network in a deep neural network architecture (such as a long short-term memory network or a temporal convolutional network, combined with an attention mechanism if necessary) is used for fitting and training. During training, the mean squared error loss function is used, supplemented by regularization and early stopping strategies for hyperparameter optimization and to prevent overfitting. The trained energy carbon emission model is evaluated on the validation and test sets. The mean absolute error is calculated by averaging the absolute values ​​of the pointwise differences between the predicted and actual carbon emission values, expressed as:

[0071] ;

[0072] in, Indicates the mean absolute error. This represents the total carbon emissions. An index representing carbon emission values, The first carbon emission predicted by the energy carbon emission model One carbon emission value, Indicates the actual measurement of the first One carbon emission value;

[0073] The root mean square error (RMSE) is calculated by averaging and taking the square root of the squared differences between the predicted and actual values. The expression is:

[0074] ;

[0075] in, Indicates the root mean square error;

[0076] according to and The calculation results can be used to adjust hyperparameters of the deep neural network, such as the number of network layers, hidden layers, learning rate, time step, and regularization coefficient. For example, when the error is large, the number of hidden layers or the number of convolutional kernels in the temporal convolutional layer can be increased to enhance the fitting ability, or the learning rate and regularization coefficient can be adjusted to improve the convergence effect. When the error reaches the preset performance index, the combination of deep neural network and hyperparameters is considered to meet the preset performance index. The output of the energy carbon emission model is normalized and integrated with the original features of standardized energy data and carbon emission data to form an energy carbon mapping matrix for subsequent dynamic analysis and scheduling, thereby completing the construction of the energy carbon emission model.

[0077] It should also be noted that the preset performance indicators can be set based on the statistical characteristics of historical standardized energy data and carbon emission data, as well as the accuracy requirements of the prediction task. Specific steps include: segmenting historical energy and carbon emission data into time series; calculating the error distribution between the actual observed values ​​and the preliminary predicted values ​​at each time step to obtain the statistical range of the mean absolute error (MAE) and root mean square error (RMSE); selecting the upper limit of the error distribution as a performance indicator reference based on the actual accuracy requirements of the park's energy dispatch and carbon emission control, for example, setting MAE ≤ 0.03 or RMSE ≤ 0.08 as an example range; using the error range as a judgment criterion during the training and optimization process of the deep neural network to determine whether the network structure and hyperparameter combination meet the preset performance indicators, thereby ensuring that the prediction accuracy of the energy carbon emission model output is suitable for subsequent energy-carbon mapping matrix generation and dynamic dispatch analysis.

[0078] The anomaly pattern removal mechanism identifies outliers that deviate from the normal range in time series or multidimensional features and removes abnormal data during the training or prediction process of deep neural networks to ensure the fitting accuracy of the mapping between energy consumption and carbon emission factors.

[0079] S2.2 Input standardized energy data and carbon emission data into the energy carbon emission model, perform preliminary mapping and matching of various energy and carbon emission indicators, and generate the original correspondence matrix.

[0080] Specifically, standardized energy data and carbon emission data are sequentially input into the energy carbon emission model according to time steps. Through a multi-dimensional feature extraction layer, high-dimensional feature representations of energy data such as electricity, heat, cooling capacity, and gas consumption, as well as direct emission data and indirect emission factor data, are obtained respectively. At each time step, preliminary correlation coefficients between various energy features and carbon emission features are calculated, for example, using the Pearson correlation coefficient, to form a preliminary matching matrix between energy categories and carbon emission indicators. The preliminary matching results of all time steps are summarized according to the dimensions of energy category and carbon emission indicator to generate the original correspondence matrix.

[0081] S2.3. Perform dynamic weighting and multi-dimensional feature fusion on the original correspondence matrix to adjust the correlation strength between energy category and carbon emission factor, and generate a weighted correlation matrix.

[0082] Specifically, the original correspondence matrix is ​​assigned dynamic weights according to the time series and energy category dimensions. For example, the weight of recent data or high fluctuation periods is set to 0.6. Energy characteristics such as electricity, heat, cooling capacity and gas consumption are fused with high-dimensional features of direct emission data and indirect emission factor data in a multidimensional manner. For example, the correlation and importance between different features are uniformly calculated by feature splicing and weighted summation methods to obtain the adjusted correlation strength matrix between energy category and carbon emission factor, forming a weighted correlation matrix.

[0083] S2.4. Using multi-scale time series analysis and deep learning prediction methods, the weighted correlation matrix is ​​fitted and optimized to eliminate noise and abnormal patterns and generate an accurate correspondence matrix.

[0084] Specifically, the weighted correlation matrix is ​​decomposed into time series at different time scales of daily, weekly, and monthly, and trend, seasonality, and residual components are extracted respectively. For example, the daily residual is set to ±0.05 as the noise range. The decomposed energy consumption and carbon emission sequences are trained using a deep neural network. The decomposed sequences are input into the network, and the fitted values ​​at each time point are calculated through forward propagation. The results are dynamically corrected based on the residuals between the fitted values ​​and the actual observed values. Abnormal patterns that exceed the residual threshold are removed and smoothed. Finally, the prediction results at each time scale are weighted and fused to generate an accurate correspondence matrix after noise and abnormal patterns are eliminated.

[0085] It should also be explained that the specific steps for setting the residual threshold are as follows: the weighted correlation matrix is ​​decomposed into a time series using wavelet decomposition or moving average methods at different time scales such as daily, weekly, or monthly, and the trend component and residual component are extracted; the residual mean and residual standard deviation are obtained by calculating the mean and standard deviation of the residual components, respectively, and the residual threshold range is determined by interval estimation method based on the residual distribution characteristics. For example, the residual mean ± 2 times the residual standard deviation is taken as the residual threshold. The example range of the preset residual threshold is ±0.05, which is used for subsequent abnormal pattern removal and smoothing.

[0086] S2.5. The exact correspondence matrix is ​​normalized and integrated through multidimensional mapping to generate an energy-carbon mapping matrix.

[0087] Specifically, each feature dimension of the precise correspondence matrix is ​​normalized, and the min-max normalization method is used to unify the feature scales to comparable dimensions, outputting a normalized feature matrix. The normalized feature matrix and the weighted correlation matrix are then weighted and fused according to the energy category and carbon emission factor dimensions, using element-level multiplication and row-wise methods to obtain a preliminary mapping array. The preliminary mapping array is then integrated into a multidimensional mapping. First, the time dimension, energy category dimension, and factor dimension are expanded as tensors to form a high-dimensional feature tensor. Then, singular value decomposition is used to reduce the dimensionality and compress the features of the high-dimensional feature tensor, outputting a set of dimensionality-reduced mapping vectors. Finally, the dimensionality-reduced mapping vectors are reconstructed into a matrix form according to energy category and time order, and the reconstructed matrix is ​​normalized again to ensure dimensional consistency, outputting an energy-carbon mapping matrix.

[0088] S3. Based on the energy-carbon mapping matrix, conduct dynamic analysis on the energy consumption patterns and carbon emission changes at different times, obtain analysis results, identify peak load intervals and high emission periods, and form preliminary optimization requirements.

[0089] S3.1 Decompose the energy-carbon mapping matrix according to different time scales of day, week and month to generate energy consumption sequence and carbon emission sequence.

[0090] Specifically, the energy-carbon mapping matrix is ​​resampled and aggregated into initial matrices at three frequencies: daily, weekly, and monthly, based on time indices. For example, it is aggregated daily to obtain the daily total or daily average for each energy category and factor, and outputs the daily-scale initial matrix; it is aggregated weekly and outputs the weekly-scale initial matrix; and it is aggregated monthly and outputs the monthly-scale initial matrix. The seasonal-trend decomposition method is used to separate the trend component, seasonal component, and residual component from the daily-scale initial matrix, and the trend and seasonal components of the daily-scale energy consumption and carbon emission sequences are extracted and output. The seasonal-trend decomposition method is repeated on the weekly-scale initial matrix to extract the weekly-scale trend and seasonality, and the weekly-scale energy consumption and carbon emission sequences are output. The same method is used to extract and output the monthly-scale energy consumption and carbon emission sequences from the monthly-scale initial matrix. The trend components at each time scale and the smoothed seasonal or residual components are reconstructed into complete energy consumption and carbon emission sequences based on time series.

[0091] S3.2 Perform dynamic pattern recognition and trend analysis on energy consumption and carbon emission sequences, extract periodic fluctuations, abnormal peaks and potential change patterns, and generate time series analysis results.

[0092] Specifically, for energy consumption and carbon emission series, a sliding window statistical method is used to segment the series according to a preset window length and step size (e.g., a window length of 24 hours and a step size of 1 hour). The mean, variance, standard deviation, and autocorrelation coefficient within each window are calculated to form local statistical characteristics. The Fast Fourier Transform (FFT) method is used to perform frequency domain analysis on the series within each window, calculating the amplitude and frequency spectra to extract periodic fluctuation characteristics. A threshold method is used to identify abnormal peaks; when the instantaneous energy consumption or carbon emission value exceeds the window mean ± 3 times the standard deviation, it is marked as an abnormal peak. A trend decomposition method is used to separate the seasonal, trend, and residual components of the daily, weekly, and monthly timescale series. Linear regression and seasonal decomposition methods are used to calculate the trend change rate and seasonal amplitude changes, forming potential variation patterns. The local statistical characteristics, periodic fluctuations, abnormal peaks, and potential variation patterns are integrated in chronological order to generate time series analysis results.

[0093] It should also be noted that seasonal decomposition is a method that decomposes a time series into trend, seasonal, and residual components for analyzing periodic fluctuations and patterns.

[0094] S3.3 Based on the time series analysis results, sort and statistically analyze the energy consumption series and carbon emission series according to the time scale, identify peak load intervals and high emission periods, and sort them together with the energy consumption series and carbon emission series at the corresponding time scale to form preliminary optimization requirements.

[0095] Specifically, the energy consumption and carbon emission sequences are sorted according to different time scales of day, week, and month. Based on the sorting results, the peak values ​​at each time scale are counted, and the time intervals corresponding to the peak values ​​are identified as peak load intervals and high emission periods. The energy consumption and carbon emission sequences at each time scale are sorted in descending order according to the size of the peak values, and preliminary optimization requirements are formed by combining the time intervals where the peak values ​​are located.

[0096] S4. Using predictive scheduling algorithms, the initial optimization requirements are processed, and an energy optimization scheduling scheme is generated through the collaborative operation of rapid prediction at the edge and global optimization in the cloud.

[0097] S4.1 Based on the preliminary optimization requirements, the predictive scheduling algorithm is used to make short-term predictions of the energy consumption sequence and carbon emission sequence, generating rapid prediction results at the edge.

[0098] Specifically, based on the initial optimization requirements, the energy consumption sequence and carbon emission sequence are segmented into time series. The segmented historical sequence is used as input features. Short-term prediction input is constructed based on the preset time step and sliding window length. The prediction algorithm is used for step-by-step iterative calculation. At each prediction time step, the energy consumption value and carbon emission value are calculated. The prediction results are combined to perform error verification and adjustment, and generate rapid prediction results on the edge side, such as predicting the corresponding values ​​of the energy consumption sequence and carbon emission sequence in the next 1 hour or the next 24 hours.

[0099] It should also be explained that the specific steps for setting the time step and sliding window length are as follows: Determine the time granularity based on the sampling frequency of the energy consumption and carbon emission sequences, for example, collecting data every 5 minutes; statistically analyze the typical change cycles and short-term fluctuation amplitudes of historical sequences, such as intraday peak cycles or weekly patterns, in order to select an appropriate prediction range; determine the time step, for example, choosing 5 minutes or 15 minutes as the prediction interval; determine the sliding window length based on the prediction accuracy requirements and sequence length, for example, selecting data from the past 12 or 24 time steps as prediction input; apply the time step and sliding window length to sequence segmentation and input construction, ensuring that each sliding window covers continuous historical data and moves according to the set time step to generate the next set of prediction inputs;

[0100] Prediction algorithms are calculation methods that use mathematical or statistical methods to extrapolate future values ​​based on historical energy consumption and carbon emission sequences.

[0101] S4.2 Based on the rapid prediction results from the edge side, the energy consumption sequence and carbon emission sequence for each time period are integrated, and a global optimization result is generated through global optimization calculation.

[0102] Specifically, the edge-side rapid prediction results are aligned in chronological order to ensure that the energy consumption sequence and carbon emission sequence correspond consistently for each time period. For overlapping or missing time period data, linear weighted averaging or interpolation methods are used for integration. For example, the weighted average of the prediction results for 5 consecutive minutes is taken to generate an integrated sequence. A global optimization calculation input table is constructed, and the integrated energy consumption sequence and carbon emission sequence are used as parameters for constraints and objective functions. Based on historical operating data and energy management strategies, energy allocation constraints, carbon emission limits, and load demands are obtained, and initial conditions are set. The optimization algorithm is executed to generate global optimization results, including the energy consumption allocation and corresponding carbon emission prediction values ​​for each time period.

[0103] S4.3 Compare and fine-tune the global optimization results with the preliminary optimization requirements to generate an energy optimization scheduling scheme.

[0104] Specifically, the global optimization results and preliminary optimization requirements are aligned on the same time scale to ensure that the energy consumption and carbon emission data for each time period are consistent; the difference value is calculated for each time period; based on the difference value and the preset energy consumption deviation threshold and carbon emission deviation threshold, the energy allocation and load strategies are adjusted for time periods where the deviation exceeds the energy consumption deviation threshold and carbon emission deviation threshold. For example, the adjustable energy allocation is increased for peak load intervals, and the energy supply is reduced for low load intervals. At the same time, the carbon emission allocation is updated to meet the total constraint; the adjusted energy consumption and carbon emission data for each time period are integrated to generate a complete energy optimization scheduling scheme.

[0105] It should also be explained that the specific steps for setting the energy consumption deviation threshold and carbon emission deviation threshold are as follows: Collect energy consumption and carbon emission sequences from historical operating data, perform statistical analysis according to daily, weekly, and monthly time scales, and calculate the average and standard deviation for each time period. For example, the daily average energy consumption is 1000 kWh, and the standard deviation is 50 kWh. Determine the deviation tolerance range based on historical fluctuations and management requirements. For example, the energy consumption deviation threshold is set to ±2 times the average standard deviation, and the carbon emission deviation threshold is set to ±1.5 times the average standard deviation. Compare and verify the set deviation range with historical data to ensure that it can cover most normal fluctuations without affecting the feasibility of the control strategy. Record the preset energy consumption deviation threshold and carbon emission deviation threshold for subsequent deviation judgment and strategy adjustment in each time period of the energy optimization scheduling scheme. Energy consumption deviation threshold: If the daily average energy consumption is 1000 kWh, the energy consumption deviation threshold can be set to ±50~±100 kWh; if the weekly average energy consumption is 7000 kWh, the energy consumption deviation threshold can be set to ±350~±700 kWh. Carbon emission deviation threshold: If the daily average carbon emission is 500 kg, the carbon emission deviation threshold can be set to ±25~±50 kg; if the weekly average carbon emission is 3500 kg, the carbon emission deviation threshold can be set to ±175~±350 kg.

[0106] It should be noted that the two-layer collaborative architecture effectively solves the problem of balancing real-time performance and global performance, avoiding the computational latency inherent in cloud-based optimization alone, and overcoming the limitations of simple edge prediction in covering global energy consumption characteristics. On the one hand, it improves the accuracy and adaptability of energy dispatching schemes at different time scales; on the other hand, it achieves a balance between rapid response and global optimization in park energy management, providing more efficient and reliable technical support for the low-carbon operation of smart parks.

[0107] S5. Allocate the energy optimization scheduling scheme to various energy control links within the park. During the execution process, collect and transmit actual energy and carbon emission data simultaneously, and compare it with the energy-carbon mapping matrix to generate execution deviation information.

[0108] S5.1 Transform the energy optimization scheduling scheme into control instructions and distribute them to each energy control link, generating feedback energy and emission data.

[0109] Specifically, the energy optimization scheduling scheme is analyzed into specific energy consumption and carbon emission control amounts by time period. For example, 50 kWh of adjustable energy is allocated to the peak load range, and 30 kWh of energy supply is reduced to the low load range. Based on the correspondence between various energy types (such as electricity, natural gas, and steam) and emission indicators (such as carbon dioxide emissions, carbon monoxide emissions, and nitrogen oxide emissions), the control amounts are mapped to the operating parameters of specific energy control links. After the operation of each energy control link is executed, energy consumption data and carbon emission data of each link are collected in sequence. The feedback energy and emission data of each time period are summarized in chronological order to form a complete feedback energy and emission data sequence.

[0110] S5.2 Perform time synchronization, deviation correction, and filtering on the feedback energy and emission data to obtain standardized energy and emission data.

[0111] Specifically, the feedback energy and emission data are aligned by timestamp to ensure that the energy consumption data and carbon emission data correspond consistently across time periods. The deviation between the data for each time period and the reference energy consumption and carbon emission sequences is calculated, and energy consumption and carbon emission data exceeding the example ±5% deviation range are corrected based on the deviation value. The corrected data is smoothed using a moving average filtering method, for example, by taking the average of the data from three adjacent time steps for each time period. The corrected and filtered energy consumption and carbon emission data for each time period are integrated in chronological order to generate standardized energy and emission data.

[0112] S5.3. Compare standardized energy and emission data with the carbon mapping matrix, calculate the difference between energy and emissions, and generate execution deviation information.

[0113] Specifically, standardized energy and emission data are compared with the corresponding energy categories and carbon emission factors in the energy-carbon mapping matrix according to time periods; the difference between the energy consumption value and the corresponding value in the energy-carbon mapping matrix for each time period is calculated, for example, using the absolute difference calculation method or the relative percentage difference calculation method; for the carbon emission data for each time period, the difference is calculated using both the absolute difference calculation method and the relative percentage difference calculation method; the difference in energy consumption and carbon emission for each time period is integrated according to energy category and carbon emission factor to form execution deviation information, which can be calculated hourly and recorded in matrix form.

[0114] S6. Based on the execution deviation information, adjust the parameter thresholds and optimization weights of the predictive scheduling algorithm, generate scenario-based energy and carbon emission time-series data, generate a corrected energy optimization scheduling scheme through multi-objective robust optimization, and perform feasibility verification and operation comparison in combination with actual energy and carbon emission data to generate energy and carbon control instructions.

[0115] S6.1 Extract the hierarchical deviation index set from the execution deviation information, and recalibrate the parameter thresholds and optimization weights of the predictive scheduling algorithm to generate the updated parameter set.

[0116] Specifically, the execution deviation information is hierarchically summarized by time period, energy category, and carbon emission factor to form a hierarchical data matrix. For each hierarchical data matrix, the average deviation value is calculated using the average method, the maximum deviation value is calculated using the maximum value method, and the standard deviation is calculated using the standard deviation formula to generate a hierarchical deviation index set. Based on the hierarchical deviation index set, the threshold values ​​of each parameter in the updated parameter set are calculated and compared. For example, each parameter threshold is compared one by one with the average deviation value, maximum deviation value, and standard deviation of the corresponding energy category or carbon emission factor, and adjustments are made according to the comparison results. For example, the threshold values ​​of parameters with large deviation values ​​are increased or decreased by a fixed proportion, while the threshold values ​​of parameters with small deviation values ​​are kept at their original values ​​or finely adjusted. The value and proportion of each adjustment are recorded. The optimization weights of all energy categories and carbon emission factors are corrected using a proportional adjustment method to keep the total optimization weights balanced, and the edge prediction step size and sliding window length are corrected using time series analysis to match the deviation characteristics. The updated parameter thresholds and optimization weights are summarized to form the updated parameter set.

[0117] It should also be noted that the parameter threshold is a numerical limit used in the predictive scheduling algorithm to limit the allowable deviation range of the energy consumption sequence and carbon emission sequence. Its function is to judge the difference between the predicted value and the actual value. When the deviation exceeds the parameter threshold, the predictive scheduling algorithm will trigger corresponding correction or optimization operations to ensure the stability and accuracy of the scheduling scheme. The parameter threshold can be set with different values ​​for different energy categories, time periods or carbon emission factors. For example, a deviation of ±5% can be set for the hourly energy consumption sequence and a deviation of ±3% can be set for the carbon emission sequence, thereby forming a control basis for dynamically adjusting the predictive scheduling strategy.

[0118] S6.2. Dynamically reorganize and extend the updated parameter set in time series to generate scenario-based energy and carbon emission time series data, and perform multi-objective robust optimization calculations to obtain the corrected energy optimization scheduling scheme.

[0119] Specifically, the updated parameter set is dynamically reorganized according to the time series. Each updated parameter set is mapped to a corresponding time period based on different energy categories and carbon emission factors. Linear interpolation is used for time series expansion to generate scenario-based energy and carbon emission time series data. The generated time series data is robustly optimized using multi-objective optimization algorithms, such as Pareto optimization, according to the two objectives of minimizing energy consumption and carbon emissions. During the optimization process, the energy allocation, the updated parameter thresholds for load adjustment, and the carbon emission allocation ratio are iteratively adjusted to ensure that the energy consumption and carbon emissions in each time period meet the total constraints and peak-valley load limits. The optimized energy consumption and carbon emission values ​​are integrated into a matrix or table to generate a corrected energy optimization scheduling scheme. The example records the allocation and corresponding carbon emissions for each energy category with an hourly time step.

[0120] S6.3. Conduct feasibility verification and compare the revised scheduling scheme with actual operation, and generate energy and carbon control instructions based on actual energy and carbon emission data.

[0121] Specifically, the revised energy optimization scheduling scheme is unfolded according to time steps. Based on the energy category and carbon emission factor corresponding to each time period, the energy consumption and carbon emissions in the scheduling scheme are compared with standardized energy and emission data collected during actual operation. The deviation value for each time period is calculated using the absolute difference method. Based on preset energy consumption deviation thresholds and carbon emission deviation thresholds, the feasibility of the scheduling scheme for each time period is determined. For time periods where the deviation exceeds the energy consumption deviation threshold and carbon emission deviation threshold, the energy allocation and load strategy are adjusted, and the carbon emission allocation ratio is corrected accordingly. The adjusted time period energy consumption and carbon emission data are integrated by energy category. By mapping the energy consumption value corresponding to each time period to identifiable operational parameters of the energy control link, such as gas volume, electricity load, or heat supply, and mapping carbon emissions to emission control indicators or emission quotas, the conversion from numerical values ​​to control commands is achieved. A set of control values ​​is generated every hour, forming an energy and carbon control command. An example is shown where the control values ​​for each time period are recorded in matrix form.

[0122] It should be noted that by collecting execution deviation information in real time, the parameter thresholds and optimization weights of the predictive scheduling algorithm are dynamically adjusted. Combined with scenario-based energy and carbon emission time-series data, a multi-objective robust optimization method is introduced to generate a correction scheme, thereby continuously correcting and optimizing the scheduling strategy in actual operation. On the one hand, this improves the flexibility and robustness of the energy optimization scheme, making it feasible under different operating scenarios. On the other hand, through feasibility verification and operational comparison, a closed-loop feedback of energy and carbon control commands is achieved, thus ensuring the dynamic optimization and continuous carbon reduction goals of the smart park under complex operating conditions.

[0123] This embodiment also provides a smart park energy and dual-carbon management system, including: a data acquisition module for collecting real-time operational data and carbon emission monitoring data, and performing unified formatting and noise reduction processing to generate standardized energy data and carbon emission data; a matrix mapping module for inputting standardized energy data and carbon emission data into a pre-built energy carbon emission model, and calculating the correspondence between various energy consumption and carbon emission factors to generate an energy-carbon mapping matrix; a demand analysis module for dynamically analyzing energy consumption patterns and carbon emission changes at different times based on the energy-carbon mapping matrix, obtaining analysis results, identifying peak load intervals and high emission periods, and forming preliminary optimization requirements; and a scheduling generation module for using predictive scheduling algorithms... The system processes initial optimization requirements and generates an energy optimization scheduling scheme through a collaborative operation of rapid edge prediction and global optimization in the cloud. A deviation comparison module distributes the energy optimization scheduling scheme to various energy control links within the park, synchronously collecting and transmitting actual energy and carbon emission data during execution, and then comparing it with the energy-carbon mapping matrix to generate execution deviation information. A closed-loop optimization module adjusts the parameter thresholds and optimization weights of the predictive scheduling algorithm based on the execution deviation information, generating scenario-based energy and carbon emission time-series data. Multi-objective robust optimization is then performed to generate a revised energy optimization scheduling scheme, which is then combined with actual energy and carbon emission data for feasibility verification and operational comparison, generating energy-carbon control instructions.

[0124] This embodiment also provides a computer device applicable to the smart park energy and dual-carbon management method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the smart park energy and dual-carbon management method proposed in the above embodiment.

[0125] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0126] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the smart park energy and dual-carbon management method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0127] In summary, this invention achieves accurate nonlinear and multidimensional fitting between energy consumption behavior and carbon emission factors by inputting standardized energy data and carbon emission data into an energy carbon emission model with a deep neural network architecture, and generating an energy-carbon mapping matrix by combining dynamic weighting and multi-scale time series analysis. This effectively bridges the data gap between the energy side and the carbon emission side, providing a high-precision causal relationship foundation for energy consumption pattern analysis and optimized scheduling, and significantly improving the accuracy and stability of energy-carbon correspondence modeling, enabling scheduling strategies to take into account both energy efficiency and carbon emission reduction targets.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart park energy and dual-carbon management method, characterized by: include, Collect real-time operational data and carbon emission monitoring data, and perform unified formatting and noise reduction processing to generate standardized energy data and carbon emission data; Standardized energy data and carbon emission data are input into a pre-built energy-carbon emission model, and various energy consumption and carbon emission factors are mapped to generate an energy-carbon mapping matrix. The specific steps are as follows: Standardized energy data and carbon emission data are input into the energy carbon emission model to perform preliminary mapping and matching of various energy and carbon emission indicators, generating an original correspondence matrix. The original correspondence matrix is ​​dynamically weighted and multi-dimensional features are fused to adjust the correlation strength between energy categories and carbon emission factors, thereby generating a weighted correlation matrix. By using multi-scale time series analysis and deep learning prediction methods, the weighted correlation matrix is ​​fitted and optimized to eliminate noise and abnormal patterns and generate an accurate correspondence matrix. The precise correspondence matrix is ​​normalized and integrated through multi-dimensional mapping to generate an energy-carbon mapping matrix; The pre-built energy carbon emission model is based on a deep neural network architecture. It receives standardized energy data and carbon emission data, obtains high-dimensional feature representations of energy consumption and carbon emission factors through a multi-dimensional feature extraction layer, introduces an anomaly pattern removal mechanism in the time series modeling and dynamic analysis process, uses a deep learning network to fit the nonlinear mapping relationship between energy consumption and carbon emission factors, and integrates and normalizes the constructed model. Based on the energy-carbon mapping matrix, the energy consumption patterns and carbon emission changes at different times are dynamically analyzed to obtain the analysis results, identify peak load intervals and high emission periods, and form preliminary optimization requirements. Predictive scheduling algorithms are used to process initial optimization requirements, and energy optimization scheduling schemes are generated through a collaborative operation mode of rapid prediction at the edge and global optimization in the cloud. The energy optimization and scheduling scheme is allocated to various energy control links within the park. During the execution process, actual energy and carbon emission data are collected and transmitted back simultaneously, and compared with the energy-carbon mapping matrix to generate execution deviation information. Based on the execution deviation information, the parameter thresholds and optimization weights of the predictive scheduling algorithm are adjusted to generate scenario-based energy and carbon emission time-series data. Through multi-objective robust optimization, a revised energy optimization scheduling scheme is generated. Feasibility verification and operational comparison are then conducted using actual energy and carbon emission data to generate energy and carbon control instructions. The specific steps are as follows: Extract a hierarchical set of deviation indicators from the execution deviation information, and recalibrate the parameter thresholds and optimization weights of the predictive scheduling algorithm to generate an updated parameter set; The updated parameter set is dynamically reorganized and extended over time to generate scenario-based energy and carbon emission time-series data. Multi-objective robust optimization calculations are then performed to obtain the corrected energy optimization scheduling scheme. The revised scheduling scheme is tested for feasibility and compared with actual operation. Combined with actual energy and carbon emission data, energy and carbon control instructions are generated.

2. The smart park energy and dual-carbon management method as described in claim 1, characterized in that: The collected real-time operating data includes electricity, heat, cooling capacity, and gas consumption data; The carbon emission monitoring data includes direct emission data and indirect emission factor data.

3. The smart park energy and dual-carbon management method as described in claim 1, characterized in that: The specific steps for generating standardized energy data and carbon emission data are as follows. The real-time operating data is time-aligned, outlier removed, dimensionless, smoothed, denoised, and filtered to generate standardized energy data. Carbon emission monitoring data is processed by time alignment, outlier removal, missing value completion, and data interpolation to generate carbon emission data.

4. The smart park energy and dual-carbon management method as described in claim 1, characterized in that: The specific steps for forming preliminary optimization requirements are as follows. The energy-carbon mapping matrix is ​​decomposed according to different time scales of day, week and month to generate energy consumption sequence and carbon emission sequence. Dynamic pattern recognition and trend analysis are performed on energy consumption and carbon emission sequences to extract periodic fluctuations, abnormal peaks and potential change patterns, and generate time series analysis results. Based on the time series analysis results, the energy consumption series and carbon emission series are sorted and peak statistics are performed according to the time scale to identify peak load intervals and high emission periods. The energy consumption series and carbon emission series at the corresponding time scale are then sorted to form preliminary optimization requirements.

5. The smart park energy and dual-carbon management method as described in claim 1, characterized in that: The specific steps for generating the energy optimization scheduling scheme are as follows: Based on the initial optimization requirements, a predictive scheduling algorithm is used to make short-term predictions of energy consumption and carbon emission sequences, generating rapid prediction results at the edge. Based on the rapid prediction results from the edge side, the energy consumption and carbon emission sequences for each time period are integrated, and global optimization results are generated through global optimization calculations. The global optimization results are compared and adjusted with the preliminary optimization requirements to generate an energy optimization scheduling scheme.

6. The smart park energy and dual-carbon management method as described in claim 1, characterized in that: The specific steps for generating execution deviation information are as follows: The energy optimization scheduling scheme is transformed into control instructions and distributed to various energy control links, generating feedback energy and emission data. The feedback energy and emission data are time-synchronized, bias-corrected, and filtered to obtain standardized energy and emission data. By comparing standardized energy and emission data with the carbon mapping matrix, the magnitude of the difference between energy and emissions is calculated, and execution deviation information is generated.

7. A smart park energy and dual-carbon management system, based on the smart park energy and dual-carbon management method according to any one of claims 1 to 6, characterized in that: include, The data acquisition module is used to collect real-time operational data and carbon emission monitoring data, and perform unified formatting and noise reduction processing to generate standardized energy data and carbon emission data. The matrix mapping module is used to input standardized energy data and carbon emission data into a pre-built energy carbon emission model, calculate the correspondence between various energy consumption and carbon emission factors, and generate an energy-carbon mapping matrix. The demand analysis module is used to dynamically analyze the energy consumption patterns and carbon emission changes at different times based on the energy-carbon mapping matrix, obtain analysis results, identify peak load intervals and high emission periods, and form preliminary optimized demand. The scheduling generation module is used to process the initial optimization requirements using predictive scheduling algorithms, and generate energy optimization scheduling schemes through a collaborative operation mode of rapid prediction at the edge and global optimization in the cloud. The deviation comparison module is used to allocate energy optimization scheduling schemes to various energy control links within the park. During the execution process, it synchronously collects and transmits actual energy and carbon emission data, and then compares them with the energy-carbon mapping matrix to generate execution deviation information. The closed-loop optimization module is used to adjust the parameter thresholds and optimization weights of the predictive scheduling algorithm based on the execution deviation information, generate scenario-based energy and carbon emission time-series data, perform multi-objective robust optimization, generate a corrected energy optimization scheduling scheme, and perform feasibility verification and operation comparison with actual energy and carbon emission data to generate energy and carbon control instructions.

Images