Building carbon emission monitoring system and method based on big data acquisition

Abstract

The invention discloses a building carbon emission monitoring system and method based on big data acquisition, particularly relates to the field of building carbon emission monitoring, and is used for solving the problems of energy use optimization and carbon emission dynamic regulation and control. Through combination of a recursive feature elimination algorithm and Kalman filter dynamic correction, efficient identification of an electric appliance operation state and accurate correction of load data are realized. Time-delay-free synchronization of distributed energy and load requirements is realized through a time sequence cross-correlation algorithm, and the efficiency and precision of energy allocation and carbon emission monitoring are greatly improved. The dynamic carbon load pressure index and the global energy consumption difference index cooperate to assess changes of load and carbon emission, rapid response to abnormal energy consumption and carbon emission is realized, generation and prediction optimization of a regulation and control strategy are carried out in combination with a fuzzy control algorithm and an autoregression model, the carbon emission of buildings is effectively reduced, and the construction efficiency is improved. And intelligent and dynamic load and carbon emission management is realized.

Description

Technical Field

[0001] The present invention relates to the field of building carbon emission monitoring, and more specifically, to a building carbon emission monitoring system and method based on big data collection. Background Art

[0002] In recent years, the rapid growth of global energy consumption and carbon emissions has exacerbated the greenhouse effect, leading to increasingly serious climate change issues. To cope with this situation, various industries are required to improve energy efficiency and reduce carbon emissions. Therefore, how to effectively monitor the energy consumption and carbon emissions of buildings, especially in a complex energy supply and demand environment, has become one of the key issues in energy conservation and emission reduction.

[0003] Most existing building carbon emission monitoring methods rely on static models, usually based on macro-statistical data or annual reports to estimate carbon emissions, which makes it difficult to accurately reflect the real-time operating status and energy consumption fluctuations of electrical appliances in buildings. This static method is insufficient in monitoring accuracy and dynamic control capabilities when faced with complex energy usage in buildings, and is difficult to cope with changes in electrical loads and the variability of real-time output of distributed energy. In addition, with the popularization of distributed energy sources such as photovoltaics and energy storage, energy usage patterns in buildings have become more complex. Existing technologies make it difficult to achieve source-load coordinated management of energy and carbon emissions in buildings, and are unable to effectively optimize energy use and dynamically adjust carbon emissions.

[0004] Current technical means have significant limitations in the fine identification of electrical loads, energy consumption and carbon emission monitoring. Existing technologies cannot perform real-time and synchronous analysis of the instantaneous energy consumption of electrical appliances and the output of distributed energy. The accuracy and response speed of the carbon emission control system are low, and cannot meet the needs of dynamic control. Therefore, there is an urgent need for an innovative carbon emission monitoring and management system that can accurately identify the operating status of electrical appliances based on real-time data, analyze energy usage patterns, and dynamically optimize the load and carbon emission management inside the building in combination with the real-time output of distributed energy, to help achieve energy conservation and emission reduction goals.

[0005] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiment of the present invention provides a building carbon emission monitoring system and method based on big data collection, which can quickly identify the operating status of different types of electrical appliances by accurately extracting the steady-state and transient characteristics of residential building electrical signals, improve the accuracy of electrical load identification, and effectively capture the energy consumption characteristics of instantaneous changes and stable operation stages. Combined with the recursive feature elimination algorithm and the dynamic correction of the Kalman filter, the system performance optimization and the stability and continuity of the data are ensured. At the same time, through the time series cross-correlation algorithm, the energy output of distributed power generation equipment and the electrical load demand are synchronously mapped without time delay, which greatly improves the energy allocation efficiency, and captures small load fluctuations and energy consumption differences in carbon emission monitoring, thereby improving the accuracy of carbon emission calculation. On the basis of real-time carbon emission calculation and short-term trend prediction, the system can accurately identify abnormal fluctuations in energy consumption and carbon emissions, intervene in regulation in time, and avoid the accumulation of peak energy consumption and excess carbon emissions. Through the collaborative analysis of the dynamic carbon load pressure index and the global energy consumption difference index, it is possible to quickly respond to subtle load fluctuations and make adjustments. Combined with the dynamic control strategy generated by the fuzzy control algorithm, the autoregressive model is used for prediction and load optimization, the control response speed and carbon emission management accuracy are improved, the electrical load in the building is effectively managed, the use of distributed energy is optimized, and the impact of high-energy consumption equipment on overall carbon emissions is reduced, thereby achieving highly integrated and intelligent building energy use and carbon emission management to solve the problems raised in the above background technology.

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

[0008] S1, using low-frequency sampling equipment to obtain electrical signals in residential buildings, classified according to steady-state and transient characteristics, transient characteristics extracted by adaptive wavelet analysis, steady-state characteristics based on Fourier transform processing, respectively record the instantaneous start-up and stable operation characteristics of electrical appliances; the extracted steady-state and transient characteristic data are stored in a local database;

[0009] S2, through the recursive feature elimination algorithm, the classified and extracted electrical appliance steady-state and transient features are screened, the key features with high contribution to electrical appliance identification are retained, and the Kalman filter is used to dynamically correct the key feature values, filter out signal fluctuations and noise, and ensure that the key feature data is consistent and accurate at different times;

[0010] S3, based on the corrected key feature data, obtains the real-time output data of distributed power generation equipment, aligns the energy output of distributed energy with the load demand of electrical appliances through the time series cross-correlation algorithm, and ensures the time synchronization and time-delay-free mapping of source-load data;

[0011] S4, based on the corrected source-load synchronization data, calculate the carbon emissions of each appliance, conduct a comprehensive assessment of energy consumption and carbon emissions by evaluating the instantaneous impact of appliance energy consumption deviation and load growth, and predict the carbon emission trend through an autoregressive model when the assessment result deviates from expectations, and use the assessment and prediction results as the basis for load optimization and regulation;

[0012] S5, based on the carbon emission calculation results and trend forecast data, generates a dynamic load control strategy based on the fuzzy control algorithm and automatically executes it through the load adaptive control module.

[0013] In a preferred embodiment, step S1 includes the following contents:

[0014] S1.1, use low-frequency sampling equipment to continuously sample the electrical signals of residential buildings to capture the steady-state and transient characteristics of electrical appliance operation; the collected data is recorded with a timestamp and synchronously written to the local cache;

[0015] S1.2, the transient characteristics of electrical signals refer to the signal fluctuations during the instantaneous change of electrical start-up and shutdown; the transient characteristics are extracted by adaptive wavelet transform, and the specific steps are as follows:

[0016] Use the selected wavelet basis function; decompose the electrical signal in the frequency domain through multi-scale wavelet decomposition to capture the transient characteristics of the electrical appliance during start-up and shutdown; after the extraction is completed, the transient characteristic data is marked according to the event triggering time and then stored in the local database;

[0017] S1.3, steady-state characteristics refer to the characteristic parameters of the electrical appliance when it is running stably; the steady-state characteristics are extracted through Fourier transform, and the specific steps are as follows:

[0018] First, the steady-state signal is windowed; the electrical signal is converted from the time domain to the frequency domain using the fast Fourier transform, and the spectral characteristics of the signal are analyzed; the steady-state characteristic data is stored in time sequence and associated with the corresponding electrical load state;

[0019] S1.4, using an event-driven trigger mechanism, first identifies sudden changes in electrical signals through a mutation detection algorithm; if the instantaneous change of the signal exceeds a preset threshold, the corresponding segment signal is identified as a transient feature; otherwise, the corresponding segment signal is identified as a steady-state feature by default;

[0020] S1.5, based on the classification results, the steady-state and transient characteristics are stored in different data sets, with timestamps and device IDs attached for quick indexing.

[0021] In a preferred embodiment, step S2 includes the following contents:

[0022] S2.1, based on the steady-state and transient characteristics of the electrical appliance extracted in step S1, first normalize the extracted characteristics;

[0023] S2.2, in feature screening, a recursive feature elimination algorithm is used to progressively screen the normalized steady-state and transient feature data; the output is a screened key feature set, including steady-state features and transient features;

[0024] S2.3, dynamically correct the selected key feature data, and use Kalman filter to filter noise and smooth the signal; the output is the dynamically corrected key feature data, including the correction values ​​of steady-state and transient features;

[0025] S2.4, the key feature data after dynamic correction is recorded and stored in the local database according to the timestamp; different storage strategies are set for transient features and steady-state features: transient feature data is marked and stored according to event triggers, and steady-state features are stored according to fixed time windows.

[0026] In a preferred embodiment, step S3 includes the following contents:

[0027] S3.1, obtaining real-time output data of distributed power generation equipment, including output power, real-time voltage and current parameters of distributed energy equipment;

[0028] S3.2, in order to ensure that the output data of the distributed generation equipment matches the electrical load characteristic data in time, the timestamps of the two need to be preprocessed and synchronized:

[0029] Standardize the timestamps of the real-time output data of distributed power generation equipment and the corrected load characteristic data to ensure that all data are recorded according to the same time reference; if the sampling frequencies of distributed energy data and load characteristic data are different, use interpolation or downsampling technology to reduce the frequency of the higher frequency data to align it with the lower frequency data to ensure that the data of the two are compared and analyzed within the same time window;

[0030] S3.3, in order to achieve the time-delay-free alignment of the output data of distributed power generation equipment and the electrical load demand, the time series cross-correlation algorithm is used to associate and map the data. The specific steps are as follows:

[0031] A sliding time window is set on the load demand data and distributed energy output data, and the time relationship between load demand and distributed energy output is analyzed by calculating the cross-correlation function between the two time series; the cross-correlation function is defined as follows: R xy (τ)=∑ t x(t)·y(t-τ);

[0032] Among them, x(t) is the load demand data, y(t) is the output data of distributed generation, and τ is the time lag. By finding the peak position of the cross-correlation function, the optimal alignment time point between distributed energy output and load demand is determined to ensure a time-lag-free mapping between the two.

[0033] According to the calculation results of the cross-correlation function, the time alignment of the distributed generation equipment output data and the load demand data is adjusted; if a delay τ is detected, it is corrected by data shifting or interpolation technology to ensure that the source-load data can be matched without lag;

[0034] S3.4, after completing the time series alignment, establish the mapping relationship between the energy output of distributed energy and the electrical load demand; the specific method is:

[0035] Based on the time-aligned data, a functional relationship between the output power of distributed energy and the electrical load demand is established, and a linear regression or multivariate regression model is used to calculate the power matching degree between the two.

[0036] By calculating the total output power of distributed energy and the load demand of electrical appliances in each time window, analyze whether the two match; if the energy output is less than the load demand, record the energy gap information; if the energy output is greater than the load demand, analyze the energy storage or feedback path;

[0037] S3.5, after completing time synchronization and data mapping, the aligned and corrected distributed energy output data and electrical load demand data are stored in the local database.

[0038] In a preferred embodiment, step S4 includes the following contents:

[0039] S4.1, based on the real-time power generation structure of the power grid and the types and actual operation of distributed power generation equipment, obtain the dynamic carbon emission factors of the power grid and distributed energy to reflect the changes in the carbon emission intensity of power sources;

[0040] S4.2, calculate the real-time carbon emissions of each appliance using the following formula:

[0041] Among them, C(t) is the carbon emission of the appliance at time t, P(t) is the power consumption of the appliance, and P grid and P total are the power supply of the grid and distributed energy, respectively, grid (t) and f dist (t) Dynamic carbon emission factors for the power grid and distributed energy, respectively;

[0042] S4.3, conduct an assessment of the deviation of electrical appliance energy consumption and calculate the global energy consumption difference index, which is used to quantify the difference between the current energy consumption of electrical appliances and the historical benchmark energy consumption. The specific acquisition process is as follows:

[0043] Based on the corrected source-load synchronization data, the energy consumption characteristics of the electrical appliances at the current time t are extracted to form an energy consumption characteristic vector Among them, e i (t) represents the instantaneous power consumption of the i-th appliance at time t, and N is the number of appliances;

[0044] Select the energy consumption feature vectors of the past M time windows from the historical data and construct the historical energy consumption feature space matrix H: H = [E(t-1), E(t-2), ..., E(tM)]; where E(tk) is the energy consumption feature vector at time tk;

[0045] Perform singular value decomposition on the historical energy consumption feature space matrix and obtain: Among them, U and V are orthogonal matrices, ∑ is a diagonal singular value matrix;

[0046] Select the singular vectors corresponding to the first K largest singular values ​​to construct the energy consumption pattern basis matrix B: B = [u1,u2,…,u K ]; where u K is the kth column of U, K satisfies K <N;

[0047] Project the current energy consumption feature vector onto the energy consumption pattern basis matrix to obtain the projection coefficient vector

[0048] Using the projection coefficient vector and the energy consumption pattern basis matrix, the energy consumption feature vector E′(t) is reconstructed: E′(t)=Ba(t);

[0049] The difference between the actual energy consumption feature vector and the reconstructed energy consumption feature vector is obtained to obtain the reconstruction error vector δ(t): δ(t) = E(t) - E′(t);

[0050] Perform nonlinear transformation on each element of the reconstructed error vector to obtain the difference metric function value Among them, δ i (t) is the i-th element of the reconstruction error vector, λ i is the scale parameter of the i-th appliance, controlling the sensitivity of the difference measure;

[0051] Normalize the difference metric function value to get the global energy consumption difference index Where D maxIt is the maximum value of the difference measurement function and is used for normalization processing so that the value range of the global energy consumption difference index is between [0,1].

[0052] In a preferred embodiment, step S4 further includes the following contents:

[0053] S4.4, the instantaneous impact of load growth is evaluated by calculating the rate of impact of load change on carbon emissions, and the dynamic carbon load pressure index is calculated. The specific acquisition process is:

[0054] According to the carbon emission C(t) obtained in step S4.2, its first-order time derivative V(t) is calculated to represent the instantaneous rate of change of carbon emission over time: Among them, V(t) represents the instantaneous rate of change of carbon emissions, reflecting the dynamic trend of carbon emissions;

[0055] Define the nonlinear response function R(t) of carbon emission changes to amplify the changes in carbon emission rate and capture sudden carbon emission pressure: R(t) = κ·ln(1+[V(t)] 2 ), where k is the proportional coefficient used to adjust the sensitivity of the function, ln is the natural logarithmic function, [V(t)] 2 represents the square of the carbon emission rate, ensuring that the response function responds to both positive and negative changes;

[0056] The space-time convolution operator is introduced to perform convolution operation on the nonlinear response function of carbon emissions to obtain the cumulative carbon emission pressure function. Among them, φ(t-τ) is the kernel function, and the Gaussian kernel function is selected: σ is the scale parameter of the kernel function, which controls the sensitivity of the convolution operation to the time difference;

[0057] The cumulative carbon emission pressure function is transformed nonlinearly to define the dynamic carbon load pressure index. Among them, λ is the adjustment parameter, which controls the steepness of the exponential function, and μ is the translation parameter, which sets the midpoint position of the exponential function;

[0058] S4.5, comprehensively evaluate and calculate the global energy consumption difference index and the dynamic carbon load pressure index to obtain the energy-carbon synchronization regulation index;

[0059] If the energy-carbon synchronization control index is greater than or equal to the control threshold, it means that the energy consumption and carbon emission pressure of the appliance are both at a high level, and there is an obvious abnormality. The control decision needs to be executed immediately, and a control signal is generated;

[0060] S4.6: If the assessment results show that regulation is needed, the autoregressive model is used to predict the carbon emission trend. The specific steps are as follows:

[0061] Based on historical carbon emission data t-1 ,C t-2 ,…,C t-n}, build an autoregressive model to predict carbon emissions in the future: in, is the prediction step size, α1,α2,…,α n is the model coefficient, ∈(t) is the random error term.

[0062] In a preferred embodiment, step S5 includes the following contents:

[0063] S5.1, if the assessment results show that regulation is needed, fuzzification processing is performed on the carbon emission calculation results and trend forecast data to convert continuous input variables into fuzzy linguistic variables;

[0064] S5.2, construct a fuzzy control rule base to map different combinations of carbon emissions and change trends to control strategies;

[0065] S5.3, using the fuzzy inference engine, matching the input fuzzy variables with the rule base; based on the matching situation in the rule base, using the Max-Min inference method or other appropriate inference methods to generate the membership of different control strategies;

[0066] S5.4, defuzzify the fuzzy reasoning results, convert the fuzzy control strategy into a specific load adjustment operation; use the center of gravity method to calculate the exact value of the control strategy;

[0067] S5.5, passing the defuzzified specific control strategy to the load adaptive control module;

[0068] S5.6, after the load control is executed, the load adaptive control module will continuously monitor the operating status and carbon emission changes of the electrical appliances, and compare the actual feedback data with the control target; if the control result does not meet the expectation, the control strategy will be automatically adjusted.

[0069] A building carbon emission monitoring system based on big data collection, comprising: a signal acquisition unit, a feature screening and correction unit, an energy alignment unit, a carbon emission calculation and evaluation unit and a dynamic control module unit;

[0070] Signal acquisition unit: Use low-frequency sampling equipment to obtain electrical signals in residential buildings, and classify and process the signals according to steady-state and transient characteristics; the steady-state characteristics are extracted through Fourier transform, and the transient characteristics are extracted through adaptive wavelet analysis, and the instantaneous start-up and stable operation characteristics of the electrical appliances are recorded respectively; the extracted steady-state and transient characteristic data are transmitted to the feature screening unit;

[0071] Feature screening and correction unit: The classified and extracted electrical steady-state and transient features are screened through a recursive feature elimination algorithm, and the key features that contribute most to electrical identification are retained. The screened key feature values ​​are dynamically corrected through a Kalman filter to filter out signal fluctuations and noise to ensure the consistency of feature data. The corrected key feature data is transmitted to the energy alignment unit.

[0072] Energy alignment unit: Based on the corrected key feature data, the real-time output data of distributed power generation equipment is obtained, and the time series cross-correlation algorithm is used to align the energy output of distributed energy with the electrical load demand without time lag, ensuring data linkage and synchronization between source and load; the aligned source-load synchronization data is transmitted to the carbon emission calculation and evaluation unit;

[0073] Carbon emission calculation and evaluation unit: Based on the aligned source-load synchronization data, the carbon emission of each appliance is calculated, and the instantaneous impact of the energy consumption deviation and load growth of the appliance is evaluated to form a comprehensive evaluation result of energy consumption and carbon emissions; when the evaluation result shows an abnormality, the autoregressive model is used to predict the carbon emission trend; the evaluation result and trend prediction data are transmitted to the dynamic control unit;

[0074] Dynamic control module unit: Based on the carbon emission calculation results and trend forecast data, a dynamic load control strategy is generated through a fuzzy control algorithm to automatically adjust the operating status of electrical appliances, optimize the proportion of distributed energy use, and finally automatically execute it through the load adaptive control module.

[0075] Technical effects and advantages of a building carbon emission monitoring system and method based on big data collection in the present invention:

[0076] 1. The present invention can quickly identify the operating status of different types of electrical appliances in a complex load environment by accurately classifying and extracting the steady-state and transient characteristics of electrical signals in residential buildings, greatly improving the recognition accuracy of electrical loads, and effectively capturing the energy consumption characteristics of instantaneous changes and stable operation stages. The application of the recursive feature elimination algorithm ensures that only the key features that contribute to electrical appliance identification are retained, avoiding the impact of redundant data on system performance. The dynamic correction of the Kalman filter effectively eliminates noise interference and ensures the stability and continuity of the data. Through the time series cross-correlation algorithm, the energy output of distributed power generation equipment and the electrical load demand are synchronized without time delay, so that the data between source and load are accurately mapped. This real-time linkage mechanism not only improves the efficiency of energy allocation, but also can capture tiny load fluctuations and energy consumption differences during carbon emission monitoring, thereby significantly improving the accuracy of carbon emission calculations, providing reliable data support for subsequent dynamic load regulation and carbon emission management, and realizing full process optimization from data collection to carbon emission assessment.

[0077] 2. The present invention can accurately identify abnormal fluctuations in energy consumption and carbon emissions in buildings by combining real-time calculation of carbon emissions with short-term trend prediction, ensuring timely intervention and regulation when load increases or carbon emission pressure increases, thereby avoiding the accumulation of peak energy consumption and excess carbon emissions. The collaborative analysis of the dynamic carbon load pressure index and the global energy consumption difference index effectively improves the sensitivity to load changes, ensuring that even subtle load fluctuations can be responded to and adjusted in a timely manner. The dynamic control strategy generated by the fuzzy control algorithm not only makes adjustments based on real-time data, but also optimizes the load in advance through the prediction of the autoregressive model, significantly improving the response speed of building load control and the accuracy of carbon emission management. It can then intelligently manage the electrical load in the building, optimize the proportion of distributed energy use, and reduce the impact of high-energy-consuming equipment on overall carbon emissions, thereby effectively controlling and reducing the overall carbon emission level of the building, and achieving a high degree of integration and intelligence in building energy use and carbon emission management. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A schematic diagram of a process of a building carbon emission monitoring method based on big data collection according to the present invention;

[0079] Figure 2 It is a schematic diagram of the process of a building carbon emission monitoring system based on big data collection of the present invention. DETAILED DESCRIPTION

[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] Example 1

[0082] Figure 1 The present invention provides a building carbon emission monitoring method based on big data collection, comprising:

[0083] S1, use low-frequency sampling equipment to obtain electrical signals in residential buildings, classify them according to steady-state and transient characteristics, extract transient characteristics through adaptive wavelet analysis, and process steady-state characteristics based on Fourier transform to record the instantaneous start-up and stable operation characteristics of electrical appliances respectively; the extracted steady-state and transient characteristic data are stored in the local database.

[0084] S2, through the recursive feature elimination algorithm, the classified and extracted electrical steady-state and transient features are screened, the key features that contribute most to electrical appliance identification are retained, and the Kalman filter is used to dynamically correct the key feature values, filter out signal fluctuations and noise, and ensure that the key feature data is consistent and accurate at different times.

[0085] S3, based on the corrected key feature data, obtains the real-time output data of distributed power generation equipment, aligns the energy output of distributed energy with the load demand of electrical appliances through the time series cross-correlation algorithm, and ensures the time synchronization and time-delay-free mapping of source-load data.

[0086] S4, based on the corrected source-load synchronization data, calculates the carbon emissions of each appliance, and conducts a comprehensive assessment of energy consumption and carbon emissions by evaluating the instantaneous impact of appliance energy consumption deviation and load growth. When the assessment results deviate from expectations, the carbon emission trend is predicted through an autoregressive model, and the assessment and prediction results are used as the basis for load optimization and regulation.

[0087] S5, based on the carbon emission calculation results and trend forecast data, generates a dynamic load control strategy based on the fuzzy control algorithm and automatically executes it through the load adaptive control module.

[0088] In residential buildings, the operating status of electrical appliances not only directly determines the overall energy consumption, but also has a profound impact on the accurate measurement of building carbon emissions. There are significant differences in the electrical signal performance of electrical appliances in different states, especially when the appliances are started and shut down, there will be instantaneous signal fluctuations (transient characteristics), while when the appliances are running continuously for a long time, they will show stable electrical characteristics (steady-state characteristics). In order to accurately identify the load status, energy consumption characteristics and corresponding carbon emissions of electrical appliances, it is a key step for the system to refine the electrical signals and distinguish between steady-state and transient characteristics.

[0089] Through real-time acquisition of electrical signals and combined with advanced signal processing algorithms (such as adaptive wavelet analysis and Fourier transform), steady-state and transient features can be effectively extracted and classified. This not only helps to accurately evaluate the instantaneous energy consumption of electrical appliances, but also captures subtle changes during the operation of electrical appliances, improving the accuracy and timeliness of overall identification. Step S1 will accurately extract steady-state and transient features to provide a solid data foundation for subsequent electrical appliance identification, energy consumption calculation, and carbon emission analysis, ensuring that the system can achieve dynamic regulation of load and carbon emissions.

[0090] Step S1 includes the following contents:

[0091] S1.1, use low-frequency sampling equipment to continuously sample the electrical signals of residential buildings, and the sampling frequency is set between 50Hz and 500Hz to ensure that the steady-state and transient characteristics of electrical operation are captured. The collected electrical signals include voltage, current, active power, reactive power, power factor, and harmonic components. The collected data is recorded with precise timestamps and synchronously written to the local cache to ensure the timing consistency and data integrity of the signal.

[0092] S1.2, the transient characteristics of electrical signals refer to the signal fluctuations during the instantaneous changes of electrical appliances such as starting and shutting down. The transient characteristics are extracted by adaptive wavelet transform. The specific steps are as follows:

[0093] Select a wavelet basis function suitable for capturing instantaneous changes, such as Daubechies wavelet or Morlet wavelet. Through multi-scale wavelet decomposition, the electrical signal is decomposed in the frequency domain to capture instantaneous features such as voltage drops and current spikes during the start-stop process of the electrical appliance.

[0094] During the signal decomposition process, according to the characteristics of transient signals, the adaptive wavelet packet decomposition method is used to automatically adjust the scale of the wavelet basis function to ensure that the high-frequency components can be accurately extracted.

[0095] The extracted transient features include voltage sag amplitude, current transient change rate, impact power, etc. These features reflect the dynamic characteristics of the electrical startup and shutdown process.

[0096] After the extraction is completed, the transient characteristic data is marked according to the event triggering time, and then stored in the local database, providing data support for subsequent appliance identification.

[0097] S1.3, steady-state characteristics refer to the characteristic parameters of the electrical appliance when it is running stably, including the stable values ​​of voltage, current, power, etc. The steady-state characteristics are extracted through Fourier transform. The specific steps are as follows:

[0098] First, the steady-state signal is windowed to ensure that the signal remains stable within the selected time period.

[0099] Fast Fourier Transform (FFT) is used to convert electrical signals from the time domain to the frequency domain and analyze the spectral characteristics of the signal, especially the amplitude and phase of each frequency component.

[0100] The extracted steady-state features include power factor, harmonic distortion rate, steady-state voltage and current amplitude, harmonic amplitude, etc. These features are used to identify the power characteristics and power quality of electrical appliances during continuous operation.

[0101] Steady-state characteristic data are stored in chronological order and associated with the corresponding electrical load status to ensure the accuracy of subsequent identification and calculation.

[0102] S1.4, using an event-driven trigger mechanism, first identify sudden changes in electrical signals through a mutation detection algorithm. If the instantaneous change of the signal (such as a sudden change in voltage or current) exceeds a preset threshold, the corresponding segment signal is identified as a transient feature. Otherwise, the corresponding segment signal will be identified as a steady-state feature by default.

[0103] Transient feature recognition depends on the severity of the instantaneous signal change. The specific steps are as follows:

[0104] By detecting the rate of change of current and voltage (ΔI / Δt or ΔV / Δt), it is determined whether the instantaneous change amplitude of the signal exceeds the set burst threshold. If the change amplitude exceeds the threshold, it is considered that the appliance is undergoing a startup or shutdown process, triggering transient feature classification.

[0105] For the specific time period when the transient event occurs, the timestamp is automatically marked, the period is marked as the transient characteristic section, and the relevant data is stored in the transient characteristic data set.

[0106] Steady-state feature recognition mainly targets signal data under the stable operation state of the electrical appliance. The specific steps are as follows:

[0107] During the operation of the appliance, the mean and variance of the signal changes are detected. If the amplitude of the signal change (such as current or voltage) remains within a stable range and the fluctuation is small, it is identified as a steady-state feature.

[0108] By analyzing the signal mean and fluctuation within the time window, the period with smaller fluctuation is marked as a steady-state feature segment, and the feature data of this period is classified and stored in the steady-state feature data set.

[0109] When the transient event ends, it automatically switches to the steady-state detection mode. The specific steps are as follows:

[0110] When the transient characteristics of the appliance (such as startup or shutdown) end and the rate of change of the monitoring signal gradually stabilizes, the transient detection algorithm stops and automatically switches to the steady-state detection algorithm. This conversion is judged by the threshold to ensure that the transient-steady-state switching is correct.

[0111] In the process of feature classification, in order to ensure accuracy, the transient and steady-state classification results will be cross-validated, that is, by calculating the rate of change of the signal and comparing it with the characteristic value to ensure that the classification result matches the actual state of the appliance.

[0112] S1.5, based on the classification results, the steady-state and transient characteristics are stored in different data sets, with timestamps and device IDs attached for quick indexing. The characteristic data is stored in an independent index form to ensure quick query and accurate identification of the subsequent operation status of the appliance.

[0113] In residential buildings, the steady-state and transient characteristics of electrical appliances are crucial for appliance identification and energy consumption analysis. However, due to the noise interference and environmental fluctuations in electrical signals, directly using the original feature data for analysis may lead to inaccurate results. In order to ensure that the extracted appliance feature data has a high recognition contribution and maintains consistency and accuracy at different times, the feature data needs to be screened and dynamically corrected. Step S2 uses a recursive feature elimination algorithm to screen out key features with high contribution to appliance identification, and uses a Kalman filter to dynamically correct the feature data to filter out noise and signal fluctuations to ensure the stability of the appliance feature data. This process lays a solid data foundation for subsequent load analysis and carbon emission calculations.

[0114] Step S2 includes the following contents:

[0115] S2.1, based on the steady-state and transient characteristics of the electrical appliances extracted in step S1, the extracted characteristics are first normalized to ensure that the numerical ranges of the characteristics of different dimensions are unified and to eliminate the influence of the characteristic values ​​on the recognition effect due to different magnitudes. The normalization method uses Z-score standardization, setting the mean of all characteristics to zero and the standard deviation to one to ensure the fairness of subsequent screening.

[0116] S2.2, in feature screening, the recursive feature elimination algorithm (RFE) is used to progressively screen the normalized steady-state and transient feature data. The specific technical steps of the algorithm include:

[0117] The input data is the steady-state and transient feature sets of the electrical equipment extracted in step S1. These feature data have been normalized to ensure that the numerical ranges of features of different dimensions are unified. The input data set includes multi-dimensional electrical features, including steady-state features (such as power factor, harmonic distortion rate, steady-state voltage and current, etc.) and transient features (such as voltage mutation amplitude, current spike, transient power fluctuation, etc.).

[0118] Initialize the main parameters of the Recursive Feature Elimination (RFE) algorithm:

[0119] Evaluation model: Support vector machine (SVM) was selected as the classifier for feature weight evaluation in the feature elimination process, and the linear kernel function (LinearKernel) was used for initial setting.

[0120] Number of features: Based on the dimension of the input data, all features are initially retained, and the recursive goal is to gradually reduce the feature dimension until the most discriminative features are retained.

[0121] Weight calculation: During the training process, the SVM classifier generates a feature weight vector. The weight of each feature reflects its contribution to appliance identification.

[0122] In the recursive process, each round will delete the feature with the smallest contribution according to the order of feature weight values, and repeat this process until the number of remaining features meets the preset conditions. The specific process is as follows:

[0123] Step 1: Training the model

[0124] Use all input features to train the support vector machine (SVM) model, classify the electrical operation status, and calculate the weight coefficient corresponding to each feature. The weight value reflects the importance of the feature for identifying the electrical load status. The formula is expressed as: Among them, w o Represents feature x o The weight of , L is the loss function of the SVM classifier.

[0125] Step 2: Feature sorting and elimination

[0126] Sort the features by weight, placing the features with lower weights at the end. Delete the features with the lowest weights by a certain ratio (such as 10%) or a preset number. For example, if there are 50 features initially, delete the 5 features with the lowest weights in the first round, leaving 45 features.

[0127] Step 3: Retraining

[0128] After removing the feature with the lowest weight, the SVM model is retrained using the retained features, and the above training and feature elimination steps are repeated. Each cycle will reduce a number of features and retain features with higher weights, thereby gradually optimizing the feature set for appliance identification.

[0129] During the recursive feature elimination process, the feature deletion ratio and recursive rounds can be dynamically adjusted according to the appliance recognition accuracy:

[0130] Stop condition for recursive rounds: preset a lower limit for the number of features or a threshold for reaching the required classification accuracy. You can set the recursion to stop when the classification accuracy is no longer significantly improved (for example, the accuracy improvement is less than 0.01).

[0131] Number or ratio of features to be deleted: The number of features to be deleted can be set for each recursion, usually 5% to 10% of the original number of features. The deletion ratio is dynamically adjusted according to the complexity of the specific appliance category.

[0132] Recursive termination condition: The recursive feature elimination algorithm will be repeated until the number of remaining features reaches the preset target or the performance indicators of the classifier (such as accuracy, recall, etc.) are no longer significantly improved. The final retained feature set is the key feature set with the highest contribution to appliance identification.

[0133] Output results: The final output is a selected key feature set, including steady-state features and transient features. The specific number is determined by adjusting the recursive rounds and deletion ratio. The output feature set is used as the input for electrical appliance identification, for further electrical load classification and carbon emission calculation in subsequent steps.

[0134] The recursive feature elimination algorithm ensures that the final retained features contribute the most to appliance identification by cyclically training the SVM model, calculating feature weights, gradually eliminating features with low weights, and continuously optimizing the process of retaining features, thereby ensuring that the system has high accuracy and efficiency when processing appliance load status identification.

[0135] S2.3, dynamically correct the selected key feature data, and use the Kalman filter to perform noise filtering and signal smoothing. The specific steps are as follows:

[0136] The input data is the key feature data filtered by the recursive feature elimination algorithm, including steady-state and transient features. The input data is in the form of a time series, including the feature values ​​and their changes at each time point, such as sudden changes in voltage and current, fluctuations in power factor, etc. These data may be interfered by noise and power environment and need to be dynamically corrected.

[0137] Initialization of Kalman filter model: The correction process of Kalman filter is based on the design of state space model. The specific initialization steps are as follows:

[0138] State variable definition: Set the system's state variable x t , represents the real characteristic value at the current time point t. This state variable describes the characteristic value of the appliance in a steady state or transient state, reflecting its operating state.

[0139] Measurement variable definition: Define the measurement variable z t , which is the actual observed characteristic value, which may be affected by noise. The measured variable represents the observed data obtained by the electrical signal acquisition device.

[0140] State transfer matrix A: Define the state transfer matrix A to describe the process of eigenvalues ​​changing over time. Assuming that the eigenvalues ​​change relatively smoothly in continuous time, the state transfer matrix is ​​usually set to the identity matrix A = I.

[0141] Observation matrix H: Observation matrix H is defined to map state variables to measurement space. In this scenario, it is assumed that the eigenvalues ​​are measured directly, and the observation matrix is ​​set to the identity matrix H = I.

[0142] Process noise covariance matrix Q: describes the process noise of the state variables. The process noise is assumed to be Gaussian distributed, and the covariance matrix Q is adjusted according to the stability of the appliance characteristics. The process noise of steady-state characteristics is small, while the process noise of transient characteristics may be large.

[0143] Measurement noise covariance matrix R: describes the measurement noise and reflects the noise level of the electrical signal acquisition equipment. The covariance matrix R is adjusted according to the noise characteristics of the actual electrical equipment.

[0144] Prediction step: Based on the state space model of the Kalman filter, use the prior information to predict the state variables at the next moment (i.e. the eigenvalue at the next moment):

[0145] Among them, u t It is the control input, which describes the impact of external interference on the characteristics of the appliance (such as voltage fluctuations and load changes). It is assumed that B=0 when there is no control input.

[0146] At the same time, the covariance matrix of the predicted state is: P t|t-1 =A·P t-1 ·A T +Q;

[0147] Among them, P t|t-1 It represents the covariance of the predicted state and reflects the uncertainty of the feature value.

[0148] Update step: Every time new feature data z is observed t Then, calculate the Kalman gain K t , used to update the state estimate: K t =P t|t-1 ·H T ·(H·P t|t-1 ·H T +R) -1 ;

[0149] Kalman gain = measures the contribution of the measurement value to the feature estimation. When the measurement noise is small R, the Kalman gain is large and the measurement value has a greater impact on the estimation.

[0150] Use the Kalman gain to update the state:

[0151] Among them, z t is the observed value at the current moment, is the updated feature estimate. The updated estimate is a weighted average of the prior prediction and the observed value, and the Kalman gain determines the weight of the two.

[0152] At the same time, update the covariance matrix: P t =(IKt ·H)·P t|t-1 ;

[0153] Updated covariance matrix P t Reflects the uncertainty of the updated feature estimate.

[0154] Loop execution and feature data correction: The Kalman filter runs in a loop, repeatedly performing the prediction and update steps. At each time point, the state variables are corrected using new observations to ensure that the key feature values ​​can dynamically reflect the operating status of the appliance, filter out signal noise, and ensure data stability and continuity.

[0155] Especially during transient changes of electrical appliances, such as the instantaneous characteristics of startup and shutdown, the Kalman filter can quickly adapt to the drastic fluctuations of characteristic values ​​and provide smooth estimation results.

[0156] Output and corrected data: The final output is the key characteristic data after dynamic correction, including the correction values ​​of steady-state and transient characteristics. Each characteristic data is accompanied by a timestamp to ensure the consistency and accuracy of the characteristic data at different times. These corrected characteristic data will be used for appliance identification and subsequent carbon emission calculations to ensure high-precision load analysis and regulation.

[0157] The dynamic correction process of the Kalman filter achieves dynamic adjustment of the electrical characteristic data through a prediction-update cycle. This process can filter out noise interference and provide real-time correction results for the electrical state, which is particularly suitable for accurate estimation of steady-state and transient characteristics in complex electrical environments.

[0158] S2.4, the key feature data after dynamic correction is recorded and stored in the local database by timestamp. Different storage strategies are set for transient features and steady-state features: transient feature data is marked and stored according to event triggers, and steady-state features are stored according to fixed time windows. The system will regularly update and verify the stored feature data to ensure efficient call and data consistency in subsequent analysis.

[0159] In the above steps, the corrected key characteristic data has been subjected to noise filtering and timing correction by the Kalman filter, ensuring the accuracy and consistency of the electrical load characteristic data. In order to further enhance the system's dynamic perception of the matching of distributed energy and electrical load, it is necessary to achieve the time-delay-free alignment of the real-time output data of distributed power generation equipment and the electrical load demand. Distributed energy (such as photovoltaics, energy storage, etc.) plays an important role in building electricity consumption, and its energy output directly affects the load demand balance of the building. By synchronizing and accurately mapping the energy output of distributed energy with the electrical load characteristics, it can be ensured that the dynamic energy regulation between the source and the load has high precision and real-time performance, providing data support for subsequent carbon emission calculations and load optimization. Therefore, the core of step S3 is to ensure that the energy output and demand between distributed generation and electrical loads can be seamlessly connected in time through a time series cross-correlation algorithm, so as to achieve time-delay-free mapping and dynamic linkage of data.

[0160] Step S3 includes the following contents:

[0161] S3.1, obtain the real-time output data of distributed generation equipment, including the output power, real-time voltage and current of distributed energy equipment (such as photovoltaic power generation or energy storage system). The data is collected at a high frequency through the monitoring system of distributed generation equipment, and is accompanied by a precise timestamp to ensure time alignment with load demand.

[0162] S3.2, in order to ensure that the output data of the distributed generation equipment and the electrical load characteristic data match accurately in time, the timestamps of the two need to be preprocessed and synchronized:

[0163] First, the real-time output data of distributed generation equipment and the timestamps of the corrected load characteristic data are standardized to ensure that all data are recorded according to the same time reference. The Network Time Protocol (NTP) is used for time synchronization to ensure that the time recording accuracy of all devices reaches the millisecond level.

[0164] If the sampling frequencies of distributed energy data and load characteristic data are different, the higher frequency data needs to be downsampled through interpolation or downsampling technology to align it with the lower frequency data to ensure that the data of both are compared and analyzed within the same time window.

[0165] S3.3, in order to achieve the time-delay-free alignment of the output data of distributed power generation equipment and the electrical load demand, the time series cross-correlation algorithm is used to associate and map the data. The specific steps are as follows:

[0166] A sliding time window is set on the load demand data and distributed energy output data. The window size is dynamically adjusted according to the load fluctuation frequency and the response speed of energy output. It is usually set in the range of several seconds to tens of seconds to ensure that short-term fluctuations in source-load interaction are captured.

[0167] By calculating the cross-correlation function between two time series, the temporal relationship between load demand and distributed energy output is analyzed. The cross-correlation function is defined as follows: R xy (τ)=∑ t x(t)·y(t-τ);

[0168] Among them, x(t) is the load demand data, y(t) is the output data of distributed generation, and τ is the time lag. By finding the peak position of the cross-correlation function, the optimal alignment time point between distributed energy output and load demand can be determined to ensure a time-lag-free mapping between the two.

[0169] According to the calculation results of the cross-correlation function, the time alignment of the distributed generation equipment output data and the load demand data is adjusted. If a delay τ is detected, it is corrected by data shifting or interpolation technology to ensure that the source-load data can be matched without lag.

[0170] S3.4, after completing the time series alignment, establish the mapping relationship between the energy output of distributed energy and the load demand of electrical appliances. The specific method is:

[0171] Based on the time-aligned data, a functional relationship between the output power of distributed energy and the electrical load demand is established, and a linear regression or multivariate regression model is used to calculate the power matching degree between the two.

[0172] By calculating the total output power of distributed energy and the load demand of electrical appliances in each time window, we analyze whether the two match. If the energy output is less than the load demand, we record the energy gap information; if the energy output is greater than the load demand, we analyze the possible energy storage or feedback path.

[0173] S3.5, after completing time synchronization and data mapping, the aligned and corrected distributed energy output data and electrical load demand data are stored in the local database. The source-load data is classified and stored by timestamp, and provides basic data for subsequent carbon emission calculation and load regulation. The data storage process adopts a distributed storage architecture to ensure data security and real-time performance, and supports fast query and retrieval.

[0174] The present invention can quickly identify the operating status of different types of electrical appliances in a complex load environment by accurately classifying and extracting the steady-state and transient characteristics of electrical signals in residential buildings, greatly improving the recognition accuracy of electrical loads, and effectively capturing the energy consumption characteristics of instantaneous changes and stable operation stages. The application of the recursive feature elimination algorithm ensures that only key features that contribute to electrical appliance identification are retained, avoiding the impact of redundant data on system performance. The dynamic correction of the Kalman filter effectively eliminates noise interference and ensures the stability and continuity of the data. Through the time series cross-correlation algorithm, the energy output of distributed power generation equipment and the electrical load demand are synchronized without time delay, so that the data between source and load are accurately mapped. This real-time linkage mechanism not only improves the efficiency of energy allocation, but also can capture tiny load fluctuations and energy consumption differences during carbon emission monitoring, thereby significantly improving the accuracy of carbon emission calculations, providing reliable data support for subsequent dynamic load regulation and carbon emission management, and realizing full process optimization from data collection to carbon emission assessment.

[0175] In the previous steps, the source-load synchronization data was aligned without time lag through the time series cross-correlation algorithm, and the energy matching of the electrical load and the distributed energy output was completed. At the same time, the corrected key feature data provides a basis for the accurate calculation of carbon emissions. On this basis, the core task of step S4 is to calculate the carbon emissions of each electrical appliance through a time-varying function based on these aligned data, and to comprehensively evaluate the current energy consumption and carbon emission status in combination with the global energy consumption difference index and the dynamic carbon load pressure index. By analyzing the impact of electrical energy consumption deviation and load growth on instantaneous carbon emissions, it is determined whether the current load and carbon emissions need to be regulated. At the same time, when the evaluation shows abnormalities, the autoregressive model is used to predict the carbon emission trend in the short term in the future to provide a scientific basis for load optimization and regulation. The ultimate goal of step S4 is to ensure that the system can dynamically and real-time adjust the load operation strategy to achieve accurate carbon emission management and optimization.

[0176] Step S4 includes the following contents:

[0177] S4.1, based on the real-time power generation structure of the power grid and the types and actual operation of distributed power generation equipment, obtain the dynamic carbon emission factors of the power grid and distributed energy to reflect the changes in the carbon emission intensity of electricity sources.

[0178] Among them, the dynamic acquisition of the carbon emission factor per unit of electricity depends on the monitoring and analysis of the real-time power generation structure of the power grid. The specific process includes: First, real-time acquisition of the output data of various power generation sources in the power grid, such as thermal power, hydropower, wind power, solar power and other different types of power generation. The proportion of the output of each power generation source in the total power generation will directly affect the carbon emission intensity of the power grid. The carbon emission factor of fossil fuel power generation sources such as thermal power is relatively high, while the carbon emission factor of renewable energy sources such as wind power and photovoltaics is close to zero. The system monitors the output power of different power generation sources in real time, and combines the fixed carbon emission factors corresponding to each type of power generation, and uses the weighted average method to calculate the carbon emission factor per unit of electricity of the power grid at the current moment. The specific formula is: Among them, P i (t) represents the output power of the i-th power source at time t, P total (t) is the total power generation of the power grid, f i is the fixed carbon emission factor of the i-th power source. In this way, the system can accurately calculate the carbon emission factor per unit of electricity at each time point according to the dynamic changes of the real-time power generation structure, ensuring the timeliness and accuracy of the carbon emission calculation.

[0179] Among them, the calculation of dynamic carbon emission factors of distributed energy mainly depends on the type and actual operation of distributed power generation equipment. For different types of distributed energy, such as photovoltaic power generation, wind power generation, energy storage systems, etc., their carbon emission factors can be calculated or estimated based on the characteristics of the energy source.

[0180] For clean energy sources such as photovoltaics and wind power, in theory, their carbon emission factors are very close to zero, because these energy sources do not directly generate carbon emissions during the power generation process. However, in practical applications, it is necessary to consider carbon emissions indirectly generated by equipment manufacturing, transportation, installation, etc. Therefore, the dynamic carbon emission factors of this type of distributed energy are usually estimated through life cycle analysis (LCA). The system will determine a fixed basic carbon emission factor based on the power generation efficiency and actual operating load of the distributed energy, combined with historical LCA data, and then dynamically adjust it according to the current operating status of the equipment. For example, the efficiency of photovoltaic power generation may be affected by weather and temperature, so the system will correct the basic carbon emission factor based on the real-time monitored power generation efficiency data to ensure its dynamic adjustment.

[0181] For energy storage systems, when energy storage equipment supplies power to the grid or load, the system will calculate the carbon emission factor based on the source of its stored power. If the energy storage equipment's power comes from the grid, the dynamic carbon emission factor of the energy storage should be consistent with the grid's unit power carbon emission factor; if the energy storage equipment comes from renewable energy, the carbon emission factor is close to zero. The specific dynamic carbon emission factor calculation formula is: Among them, P i(t) represents the output power of different distributed energy resources at time t, P total (t) is the total output power of distributed energy. In this way, the system dynamically adjusts the carbon emission factor of distributed energy to ensure the accuracy of carbon emission calculation.

[0182] S4.2, calculate the real-time carbon emissions of each appliance using the following formula:

[0183] Among them, C(t) is the carbon emission of the appliance at time t, P(t) is the power consumption of the appliance, and P grid and P total are the power supply of the grid and distributed energy, respectively, grid (t) and f dist (t) are the dynamic carbon emission factors of the power grid and distributed energy, respectively, to ensure accurate calculation of carbon emissions under the hybrid power supply mode.

[0184] S4.3, conduct an assessment of the deviation of electrical appliance energy consumption and calculate the global energy consumption difference index, which is used to quantify the difference between the current energy consumption of electrical appliances and the historical benchmark energy consumption. The specific acquisition process is as follows:

[0185] Based on the corrected source-load synchronization data, the energy consumption characteristics of the electrical appliances at the current time t are extracted to form an energy consumption characteristic vector Among them, e i (t) represents the instantaneous power consumption of the i-th appliance at time t, and N is the number of appliances.

[0186] The energy consumption feature vectors of the past M time windows are selected from the historical data to construct the historical energy consumption feature space matrix H: H = [E(t-1), E(t-2), ..., E(tM)]; where E(tk) is the energy consumption feature vector at time tk.

[0187] Perform singular value decomposition (SVD) on the historical energy consumption feature space matrix and obtain: Among them, U and V are orthogonal matrices, and Σ is a diagonal singular value matrix.

[0188] Select the singular vectors corresponding to the first K largest singular values ​​to construct the energy consumption pattern basis matrix B: B = [u1,u2,…,u K ]; where u K is the kth column of U, K satisfies K <N。

[0189] Project the current energy consumption feature vector onto the energy consumption pattern basis matrix to obtain the projection coefficient vector

[0190] Using the projection coefficient vector and the energy consumption pattern basis matrix, the energy consumption feature vector E′(t) is reconstructed: E′(t)=Ba(t);

[0191] The difference between the actual energy consumption feature vector and the reconstructed energy consumption feature vector is obtained to obtain the reconstruction error vector δ(t): δ(t) = E(t) - E′(t);

[0192] Perform nonlinear transformation on each element of the reconstructed error vector to obtain the difference metric function value Among them, δ i (t) is the i-th element of the reconstruction error vector, λ i is the scale parameter of the i-th appliance, controlling the sensitivity of the difference measure.

[0193] Normalize the difference metric function value to get the global energy consumption difference index Where D max is the maximum possible value of the difference metric function, which is used for normalization processing so that the value range of the global energy consumption difference index is between [0,1].

[0194] The global energy consumption difference index is used to indicate the degree of difference between the overall energy consumption level in the building at the current moment and the historical benchmark energy consumption, mainly reflecting abnormal energy consumption. The index measures the difference between actual energy consumption and the main energy consumption mode to reveal whether the current energy consumption status deviates from the historical normal level. The larger the value of the global energy consumption difference index, the higher the current energy consumption level is than the historical benchmark, and there may be abnormal loads or high-energy consumption equipment in operation, which requires load adjustment or regulation; conversely, the smaller the value, the more stable the current energy consumption is, which is consistent with the historical operating status, and there is no obvious energy consumption anomaly, so there is no need to perform load regulation operations. The setting of this index helps to dynamically monitor the energy consumption status of the building and provide data support for carbon emission management and load optimization.

[0195] S4.4, the instantaneous impact of load growth is evaluated by calculating the rate of impact of load change on carbon emissions, and the dynamic carbon load pressure index is calculated. The specific acquisition process is:

[0196] According to the carbon emission C(t) obtained in step S4.2, its first-order time derivative V(t) is calculated to represent the instantaneous rate of change of carbon emission over time: Among them, V(t) represents the instantaneous change rate of carbon emissions, reflecting the dynamic change trend of carbon emissions.

[0197] Define the nonlinear response function R(t) of carbon emission changes, which is used to amplify the changes in carbon emission rate and capture sudden carbon emission pressure: R(t) = κ·ln(1+[V(t)] 2);

[0198] Where κ is the proportionality coefficient used to adjust the sensitivity of the function, ln is the natural logarithmic function, [V(t)] 2 Represents the square of the carbon emission rate, ensuring that the response function responds to both positive and negative changes.

[0199] The space-time convolution operator is introduced to perform convolution operation on the nonlinear response function of carbon emissions to obtain the cumulative carbon emission pressure function.

[0200] Among them, φ(t-τ) is the kernel function, and the Gaussian kernel function is selected:

[0201] σ is the scale parameter of the kernel function, which controls the sensitivity of the convolution operation to the time difference.

[0202] The cumulative carbon emission pressure function is transformed nonlinearly to define the dynamic carbon load pressure index.

[0203] Among them, λ is the adjustment parameter, which controls the steepness of the exponential function, and μ is the translation parameter, which sets the midpoint position of the exponential function. This form ensures that the value range of the dynamic carbon load pressure index is between (0,1), which is convenient for threshold judgment.

[0204] The dynamic carbon load pressure index is used to indicate the impact of the current rate of change of carbon emissions and its cumulative pressure on the entire system, reflecting the dynamic changes of carbon emissions over time and its immediate demand for load regulation. The index is calculated through the instantaneous rate of change of carbon emissions, nonlinear response and cumulative effect, which can measure the carbon emission pressure brought by load growth and help identify whether regulatory measures are needed. The larger the value, the faster the carbon emission rate grows and the increased carbon load pressure, which means that the urgency of carbon emission regulation and load adjustment increases, and the system may face high carbon emission risks; while the smaller the value, the slower the carbon emission changes, the lower the carbon emission pressure, and the current system is in a stable state, and no immediate regulatory action is needed. Therefore, the dynamic carbon load pressure index is a key indicator in load regulation, which is used to monitor and evaluate the impact of carbon emission changes on the system in real time, and the value directly affects the decision-making of load adjustment.

[0205] S4.5, comprehensively evaluate and calculate the global energy consumption difference index and the dynamic carbon load pressure index to obtain the energy-carbon synchronization regulation index; for example, it is calculated by the following formula:

[0206] Among them, J(t) is the energy-carbon synchronous regulation index used to evaluate the system load status, with a value range of (0,1). The closer it is to 1, the more obvious the load abnormality is and the more control measures need to be taken; the closer it is to 0, the system is in a stable state and no adjustment is required.

[0207] If the energy-carbon synchronization control index is greater than or equal to the control threshold, it means that the energy consumption and carbon emission pressure of the appliance are at a high level at the same time, there is an obvious abnormality, and the control decision needs to be executed immediately. This means that the energy consumption and carbon emission pressure of the appliance are at a high level at the same time, and the system load has exceeded the set safety range. Control measures will be initiated immediately, which may include adjusting the operation strategy of high-energy-consuming appliances, delaying the start-up of high-energy-consuming equipment, or increasing the power supply ratio of distributed energy to quickly reduce the load pressure and carbon emission load, ensure that the system returns to a stable operating state, and generate a signal that needs to be controlled.

[0208] S4.6: If the assessment results show that regulation is needed, the autoregressive model is used to predict the carbon emission trend. The specific steps are as follows:

[0209] Based on historical carbon emission data t-1 ,C t-2 ,…,C t-n}, build an autoregressive model to predict carbon emissions in the future: in, is the prediction step size, α1,α2,…,α n is the model coefficient, ∈(t) is the random error term.

[0210] Predicting the future using autoregressive models The carbon emission trend at each moment is combined with the evaluation results to determine whether load optimization and carbon emission control strategies need to be implemented.

[0211] S4.7, after comprehensive evaluation and prediction results, generate load optimization and carbon emission control strategies to ensure that the system can respond to abnormal energy consumption and carbon emission pressure in real time and dynamically adjust the operation status of electrical appliances. The result data is stored in the local database in time series for subsequent analysis and adjustment.

[0212] In step S4, the calculation of carbon emissions and short-term trend forecasting have been completed through the autoregressive model, and the current energy consumption and carbon emissions status have been comprehensively evaluated in combination with the global energy consumption difference index and the dynamic carbon load pressure index, clarifying the abnormal state and adjustment needs of the system load. Based on these analysis results, the core task of step S5 is to use the carbon emission calculation and prediction results to generate a dynamic load control strategy through a fuzzy control algorithm, and the load adaptive control module automatically executes it. This step not only relies on fuzzy reasoning of real-time data, but also converts the control strategy into specific load adjustment measures through defuzzification, ensuring that the system can respond quickly when energy consumption is abnormal or carbon emission pressure increases, optimize the electrical operation strategy, and achieve efficient management of load and carbon emissions. Step S5 aims to apply the previous analysis and prediction results to automated load control, further enhancing the system's adaptability and control accuracy.

[0213] Step S5 includes the following contents:

[0214] S5.1, if the evaluation results show that regulation is needed, fuzzification is performed on the carbon emission calculation results and trend forecast data, and the continuous input variables are converted into fuzzy language variables. Fuzzy sets are defined, such as "high carbon emissions", "medium carbon emissions", and "low carbon emissions", and the carbon emissions C(t) and the predicted rate of change ΔC(t) are mapped to the corresponding fuzzy levels through membership functions. For example, a triangular or trapezoidal membership function is used to represent the fuzzy set of carbon emissions:

[0215] Among them, C 高 , C 中 are the thresholds for high carbon emissions and medium carbon emissions, respectively. The membership function μ is used to determine the degree to which the current carbon emissions belong to each fuzzy set.

[0216] The predicted rate of change

[0217] S5.2, build a fuzzy control rule base to map different carbon emissions and change trend combinations to control strategies. Typical fuzzy control rules include:

[0218] If the carbon emissions are “high carbon emissions” and the predicted rate of change is “increasing”, the “strong load reduction” strategy is adopted;

[0219] If the carbon emission is "medium carbon emission" and the predicted rate of change is "reduction", the "moderate load adjustment" strategy is adopted;

[0220] If the carbon emissions are "low carbon emissions" and the predicted rate of change is "stable", maintain the "current load state";

[0221] S5.3, using the fuzzy inference engine, matches the input fuzzy variables (carbon emissions and change trends) with the rule base. Based on the matching in the rule base, the Max-Min inference method or other appropriate inference methods are used to generate the membership of different control strategies. The inference engine will obtain the fuzzy membership value of the corresponding output control strategy based on the membership of the input fuzzy variables in the fuzzy set and the fuzzy rule base.

[0222] S5.4, defuzzify the fuzzy reasoning results and convert the fuzzy control strategy into a specific load adjustment operation. Use the centroid method or other defuzzification methods to calculate the precise value of the control strategy. For example, the output of the control intensity may be a range, and the specific load reduction amount ΔP is calculated by defuzzification, which will be used to guide the adjustment operation of the electrical load.

[0223] Among them, μ 削减 is the membership function of the reduction strategy, and P is the load reduction amount.

[0224] S5.5, passes the defuzzified specific control strategy to the load adaptive control module. This module automatically controls the electrical appliances in the building according to the control strategy, including adjusting the operating power of high-energy-consuming appliances, delaying or shutting down some equipment, and giving priority to the use of distributed energy. The load adaptive control module automatically executes the strategy through real-time data interaction with the appliances, and dynamically adjusts the load control plan according to the feedback information to ensure the adaptive optimization of the system load.

[0225] S5.6, after the load control is executed, the load adaptive control module will continuously monitor the operating status and carbon emission changes of the electrical appliances, and compare the actual feedback data with the control target. If the control result does not meet the expectation, the control strategy will be automatically adjusted, and the self-learning algorithm will be used to optimize the future fuzzy control rules and reasoning models to ensure the gradual refinement of the load control strategy and the continuous improvement of the control effect.

[0226] The present invention can accurately identify abnormal fluctuations in energy consumption and carbon emissions in buildings by combining real-time calculation of carbon emissions with short-term trend prediction, ensuring timely intervention and regulation when load increases or carbon emission pressure increases, thereby avoiding the accumulation of peak energy consumption and excess carbon emissions. The collaborative analysis of the dynamic carbon load pressure index and the global energy consumption difference index effectively improves the sensitivity to load changes, ensuring that even subtle load fluctuations can be responded to and adjusted in a timely manner. The dynamic control strategy generated by the fuzzy control algorithm not only makes adjustments based on real-time data, but also optimizes the load in advance through the prediction of the autoregressive model, significantly improving the response speed of building load control and the accuracy of carbon emission management. It can then intelligently manage the electrical load in the building, optimize the proportion of distributed energy use, and reduce the impact of high-energy-consuming equipment on overall carbon emissions, thereby effectively controlling and reducing the overall carbon emission level of the building, and achieving a high degree of integration and intelligence in building energy use and carbon emission management.

[0227] Example 2

[0228] Figure 2 The present invention provides a building carbon emission monitoring system based on big data collection, comprising:

[0229] Signal acquisition unit, feature screening and correction unit, energy alignment unit, carbon emission calculation and evaluation unit and dynamic control module unit;

[0230] Signal acquisition unit: Use low-frequency sampling equipment to obtain electrical signals in residential buildings, and classify and process the signals according to steady-state and transient characteristics; the steady-state characteristics are extracted through Fourier transform, and the transient characteristics are extracted through adaptive wavelet analysis, and the instantaneous start-up and stable operation characteristics of the electrical appliances are recorded respectively; the extracted steady-state and transient characteristic data are transmitted to the feature screening unit.

[0231] Feature screening and correction unit: The classified and extracted steady-state and transient features of electrical appliances are screened through a recursive feature elimination algorithm, and key features that contribute greatly to electrical appliance identification are retained. The screened key feature values ​​are dynamically corrected through a Kalman filter to filter out signal fluctuations and noise to ensure the consistency of feature data. The corrected key feature data is transmitted to the energy alignment unit.

[0232] Energy alignment unit: Based on the corrected key feature data, the real-time output data of distributed power generation equipment is obtained, and the time series cross-correlation algorithm is used to align the energy output of distributed energy with the electrical load demand without time delay, ensuring data linkage and synchronization between source and load; the aligned source-load synchronization data is transmitted to the carbon emission calculation and evaluation unit.

[0233] Carbon emission calculation and evaluation unit: Based on the aligned source-load synchronization data, the carbon emissions of each electrical appliance are calculated, and the instantaneous impact of the energy consumption deviation and load growth of the electrical appliance is evaluated to form a comprehensive evaluation result of energy consumption and carbon emissions; when the evaluation result shows an abnormality, the autoregressive model is used to predict the carbon emission trend; the evaluation result and trend prediction data are transmitted to the dynamic control unit.

[0234] Dynamic control module unit: Based on the carbon emission calculation results and trend forecast data, a dynamic load control strategy is generated through a fuzzy control algorithm to automatically adjust the operating status of electrical appliances, optimize the proportion of distributed energy use, and finally automatically execute it through the load adaptive control module.

[0235] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0236] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0237] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0238] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A building carbon emission monitoring method based on big data collection, characterized in that: Includes steps: S1, using low-frequency sampling equipment to obtain electrical signals in residential buildings, classified according to steady-state and transient characteristics, transient characteristics extracted by adaptive wavelet analysis, steady-state characteristics based on Fourier transform processing, respectively record the instantaneous start-up and stable operation characteristics of electrical appliances; the extracted steady-state and transient characteristic data are stored in a local database; S2, through the recursive feature elimination algorithm, the classified and extracted electrical appliance steady-state and transient features are screened, the key features with high contribution to electrical appliance identification are retained, and the Kalman filter is used to dynamically correct the key feature values, filter out signal fluctuations and noise, and ensure that the key feature data is consistent and accurate at different times; S3, based on the corrected key feature data, obtains the real-time output data of distributed power generation equipment, aligns the energy output of distributed energy with the load demand of electrical appliances through the time series cross-correlation algorithm, and ensures the time synchronization and time-delay-free mapping of source-load data; S4, based on the corrected source-load synchronization data, calculate the carbon emissions of each appliance, conduct a comprehensive assessment of energy consumption and carbon emissions by evaluating the instantaneous impact of appliance energy consumption deviation and load growth, and predict the carbon emission trend through an autoregressive model when the assessment result deviates from expectations, and use the assessment and prediction results as the basis for load optimization and regulation; S5, based on the carbon emission calculation results and trend forecast data, generates a dynamic load control strategy based on the fuzzy control algorithm and automatically executes it through the load adaptive control module.

2. The method for monitoring building carbon emissions based on big data collection according to claim 1 is characterized in that: Step S1 includes the following contents: S1.1, use low-frequency sampling equipment to continuously sample the electrical signals of residential buildings to capture the steady-state and transient characteristics of electrical appliance operation; the collected data is recorded with a timestamp and synchronously written to the local cache; S1.2, the transient characteristics of electrical signals refer to the signal fluctuations during the instantaneous change of electrical start-up and shutdown; the transient characteristics are extracted by adaptive wavelet transform, and the specific steps are as follows: Use the selected wavelet basis function; decompose the electrical signal in the frequency domain through multi-scale wavelet decomposition to capture the transient characteristics of the electrical appliance during start-up and shutdown; after the extraction is completed, the transient characteristic data is marked according to the event triggering time and then stored in the local database; S1.3, steady-state characteristics refer to the characteristic parameters of the electrical appliance when it is running stably; the steady-state characteristics are extracted through Fourier transform, and the specific steps are as follows: First, the steady-state signal is windowed; the electrical signal is converted from the time domain to the frequency domain using the fast Fourier transform, and the spectral characteristics of the signal are analyzed; the steady-state characteristic data is stored in time sequence and associated with the corresponding electrical load state; S1.4, using an event-driven trigger mechanism, first identifies sudden changes in electrical signals through a mutation detection algorithm; if the instantaneous change of the signal exceeds a preset threshold, the corresponding segment signal is identified as a transient feature; otherwise, the corresponding segment signal is identified as a steady-state feature by default; S1.5, based on the classification results, the steady-state and transient characteristics are stored in different data sets, with timestamps and device IDs attached for quick indexing.

3. The method for monitoring building carbon emissions based on big data collection according to claim 2 is characterized in that: Step S2 includes the following contents: S2.1, based on the steady-state and transient characteristics of the electrical equipment extracted in step S1, first normalize the extracted characteristics; S2.2, in feature screening, a recursive feature elimination algorithm is used to progressively screen the normalized steady-state and transient feature data; the output is a screened key feature set, including steady-state features and transient features; S2.3, dynamically correct the selected key characteristic data, and use Kalman filter to filter noise and smooth the signal; the output is the key characteristic data after dynamic correction, including the correction values ​​of steady-state and transient characteristics; S2.4, the key feature data after dynamic correction is recorded and stored in the local database according to the timestamp; different storage strategies are set for transient features and steady-state features: transient feature data is marked and stored according to event triggers, and steady-state features are stored according to fixed time windows.

4. The method for monitoring building carbon emissions based on big data collection according to claim 3 is characterized in that: Step S3 includes the following contents: S3.1, obtaining real-time output data of distributed power generation equipment, including output power, real-time voltage and current parameters of distributed energy equipment; S3.2, in order to ensure that the output data of the distributed generation equipment matches the electrical load characteristic data in time, the timestamps of the two need to be preprocessed and synchronized: Standardize the timestamps of the real-time output data of distributed power generation equipment and the corrected load characteristic data to ensure that all data are recorded according to the same time reference; if the sampling frequencies of distributed energy data and load characteristic data are different, use interpolation or downsampling technology to reduce the frequency of the higher frequency data to align it with the lower frequency data to ensure that the data of the two are compared and analyzed within the same time window; S3.3, in order to achieve the time-delay-free alignment of the output data of distributed power generation equipment and the electrical load demand, the time series cross-correlation algorithm is used to associate and map the data. The specific steps are as follows: A sliding time window is set on the load demand data and distributed energy output data, and the time relationship between load demand and distributed energy output is analyzed by calculating the cross-correlation function between the two time series; the cross-correlation function is defined as follows: R xy (τ)=∑ t x(t)·y(t-τ); Among them, x(t) is the load demand data, y(t) is the output data of distributed generation, and τ is the time lag. By finding the peak position of the cross-correlation function, the optimal alignment time point between distributed energy output and load demand is determined to ensure a time-lag-free mapping between the two. According to the calculation results of the cross-correlation function, the time alignment of the distributed generation equipment output data and the load demand data is adjusted; if a delay τ is detected, it is corrected by data shifting or interpolation technology to ensure that the source-load data can be matched without lag; S3.4, after completing the time series alignment, establish the mapping relationship between the energy output of distributed energy and the electrical load demand; the specific method is: Based on the time-aligned data, a functional relationship between the output power of distributed energy and the electrical load demand is established, and the power matching degree between the two is calculated using a linear regression or multivariate regression model; By calculating the total output power of distributed energy and the load demand of electrical appliances in each time window, analyze whether the two match; if the energy output is less than the load demand, record the energy gap information; if the energy output is greater than the load demand, analyze the energy storage or feedback path; S3.5, after completing time synchronization and data mapping, the aligned and corrected distributed energy output data and electrical load demand data are stored in the local database.

5. The method for monitoring building carbon emissions based on big data collection according to claim 4 is characterized in that: Step S4 includes the following contents: S4.1, based on the real-time power generation structure of the power grid and the types and actual operation of distributed power generation equipment, obtain the dynamic carbon emission factors of the power grid and distributed energy to reflect the changes in the carbon emission intensity of power sources; S4.2, calculate the real-time carbon emissions of each appliance using the following formula: Among them, C(t) is the carbon emission of the appliance at time t, P(t) is the power consumption of the appliance, and P grid and P total are the power supply of the grid and distributed energy, respectively, grid (t) and f dist (t) Dynamic carbon emission factors for the power grid and distributed energy, respectively; S4.3, conduct an assessment of the deviation of electrical appliance energy consumption and calculate the global energy consumption difference index, which is used to quantify the difference between the current energy consumption of electrical appliances and the historical benchmark energy consumption. The specific acquisition process is as follows: Based on the corrected source-load synchronization data, the energy consumption characteristics of the electrical appliances at the current time t are extracted to form an energy consumption characteristic vector Among them, e i (t) represents the instantaneous power consumption of the i-th appliance at time t, and N is the number of appliances; Select the energy consumption feature vectors of the past M time windows from the historical data and construct the historical energy consumption feature space matrix H: H = [E(t-1), E(t-2), ..., E(tM)]; where E(tk) is the energy consumption feature vector at time tk; Perform singular value decomposition on the historical energy consumption feature space matrix and obtain: Among them, U and V are orthogonal matrices, ∑ is a diagonal singular value matrix; Select the singular vectors corresponding to the first K largest singular values ​​to construct the energy consumption pattern basis matrix B: B = [u1,u2,…,u K ]; where u K is the kth column of U, K satisfies K <N; Project the current energy consumption feature vector onto the energy consumption pattern basis matrix to obtain the projection coefficient vector a(t): Using the projection coefficient vector and the energy consumption pattern basis matrix, the energy consumption feature vector E′(t) is reconstructed: E′(t)=Ba(t); The difference between the actual energy consumption feature vector and the reconstructed energy consumption feature vector is obtained to obtain the reconstruction error vector δ(t): δ(t) = E(t) - E′(t); Perform nonlinear transformation on each element of the reconstructed error vector to obtain the difference metric function value D(t): Among them, δ i (t) is the i-th element of the reconstruction error vector, λ i is the scale parameter of the i-th appliance, controlling the sensitivity of the difference measure; Normalize the difference measurement function value to obtain the global energy consumption difference index GEDI(t): Where D max It is the maximum value of the difference measurement function and is used for normalization processing so that the value range of the global energy consumption difference index is between [0,1].

6. The method for monitoring building carbon emissions based on big data collection according to claim 5 is characterized in that: Step S4 also includes the following contents: S4.4, the instantaneous impact of load growth is evaluated by calculating the rate of impact of load change on carbon emissions, and the dynamic carbon load pressure index is calculated. The specific acquisition process is: According to the carbon emission C(t) obtained in step S4.2, its first-order time derivative V(t) is calculated to represent the instantaneous rate of change of carbon emission over time: Among them, V(t) represents the instantaneous change rate of carbon emissions, reflecting the dynamic change trend of carbon emissions; Define the nonlinear response function R(t) of carbon emission changes to amplify the changes in carbon emission rate and capture sudden carbon emission pressure: R(t) = κ·ln(1+[V(t)] 2 ), where κ is the proportionality coefficient used to adjust the sensitivity of the function, ln is the natural logarithmic function, [V(t)] 2 represents the square of the carbon emission rate, ensuring that the response function responds to both positive and negative changes; The spatiotemporal convolution operator is introduced to perform convolution operation on the nonlinear response function of carbon emissions to obtain the cumulative carbon emission pressure function S(t): Among them, φ(t-τ) is the kernel function, and the Gaussian kernel function is selected: σ is the scale parameter of the kernel function, which controls the sensitivity of the convolution operation to the time difference; The cumulative carbon emission pressure function is transformed nonlinearly to define the dynamic carbon load pressure index DCLPI(t): Among them, λ is the adjustment parameter, which controls the steepness of the exponential function, and μ is the translation parameter, which sets the midpoint position of the exponential function; S4.5, comprehensively evaluate and calculate the global energy consumption difference index and the dynamic carbon load pressure index to obtain the energy-carbon synchronization regulation index; If the energy-carbon synchronization control index is greater than or equal to the control threshold, it means that the energy consumption and carbon emission pressure of the appliance are both at a high level, and there is an obvious abnormality. The control decision needs to be executed immediately, and a control signal is generated; S4.6: If the assessment results show that regulation is needed, the autoregressive model is used to predict the carbon emission trend. The specific steps are as follows: Based on historical carbon emission data t-1 ,C t-2 ,…,C t-n }, build an autoregressive model to predict carbon emissions in the future: in, is the prediction step size, α1, α2, …, α n is the model coefficient, ∈(t) is the random error term.

7. The method for monitoring building carbon emissions based on big data collection according to claim 6 is characterized by: Step S5 Includes the following: S5.1, if the assessment results show that regulation is needed, fuzzification processing is performed on the carbon emission calculation results and trend forecast data to convert continuous input variables into fuzzy linguistic variables; S5.2, construct a fuzzy control rule base to map different combinations of carbon emissions and change trends to control strategies; S5.3, using the fuzzy inference engine, matching the input fuzzy variables with the rule base; based on the matching situation in the rule base, using the Max-Min inference method or other appropriate inference methods to generate the membership of different control strategies; S5.4, defuzzify the fuzzy reasoning results, convert the fuzzy control strategy into a specific load adjustment operation; use the center of gravity method to calculate the exact value of the control strategy; S5.5, passing the defuzzified specific control strategy to the load adaptive control module; S5.6, after the load control is executed, the load adaptive control module will continuously monitor the operating status and carbon emission changes of the electrical appliances, and compare the actual feedback data with the control target; if the control result does not meet the expectation, the control strategy will be automatically adjusted.

8. A building carbon emission monitoring system based on big data collection, used to implement a building carbon emission monitoring method based on big data collection as described in any one of claims 1 to 7, characterized in that: include: Signal acquisition unit, feature screening and correction unit, energy alignment unit, carbon emission calculation and evaluation unit and dynamic control module unit; Signal acquisition unit: Use low-frequency sampling equipment to obtain electrical signals in residential buildings, and classify and process the signals according to steady-state and transient characteristics; Steady-state features are extracted through Fourier transform, and transient features are extracted through adaptive wavelet analysis, respectively recording the instantaneous start-up and stable operation characteristics of the electrical appliance; the extracted steady-state and transient feature data are transmitted to the feature screening unit; Feature screening and correction unit: The classified and extracted electrical steady-state and transient features are screened through a recursive feature elimination algorithm, and the key features that contribute most to electrical identification are retained. The screened key feature values ​​are dynamically corrected through a Kalman filter to filter out signal fluctuations and noise to ensure the consistency of feature data. The corrected key feature data is transmitted to the energy alignment unit. Energy alignment unit: Based on the corrected key feature data, the real-time output data of distributed power generation equipment is obtained, and the time series cross-correlation algorithm is used to align the energy output of distributed energy with the electrical load demand without time lag, ensuring data linkage and synchronization between source and load; the aligned source-load synchronization data is transmitted to the carbon emission calculation and evaluation unit; Carbon emission calculation and evaluation unit: Based on the aligned source-load synchronization data, the carbon emissions of each appliance are calculated, and the instantaneous impact of the energy consumption deviation and load growth of the appliance are evaluated to form a comprehensive evaluation result of energy consumption and carbon emissions; when the evaluation result shows an abnormality, the autoregressive model is used to predict the carbon emission trend; The evaluation results and trend prediction data are transmitted to the dynamic control unit; Dynamic control module unit: Based on the carbon emission calculation results and trend forecast data, a dynamic load control strategy is generated through a fuzzy control algorithm to automatically adjust the operating status of electrical appliances, optimize the proportion of distributed energy use, and finally automatically execute it through the load adaptive control module.

Images