Method and system for dynamic aggregation of electricity sub-metering data based on non-intrusive monitoring

Abstract

The present application relates to the technical field of data dynamic aggregation, and particularly relates to a power sub-item data dynamic aggregation method and system based on non-intrusive monitoring. The method comprises the following steps: obtaining total power consumption data and historical time sequence records of power consumption equipment, extracting fluctuation characteristic data from the total power consumption data, combining the historical time sequence records to construct potential coupling modes among the power consumption equipment; constructing a real-time relationship matrix according to the potential coupling modes, analyzing element values in the real-time relationship matrix to generate a strong coupling link set; extracting dynamic change indicators from the strong coupling link set, grouping the dynamic change indicators to generate change path clusters. Through the deep combination of non-intrusive monitoring and dynamic aggregation technology, the present application realizes accurate decomposition and real-time updating of complex power data, effectively improves the accuracy and practicality of power sub-item data, and provides a scientific basis for energy management.

Description

Technical Field

[0001] This invention relates to the field of dynamic data aggregation technology, and in particular to a method and system for dynamic aggregation of electrical energy sub-item data based on non-intrusive monitoring. Background Technology

[0002] In modern buildings and industrial production, electricity is a fundamental energy source, and its consumption directly affects energy efficiency, cost control, and the achievement of carbon emission targets. With the increasing demand for green and low-carbon transformation and refined energy management, accurately acquiring the electricity consumption of various electrical equipment and processes has become a core prerequisite for energy optimization and scheduling, as well as a fundamental support for energy-saving renovations and energy allocation.

[0003] Currently, there are two main methods for widely used sub-metering of electricity: one is to install independent sub-meters in each power circuit and collect energy consumption through invasive deployment; the other is to estimate the sub-item ratio through short-term on-site testing and then infer the sub-item energy consumption. Both of these methods have obvious drawbacks in complex and large-scale scenarios.

[0004] On the one hand, adjustments to building functions, additions or removals of equipment, and changes in electricity usage habits can quickly invalidate the component proportions established through on-site testing, leading to a disconnect between component data and reality. Especially in complex scenarios such as renovations of old buildings and commercial complexes, fixed-proportion splitting or occasional verification cannot adapt to dynamic load changes, resulting in decreased accuracy. Furthermore, intrusive sub-meter deployment presents problems such as complex wiring, high construction and maintenance costs, and disruption to normal use during renovations.

[0005] On the other hand, electricity consumption behavior is time-varying and equipment-coupled; the operation of different devices affects each other. For example, turning on the air conditioner reduces lighting demand, while shutting down production equipment at night increases the proportion of office electricity consumption. Existing methods use fixed weights for decomposition, ignoring this dynamic interaction. This leads to significant deviations in the sub-item results when load fluctuates or equipment starts and stops simultaneously. For instance, during peak summer seasons, electricity savings from lighting may be mistakenly attributed to air conditioning energy consumption, affecting data reliability.

[0006] Non-intrusive load monitoring technology eliminates the need for individual meters on each device, obtaining load information through analysis of total load meter data. This addresses the drawbacks of traditional invasive metering and represents a significant development direction. However, a key bottleneck currently exists: how to capture real-time interactions between devices, dynamically adjust sub-item ratios, and obtain long-term stable multi-dimensional sub-item data without adding extra invasive hardware. This is the core obstacle restricting its refined and large-scale application and a pressing problem that needs to be solved in this field. Summary of the Invention

[0007] The main objective of this invention is to provide a method and system for dynamic aggregation of electrical energy sub-item data based on non-invasive monitoring. This invention achieves accurate decomposition and real-time updating of complex electrical energy data through the deep integration of non-invasive monitoring and dynamic aggregation technologies, effectively improving the accuracy and practicality of electrical energy sub-item data and providing a scientific basis for energy management.

[0008] The technical solution of the present invention is as follows:

[0009] Firstly, a method for dynamic aggregation of electricity component data based on non-intrusive monitoring is proposed, which includes the following steps:

[0010] Obtain total power consumption data and historical time series records of electrical equipment, extract fluctuation feature data from total power consumption data, and combine with historical time series records to construct potential coupling patterns between various electrical equipment.

[0011] Based on the potential coupling patterns, a real-time relationship matrix is ​​constructed, the element values ​​in the real-time relationship matrix are analyzed, and a set of strongly coupled links is generated.

[0012] Extract dynamic change indicators from the set of strongly coupled links, group the dynamic change indicators, and generate change path clusters;

[0013] For each cluster of changing paths, the proportional weights within each cluster are calculated, and the adjusted weight distribution is generated.

[0014] Based on the current total electricity consumption data and the adjusted weight distribution, a real-time input sequence is generated;

[0015] The real-time input sequence is integrated and processed according to the preset business classification attributes to determine the fused input sequence;

[0016] Based on the fused input sequence, update the component proportion model to generate the optimized component proportion;

[0017] The optimized item proportions are associated with preset business verification attributes to generate associated item proportions.

[0018] Based on the associated sub-item ratios, multi-dimensional electricity sub-item data is obtained. The verification module compares the multi-dimensional electricity sub-item data with historical accurate data to generate the final stable data.

[0019] A further improvement of the present invention is that the construction of potential coupling modes between various electrical devices includes:

[0020] Total power consumption data is collected through non-intrusive monitoring equipment, and historical time-series records are obtained from the database. The historical time-series records include total load fluctuations and equipment operation tags.

[0021] A time-series analysis algorithm is used to decompose the total electricity consumption data and extract fluctuation characteristic data;

[0022] Based on the fluctuation characteristics of each electrical device within a preset time period, the total load fluctuation, and the equipment operation tags, the element values ​​of the relationship matrix are calculated using the following formula:

[0023]

[0024] in, The element values ​​of the relation matrix represent the potential coupling patterns between devices. and Indicates different device indexes. Indicates equipment In time Fluctuation characteristic data, Indicates equipment In time Fluctuation characteristic data, Indicates the duration of historical records. Indicates equipment The average value, Indicates equipment The average value;

[0025] Based on the element values ​​of the relation matrix, construct the potential coupling patterns between various electrical devices.

[0026] A further improvement of the present invention is that the generation of the strongly coupled link set includes:

[0027] Extract element values ​​from the real-time relationship matrix;

[0028] Compare the element value with a preset threshold;

[0029] When an element value is greater than a preset threshold, it is marked, and inter-device links are generated, forming a set of strongly coupled links.

[0030] A further improvement of the present invention is that the generated change path cluster includes:

[0031] Extract multiple dynamic change indicators from a set of strongly coupled links;

[0032] The formula for calculating the distance between multiple dynamically changing indicators is:

[0033]

[0034] in, Indicates the path of change and Distance indicators between Representing a path The A dynamic indicator, Representing a path The A dynamic indicator, Indicates the dimension of the indicator;

[0035] Based on the distance index, a clustering algorithm is applied to group the changing paths, generating multiple changing path clusters.

[0036] A further improvement of the present invention is that the generation of the adjusted weight distribution includes:

[0037] For the variable path cluster, extract the number of paths and the magnitude of change within the cluster;

[0038] Based on the number and magnitude of changes in the paths within a cluster, calculate the proportional weights corresponding to the changing paths of electrical equipment within each cluster.

[0039] Sort the proportional weights and reset the weights of those below the average level to zero;

[0040] The adjusted weight distribution is obtained using the following formula:

[0041]

[0042] in, This represents the adjusted weight distribution of the r-th cluster. This represents the intra-cluster proportional weights calculated in the original dataset. This indicates the average weight level.

[0043] A further improvement of the present invention is that determining the fused input sequence includes:

[0044] Obtain the real-time input sequence and extract the data dimensions related to the preset classification attributes from the real-time input sequence;

[0045] The data in the real-time input sequence is grouped according to data dimensions to generate categorized data subsets;

[0046] The categorized data subsets are weighted and combined with multi-dimensional data to generate an integrated data structure;

[0047] Based on the integrated data structure, a fused input sequence is generated.

[0048] A further improvement of this invention is that the generation of the optimized component proportions includes:

[0049] The total load is obtained through non-invasive monitoring equipment, and the total load is decoupled into multiple sub-indicators according to preset rules;

[0050] Time series analysis is performed on the fused input sequence to extract the dynamic change characteristics of each sub-indicator;

[0051] Based on the dynamic change characteristics and the predicted value of the total load, a component proportion model is constructed;

[0052] Perform deviation analysis on the component proportion model and output the deviation value;

[0053] Compare the deviation value with the preset limit to determine whether iterative optimization is triggered;

[0054] If iterative optimization is triggered, the optimization algorithm is used to iteratively process the parameter values ​​until the deviation value converges within the preset limit, then training stops, the component proportion model is updated, and the optimized component proportion is output.

[0055] A further improvement of the present invention is that the generation of final stable data includes:

[0056] Multi-dimensional electrical energy breakdown data is input into the verification module and compared with historical accurate data to obtain the comparison results;

[0057] If the comparison results meet the preset verification rules, the multi-dimensional electrical energy sub-item data are deemed valid and are regarded as the final stable data output.

[0058] If the comparison results do not meet the preset verification rules, the multi-dimensional energy component data will be corrected, and the corrected multi-dimensional energy component data will be output as the final stable data.

[0059] Secondly, a dynamic aggregation system for electricity component data based on non-intrusive monitoring is proposed, including:

[0060] The time series analysis module acquires the total power consumption data and historical time series records of electrical equipment, extracts fluctuation feature data from the total power consumption data, and constructs the potential coupling patterns between various electrical equipment by combining the historical time series records.

[0061] The relation matrix construction module constructs a real-time relation matrix based on potential coupling patterns, analyzes the element values ​​in the real-time relation matrix, and generates a set of strongly coupled links.

[0062] The clustering and grouping module extracts dynamic change indicators from the set of strongly coupled links, groups the dynamic change indicators, and generates change path clusters.

[0063] The weight adjustment module calculates the proportional weights within each cluster for the changing path clusters and generates the adjusted weight distribution.

[0064] The real-time input tracking module generates a real-time input sequence based on the current total electricity consumption data and the adjusted weight distribution;

[0065] The real-time input fusion module integrates and processes the real-time input sequence according to the preset business classification attributes to determine the fused input sequence;

[0066] The component proportion update module updates the component proportion model based on the fused input sequence and generates the optimized component proportion.

[0067] The item ratio association module associates the optimized item ratios with preset business verification attributes to generate the associated item ratios.

[0068] The data output and verification module obtains multi-dimensional power energy sub-item data based on the associated sub-item ratios. The verification module compares the multi-dimensional power energy sub-item data with historical accurate data to generate the final stable data.

[0069] The technical effects of this invention are as follows:

[0070] This invention extracts fluctuation characteristic data from total electricity consumption data using a time-series analysis algorithm, constructs potential coupling patterns between devices by combining historical time-series records, and identifies strong coupling links through a real-time relationship matrix. This effectively avoids data distortion caused by device interaction interference and improves the accuracy of electricity item segmentation in complex coupling scenarios. Based on the dynamic change indicators of strong coupling links, this invention generates change path clusters, calculates the proportional weights within the clusters to optimize influencing factors, and integrates real-time input sequences with preset business classification attributes to achieve dynamic updates of the item segmentation proportional model. A verification module is introduced to compare the generated multi-dimensional electricity itemization data with historical accurate data. Abnormal data is fine-tuned using smoothing filters and other methods, forming a dual guarantee mechanism of "dynamic generation + verification correction." This significantly reduces the risk of data distortion caused by algorithm local extrema and transient interference, ensuring stable and reliable output data. This invention relies on non-intrusive monitoring technology to acquire total electricity consumption data, eliminating the need to install intrusive sub-meters in each power circuit, thus significantly reducing hardware deployment costs and subsequent maintenance workload. At the same time, the stable, accurate, and multi-dimensional electricity breakdown data output can provide users with a clear analysis of electricity consumption structure, providing scientific data support for energy management work such as energy optimization scheduling, energy-saving renovation, and cost control, significantly improving the engineering practicality and economic value of the technical solution. Attached Figure Description

[0071] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating an embodiment of the present invention; Figure 2 This is a schematic diagram of the modular structure of the system according to an embodiment of the present invention. Detailed Implementation

[0072] Example 1

[0073] This embodiment constructs a dynamic aggregation method for electricity consumption data based on non-intrusive monitoring. It extracts fluctuation characteristic data from total electricity consumption data using a time-series analysis algorithm, constructs potential coupling patterns between devices by combining historical time-series records, and identifies strong coupling links through a real-time relationship matrix. This effectively avoids data distortion caused by device interaction interference, improving the accuracy of electricity consumption data in complex coupling scenarios. The invention generates change path clusters based on dynamic change indicators of strong coupling links, calculates the proportional weights within the clusters to optimize influencing factors, and integrates real-time input sequences with preset business classification attributes to achieve dynamic updates of the component proportion model. A verification module is introduced to compare the generated multi-dimensional electricity consumption data with historical accurate data. Abnormal data is fine-tuned using smoothing filters and other methods, forming a dual guarantee mechanism of "dynamic generation + verification correction." This significantly reduces the risk of data distortion caused by algorithm local extrema and transient interference, ensuring stable and reliable output data. This invention relies on non-intrusive monitoring technology to acquire total electricity consumption data, eliminating the need to install intrusive sub-meters in each power circuit, thus significantly reducing hardware deployment costs and subsequent maintenance workload. At the same time, the stable, accurate, and multi-dimensional electricity breakdown data output can provide users with a clear analysis of electricity consumption structure, providing scientific data support for energy management work such as energy optimization scheduling, energy-saving renovation, and cost control, significantly improving the engineering practicality and economic value of the technical solution.

[0074] For methods of dynamic aggregation of electricity component data based on non-intrusive monitoring, such as Figure 1 As shown, the specific steps include the following:

[0075] S1. Obtain the total power consumption data and historical time series records of the electrical equipment, extract fluctuation feature data from the total power consumption data, and construct the potential coupling mode between the electrical equipment by combining the historical time series records.

[0076] In this embodiment, S1 includes the following specific steps:

[0077] Total power consumption data is collected through non-intrusive monitoring equipment, and historical time-series records are obtained from the database. The historical time-series records include total load fluctuations and equipment operation tags.

[0078] A time-series analysis algorithm is used to decompose the total electricity consumption data and extract fluctuation characteristic data;

[0079] Based on the fluctuation characteristics of each electrical device within a preset time period, the total load fluctuation, and the equipment operation tags, the element values ​​of the relationship matrix are calculated using the following formula:

[0080]

[0081] in, The element values ​​of the relation matrix represent the potential coupling patterns between devices. and Indicates different device indexes. Indicates equipment In time Fluctuation characteristic data, Indicates equipment In time Fluctuation characteristic data, Indicates the duration of historical records. Indicates equipment The average value, Indicates equipment The average value;

[0082] Based on the element values ​​of the relation matrix, construct the potential coupling patterns between various electrical devices.

[0083] In this embodiment, the non-intrusive monitoring device can be a power monitoring system or a smart meter terminal to acquire total electricity consumption data within the area to be analyzed. This total electricity consumption data is typically a power curve or current / voltage waveform that varies over time. Since the total electricity consumption data is the superposition of the operating states of multiple electrical devices within the area to be analyzed, it includes the operating characteristics of each individual device. Historical time-series records are retrieved from the database. By analyzing these records, microscopic device behaviors can be extracted from the macroscopic total data. The historical time-series records contain fluctuations in total load under different time periods and environmental conditions, as well as the operating tags of known devices. By comparing and aligning the real-time collected total electricity consumption data with the historical time-series records, the system can use time-series analysis algorithms (such as sliding window analysis, wavelet transform, or Fourier transform) to deconstruct the total load curve.

[0084] In this embodiment, the core of the time-series analysis algorithm lies in identifying the fluctuation characteristics in the load curve. The starting and stopping of equipment, or the switching of operating modes (e.g., the start and stop of an air conditioner compressor, or changes in the speed of a washing machine), will generate specific fluctuation textures on the total load curve. The system separates these fluctuations from the background noise, forming a series of feature sequences representing the operating state of the equipment.

[0085] In this embodiment, fluctuation characteristic data is used to reflect the correlation strength between devices and to explore whether there is a collaborative or mutually influential relationship between different devices.

[0086] In this embodiment, based on the fluctuation feature data obtained from the decomposition, the relationship matrix element values ​​of the potential coupling patterns between devices are constructed. For each identified device index, the system extracts its fluctuation feature data sequence within a specific time period and uses a statistical correlation calculation method to generate a relationship matrix to accurately describe the correlation strength between different devices.

[0087] Specifically, there are two electrical devices, namely equipment. and The system calculates the product of the deviations of the fluctuation characteristics of two devices from their respective means over time. To determine the cumulative effect on The numerical value, in the actual calculation process... The physical meaning of the relation matrix element values, representing potential coupling patterns between devices, lies in measuring the synchronicity of the fluctuation trends of the two devices. If... A larger value indicates that the device... When the fluctuation characteristics are higher than its average value, the equipment The fluctuation characteristics of the value also tend to be higher than its average value, and vice versa, which indicates that there is a strong positive correlation between the two.

[0088] In this embodiment, the element values ​​of the relation matrix are used to determine whether a preset threshold is exceeded, and thus decide whether to form a set of strongly coupled links.

[0089] S2. Based on the potential coupling patterns, construct a real-time relationship matrix, analyze the element values ​​in the real-time relationship matrix, and generate a set of strongly coupled links.

[0090] In this embodiment, S2 includes the following specific steps:

[0091] Extract element values ​​from the real-time relationship matrix;

[0092] Compare the element value with a preset threshold;

[0093] When an element value is greater than a preset threshold, it is marked, and inter-device links are generated, forming a set of strongly coupled links.

[0094] In this embodiment, the real-time relationship matrix is ​​used to quantify the actual correlation strength between devices at the current moment or within a specific analysis period. Although potential coupling patterns reveal possible structural connections between devices, the strength of these connections fluctuates with time, environment, and user behavior during actual operation. Therefore, this real-time relationship matrix can intuitively reflect the degree of mutual influence between devices in the current operating environment, and is a key step in transforming qualitative pattern recognition into quantitative numerical analysis.

[0095] Specifically, after constructing the real-time relationship matrix, a preset threshold determination mechanism is introduced to filter out strong correlations with practical analytical significance and eliminate accidental coincidences or background noise. This preset threshold can be adaptively generated based on the statistical distribution characteristics of historical big data, or it can be set by technical personnel according to the sensitivity requirements of the actual scenario. The system iterates through each element value in the matrix and compares it with the preset threshold.

[0096] In this embodiment, if the value of an element in the matrix exceeds a preset threshold, it indicates that there is a significant interaction or dependency between the corresponding two devices. For example, in a linkage scenario where the main device triggers the startup of a slave device, the system marks this as a strongly coupled link. Conversely, if the element value is lower than or equal to the preset threshold, the association between the two is considered weak, and it is not treated as a strongly coupled link. Finally, the system summarizes and stores all device pairs marked as strongly coupled links to obtain a set of strongly coupled links.

[0097] S3. Extract dynamic change indicators from the set of strongly coupled links, group the dynamic change indicators, and generate change path clusters.

[0098] In this embodiment, S3 includes the following specific steps:

[0099] Extract multiple dynamic change indicators from a set of strongly coupled links;

[0100] The formula for calculating the distance between multiple dynamically changing indicators is:

[0101]

[0102] in, Indicates the path of change and Distance indicators between Representing a path The A dynamic indicator, Representing a path The A dynamic indicator, Indicates the dimension of the indicator;

[0103] Based on the distance index, a clustering algorithm is applied to group the changing paths, generating multiple changing path clusters.

[0104] In this embodiment, the dynamic change index characterizes the changes in the connection between devices over time.

[0105] In this embodiment, the change path cluster represents similar change paths between devices.

[0106] In this embodiment, after determining the set of strongly coupled links, the system has grasped the significant interaction relationships between devices. However, this coupling relationship is not static, but rather exhibits dynamic evolution characteristics over time. To gain a deeper understanding of this evolutionary pattern, dynamic change indicators are further extracted from the set of strongly coupled links. Dynamic change indicators can reflect the trend, rate, and fluctuation amplitude of coupling strength changes over time, such as the growth rate of coupling strength, decay period, or abrupt change frequency.

[0107] Specifically, the system performs time-series tracking on each strongly coupled link in the set, recording its state changes over consecutive time segments, thus forming a series of dynamic change indicators describing the coupling evolution. Furthermore, to discover whether similar evolutionary patterns exist between different device pairs, clustering algorithms (such as K-means clustering, hierarchical clustering, or DBSCAN) are introduced. These algorithms aim to group links with similar dynamic change characteristics into one category, thereby identifying typical evolutionary paths within the system.

[0108] In this embodiment, to provide accurate input to the clustering algorithm, a metric for measuring path differences needs to be defined. Let... Indicates the path of change The A dynamic change index, which can be a normalized coupling strength value or a rate of change. This represents the total dimension of the indicator, i.e., the number of features considered. Similarly, Indicates the path of change The A dynamically changing indicator. The system determines the distance between two paths by calculating the differences between them across various dimensions. .

[0109] Specifically, As a path of change and The smaller the distance index between two paths, the more similar they are in their dynamic evolution. This distance can be calculated using Euclidean distance, Manhattan distance, or Dynamic Time Warping (DTW) distance, depending on the characteristics of the data and the analysis requirements. By traversing all path pairs in the set, the system constructs a comprehensive distance matrix.

[0110] In this embodiment, after obtaining the distance index, the present application uses a clustering algorithm to group all the changing paths. The algorithm aggregates paths that are close to each other, i.e., have similar evolution patterns, according to the distance matrix, forming several independent clusters. Each path cluster represents a specific device coupling evolution pattern, such as "synchronous enhancement type," "periodic fluctuation type," or "gradual decoupling type." In this way, the present application can simplify complex device networks into several typical interaction patterns. This not only helps technicians quickly grasp the overall operating status of the system but also provides a benchmark for subsequent anomaly detection—if a path suddenly deviates from its cluster, it may indicate a device failure or abnormal behavior.

[0111] S4. For the changing path clusters, calculate the proportional weights within each cluster and generate the adjusted weight distribution.

[0112] In this embodiment, S4 includes the following specific steps:

[0113] For the variable path cluster, extract the number of paths and the magnitude of change within the cluster;

[0114] Based on the number and magnitude of changes in the paths within a cluster, calculate the proportional weights corresponding to the changing paths of electrical equipment within each cluster.

[0115] Sort the proportional weights and reset the weights of those below the average level to zero;

[0116] The adjusted weight distribution is obtained using the following formula:

[0117]

[0118] in, This represents the adjusted weight distribution of the r-th cluster. This represents the intra-cluster proportional weights calculated in the original dataset. This indicates the average weight level.

[0119] In this embodiment, after the dynamic change paths between devices are summarized into several typical change path clusters using a clustering algorithm, the system needs to further evaluate the contribution or importance of each change path cluster to the overall load fluctuation. This is because, although the clustering algorithm can identify different evolution patterns, not all patterns have a significant impact on the total power load. Some change path clusters may only represent the occasional behavior of a very small number of devices, or noise interference during the data acquisition process. In order to focus on the dominant patterns that have a decisive impact on system operation, this embodiment of the application will perform weight calculation and screening on each change path cluster.

[0120] Specifically, the system first counts the number of intra-cluster paths within each change path cluster, and the corresponding change magnitudes for these paths, which are represented by the total device power or fluctuation range. Based on these statistics, the proportional weights within each cluster are calculated. This weight Reflects the first The relative proportion of a path cluster within the entire set of strongly coupled links. For example, a path cluster containing a large number of high-power devices and exhibiting frequent fluctuations, its weight... The natural ratio will be higher. This is based on obtaining the original intra-cluster proportional weights. Subsequently, to simplify the subsequent analysis model and improve computational efficiency, this embodiment introduces an average-level filtering mechanism. The system calculates the average weight of all path clusters. This serves as the criterion for judgment. Subsequently, the weights of each cluster are... Compared to the average level Compare. If Below average This indicates that the influence of this path cluster is relatively small, and it may belong to a non-critical peripheral pattern. To avoid these weak signals interfering with the analysis of the main contradiction, the system adjusts its weight to zero influence, that is, sets it to zero. It equals 0. Conversely, if Higher than or equal to the average level If so, then retain its original weight value, that is, let equal Through this "removing the weak and retaining the strong" adjustment strategy, the system ultimately obtained the adjusted weight distribution. This distribution highlights the core evolutionary patterns with significant influence in the system, providing an optimized parameter basis for subsequently constructing an accurate load decomposition model.

[0121] S5. Generate a real-time input sequence based on the current total electricity consumption data and the adjusted weight distribution.

[0122] In this embodiment, after optimizing the weights of the coupling modes between devices, the system has a parameter system that reflects the core operating characteristics of the current power network. To apply these static or quasi-static parameters to dynamic real-time monitoring, this embodiment needs to obtain the latest adjusted weight distribution and the real-time collected total power consumption data. The current total power consumption data, as the system's input source, includes the real-time power superposition information of all online devices.

[0123] In this embodiment, to accurately track the real-time status of each device from this single total data set, this application incorporates non-intrusive load monitoring technology. NILM technology can infer the operating status of the internal load solely from voltage and current data at the bus end without installing sensors inside the user's premises. The system utilizes an adjusted weight distribution as prior knowledge to guide the NILM algorithm in performing feature matching and signal separation within complex total load waveforms.

[0124] In this embodiment, the system maps the adjusted weight distribution to the probabilistic model or neural network structure of the NILM algorithm, enhancing the algorithm's sensitivity to identifying devices corresponding to high-weight path clusters. Simultaneously, leveraging the high-frequency sampling capability of NILM technology, the system can capture microsecond-level transient changes in total power consumption data, such as inrush currents during device startup or back electromotive force during shutdown. By combining these transient features with the weight distribution, the system can track the power input changes of each key device in real time, thereby generating a time-continuous real-time input sequence after tracking. This sequence not only records the real-time power values ​​of each device but also includes the switching times of its operating states, providing a high-precision dynamic data stream for subsequent load decomposition and energy efficiency analysis.

[0125] S6. Based on the preset business classification attributes, integrate the real-time input sequence to determine the fused input sequence.

[0126] In this embodiment, S6 includes the following specific steps:

[0127] Obtain the real-time input sequence and extract the data dimensions related to the preset classification attributes from the real-time input sequence;

[0128] The data in the real-time input sequence is grouped according to data dimensions to generate categorized data subsets;

[0129] The categorized data subsets are weighted and combined with multi-dimensional data to generate an integrated data structure;

[0130] Based on the integrated data structure, a fused input sequence is generated.

[0131] In this embodiment, multi-dimensional data is collected by a non-invasive detection device.

[0132] In this embodiment, after successfully tracking the real-time input sequences of each device through non-intrusive monitoring technology, the system possesses micro-level device operation data. However, this data is often discrete and lacks business semantics. To transform this technical data into information of practical value to users or power grid managers, this embodiment introduces the concepts of preset business classification attributes and dynamic aggregation. Preset business classification attributes refer to labeling devices based on their purpose, energy consumption characteristics, or region, such as "lighting system," "HVAC," or "production line A." Dynamic aggregation refers to logically merging and summarizing the real-time input data of multiple devices belonging to the same category based on actual business needs. Specifically, the system first reads a preset device business classification attribute table, mapping each tracked device ID to its corresponding business category. Subsequently, using multi-dimensional energy component data (such as reactive power, power factor, harmonic content, etc.) provided by non-intrusive monitoring technology, the system performs multi-dimensional verification and fusion of device data within the same business category. For example, when aggregating the energy consumption of the "lighting system," the system not only adds up the active power of each lamp but also comprehensively considers changes in its power factor to ensure the accuracy of the aggregation result. Based on the rules of dynamic aggregation, the system performs weighted summation or statistical analysis on the real-time input data of all devices belonging to the same business category within the same time slice, generating a new sequence. This sequence is no longer a record of a single device's operation but represents the overall energy consumption behavior of a specific business segment. Finally, the system outputs the fused real-time input sequence. This sequence, business-oriented, clearly displays the real-time energy consumption dynamics of different functional areas or systems, providing intuitive and high-value data support for subsequent refined energy management, cost accounting, and energy-saving strategy formulation.

[0133] S7. Based on the fused input sequence, update the component proportion model and generate the optimized component proportion.

[0134] In this embodiment, S7 includes the following specific steps:

[0135] The total load is obtained through non-invasive monitoring equipment, and the total load is decoupled into multiple sub-indicators according to preset rules;

[0136] Time series analysis is performed on the fused input sequence to extract the dynamic change characteristics of each sub-indicator;

[0137] Based on the dynamic change characteristics and the predicted value of the total load, a component proportion model is constructed;

[0138] Perform deviation analysis on the component proportion model and output the deviation value;

[0139] Compare the deviation value with the preset limit to determine whether iterative optimization is triggered;

[0140] If iterative optimization is triggered, the optimization algorithm is used to iteratively process the parameter values ​​until the deviation value converges within the preset limit, then training stops, the component proportion model is updated, and the optimized component proportion is output.

[0141] In this embodiment, the fused input sequence reflects the electricity consumption distribution characteristics under the service category.

[0142] In this embodiment, the parameter values ​​represent the proportional distribution of each sub-item data.

[0143] In this embodiment, after obtaining the real-time input sequence that integrates business semantics and dynamic aggregation features, the system has sufficient data foundation to finely decompose the overall load. To achieve this goal, this embodiment needs to establish and maintain a component proportion model. This model aims to describe the proportional relationship of each business segment or key equipment in the total load and is the core logical carrier of the load decomposition algorithm. As time goes by and the operating status of equipment changes, fixed model parameters are often difficult to adapt to dynamically changing load characteristics. Therefore, it is necessary to continuously update the model using the latest fused real-time input sequence. Specifically, the system will use the fused real-time input sequence as training samples or calibration data and input it into the component proportion model. The model will recalculate the theoretical proportion of each component (such as lighting, power, HVAC, etc.) in the total load based on the input sequence data. Subsequently, in order to evaluate the accuracy of the current model, this embodiment introduces a deviation judgment mechanism. The system will calculate the deviation value between the total load predicted by the model and the actual total load collected, or calculate the error between the component load inferred by the model and the known benchmark value (if any). If the deviation value in the model exceeds a preset limit, it indicates that the current model parameters are outdated or invalid and cannot accurately reflect the actual operating state of the system. At this point, the system triggers an iterative optimization process. Through algorithms such as gradient descent, genetic algorithms, or particle swarm optimization, the system automatically adjusts the weight coefficients, regression parameters, or neural network thresholds in the model to minimize prediction deviation. After multiple iterations until the deviation value converges within the preset limit, the system finally determines the optimized component proportions. This proportion not only has extremely high real-time performance but has also been repeatedly verified with dynamic data, enabling it to accurately guide subsequent energy efficiency assessments and load management decisions.

[0144] S8. Use preset business verification attributes to associate the optimized item proportions and generate associated item proportions.

[0145] In this embodiment, the preset business verification attributes cover specific rules for electricity consumption scenarios.

[0146] In this embodiment, after determining the iteratively optimized component proportions, the system has a preliminary load decomposition framework. However, to further improve the reliability and practical application value of the proportions, this embodiment introduces preset business verification attributes. These preset attributes include historical reliability indicators of equipment, priority of business scenarios, or external environmental factors such as seasonal energy consumption patterns or peak-valley electricity price impacts. Through these preset business verification attributes, the system can perform secondary calibration and correlation analysis on the optimized component proportions. Specifically, the system incorporates the preset business verification attributes as additional constraints into the multi-dimensional power component data of non-intrusive monitoring. This component data covers various indicators such as active power, reactive power, and harmonic components. Simultaneously, combining the concept of dynamic aggregation, the system logically correlates and merges similar component proportions based on the guidance of the verification attributes. For example, if the preset business verification attributes indicate that a certain business segment (such as production equipment) has higher reliability during peak periods, the system will correspondingly increase its weight in the total proportion. Through this fusion process, the system not only retains the accuracy of the optimized component proportions but also enhances their correlation with actual business scenarios. Ultimately, the system outputs the correlated percentages for each item. These correlated percentages are no longer isolated values, but dynamic indicators embedded with business semantics, enabling better support for advanced applications such as energy auditing, fault diagnosis, and optimized scheduling.

[0147] S9. Based on the associated sub-item ratios, obtain multi-dimensional power sub-item data, and compare the multi-dimensional power sub-item data with historical accurate data through the verification module to generate final stable data.

[0148] In this embodiment, S9 includes the following specific steps:

[0149] The verification module is connected to a database that stores historical accurate data. Multi-dimensional electrical energy breakdown data is input into the verification module and compared with historical accurate data to obtain comparison results.

[0150] If the comparison results meet the preset verification rules, the multi-dimensional electrical energy sub-item data are deemed valid and are regarded as the final stable data output.

[0151] If the comparison results do not meet the preset verification rules, the multi-dimensional energy component data will be corrected, and the corrected multi-dimensional energy component data will be output as the final stable data.

[0152] In this embodiment, the proportions of the associated sub-items reflect the data distribution under the business rules.

[0153] In this embodiment, after the system determines the component proportions after preset business attribute verification and dynamic aggregation association, it enters the final data generation and verification stage. At this point, the system applies these high-precision proportion coefficients to the real-time collected total power load data, and uses mathematical operations to decompose the total load into specific electrical energy parameters for each independent component. This process is not merely a simple active power breakdown, but also encompasses multiple dimensions of electrical characteristics such as reactive power, effective current value, power factor, and harmonic content, thereby outputting multi-dimensional electrical energy component data. This data provides a detailed description of the precise operation of each business segment or key equipment at the current moment. In this embodiment, to further ensure the reliability of the output data and prevent data distortion caused by algorithmic local extrema or transient interference, this application introduces a dedicated verification module. This module is connected to a database storing historical accurate data, which may include manually verified baseline records, measured values ​​from high-precision instruments at specific times, or stable operating baselines verified through long-term statistical analysis. The system inputs the newly generated multi-dimensional electrical energy component data into the verification module for horizontal comparison and trend analysis with the historical accurate data. Specifically, the verification module calculates the deviation between the current data and historical benchmarks, and checks for logical contradictions or physical impossibilities, such as the appearance of motor starting characteristics in a lighting circuit. If the data passes the verification module's check, the system confirms its validity. If anomalies are found, the system may use smoothing filtering or weighted correction to fine-tune the data. Ultimately, the system outputs stable and accurate data. This data not only has extremely high numerical accuracy but also exhibits good stability over time, providing users with reliable energy billing, equipment health status assessments, and energy-saving potential analysis services, thereby achieving deep perception and precise management of electricity load.

[0154] Example 2

[0155] This embodiment provides a dynamic aggregation system for electricity component data based on non-intrusive monitoring, such as... Figure 2 As shown, it includes:

[0156] The time series analysis module acquires the total power consumption data and historical time series records of electrical equipment, extracts fluctuation feature data from the total power consumption data, and constructs the potential coupling patterns between various electrical equipment by combining the historical time series records.

[0157] The relation matrix construction module constructs a real-time relation matrix based on potential coupling patterns, analyzes the element values ​​in the real-time relation matrix, and generates a set of strongly coupled links.

[0158] The clustering and grouping module extracts dynamic change indicators from the set of strongly coupled links, groups the dynamic change indicators, and generates change path clusters.

[0159] The weight adjustment module calculates the proportional weights within each cluster for the changing path clusters and generates the adjusted weight distribution.

[0160] The real-time input tracking module generates a real-time input sequence based on the current total electricity consumption data and the adjusted weight distribution;

[0161] The real-time input fusion module integrates and processes the real-time input sequence according to the preset business classification attributes to determine the fused input sequence;

[0162] The component proportion update module updates the component proportion model based on the fused input sequence and generates the optimized component proportion.

[0163] The item ratio association module associates the optimized item ratios with preset business verification attributes to generate the associated item ratios.

[0164] The data output and verification module obtains multi-dimensional power energy sub-item data based on the associated sub-item ratios. The verification module compares the multi-dimensional power energy sub-item data with historical accurate data to generate the final stable data.

[0165] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this disclosure can be embodied in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0166] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0167] This invention is described with reference to flowchart illustrations and block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and block diagrams, as well as combinations of blocks in the flowchart illustrations and block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0168] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and boxes Figure 1 The steps of the function specified in one or more boxes.

[0169] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for dynamic aggregation of electrical energy component data based on non-intrusive monitoring, characterized in that: The specific steps include the following: Obtain total power consumption data and historical time series records of electrical equipment, extract fluctuation feature data from total power consumption data, and combine with historical time series records to construct potential coupling patterns between various electrical equipment. Based on the potential coupling patterns, a real-time relationship matrix is ​​constructed, the element values ​​in the real-time relationship matrix are analyzed, and a set of strongly coupled links is generated. Extract dynamic change indicators from the set of strongly coupled links, group the dynamic change indicators, and generate change path clusters; For each cluster of changing paths, the proportional weights within each cluster are calculated, and the adjusted weight distribution is generated. Based on the current total electricity consumption data and the adjusted weight distribution, a real-time input sequence is generated; The real-time input sequence is integrated and processed according to the preset business classification attributes to determine the fused input sequence; Based on the fused input sequence, update the component proportion model to generate the optimized component proportion; The optimized item proportions are associated with preset business verification attributes to generate associated item proportions. Based on the associated sub-item ratios, multi-dimensional electricity sub-item data is obtained. The verification module compares the multi-dimensional electricity sub-item data with historical accurate data to generate the final stable data.

2. The method for dynamic aggregation of power component data based on non-intrusive monitoring according to claim 1, characterized in that: The potential coupling patterns between the various electrical devices include: Total power consumption data is collected through non-intrusive monitoring equipment, and historical time-series records are obtained from the database. The historical time-series records include total load fluctuations and equipment operation tags. A time-series analysis algorithm is used to decompose the total electricity consumption data and extract fluctuation characteristic data; Based on the fluctuation characteristics of each electrical device within a preset time period, the total load fluctuation, and the equipment operation tags, the element values ​​of the relationship matrix are calculated using the following formula: ； in, The element values ​​of the relation matrix represent the potential coupling patterns between devices. and Indicates different device indexes. Indicates equipment In time Fluctuation characteristic data, Indicates equipment In time Fluctuation characteristic data, Indicates the duration of historical records. Indicates equipment The average value, Indicates equipment The average value; Based on the element values ​​of the relation matrix, construct the potential coupling patterns between various electrical devices.

3. The method for dynamic aggregation of electrical energy component data based on non-intrusive monitoring according to claim 1, characterized in that: The generated set of strongly coupled links includes: Extract element values ​​from the real-time relationship matrix; Compare the element value with a preset threshold; When an element value is greater than a preset threshold, it is marked, and inter-device links are generated, forming a set of strongly coupled links.

4. The method for dynamic aggregation of electrical energy component data based on non-invasive monitoring according to claim 1, characterized in that: The generated change path cluster includes: Extract multiple dynamic change indicators from a set of strongly coupled links; The formula for calculating the distance between multiple dynamically changing indicators is: ； in, Indicates the path of change and Distance indicators between Representing a path The A dynamic indicator, Representing a path The A dynamic indicator, Indicates the dimension of the indicator; Based on the distance index, a clustering algorithm is applied to group the changing paths, generating multiple changing path clusters.

5. The method for dynamic aggregation of electrical energy component data based on non-invasive monitoring according to claim 1, characterized in that: The generation of the adjusted weight distribution includes: For the variable path cluster, extract the number of paths and the magnitude of change within the cluster; Based on the number and magnitude of changes in the paths within a cluster, calculate the proportional weights corresponding to the changing paths of electrical equipment within each cluster. Sort the proportional weights and reset the weights of those below the average level to zero; The adjusted weight distribution is obtained using the following formula: ； in, This represents the adjusted weight distribution of the r-th cluster. This represents the intra-cluster proportional weights calculated in the original dataset. This indicates the average weight level.

6. The method for dynamic aggregation of electrical energy component data based on non-intrusive monitoring according to claim 1, characterized in that: The determination of the fused input sequence includes: Obtain the real-time input sequence and extract the data dimensions related to the preset classification attributes from the real-time input sequence; The data in the real-time input sequence is grouped according to data dimensions to generate categorized data subsets; The categorized data subsets are weighted and combined with multi-dimensional data to generate an integrated data structure; Based on the integrated data structure, a fused input sequence is generated.

7. The method for dynamic aggregation of electrical energy component data based on non-intrusive monitoring according to claim 1, characterized in that: The generated optimized item proportions include: The total load is obtained through non-invasive monitoring equipment, and the total load is decoupled into multiple sub-indicators according to preset rules; Time series analysis is performed on the fused input sequence to extract the dynamic change characteristics of each sub-indicator; Based on the dynamic change characteristics and the predicted value of the total load, a component proportion model is constructed; Perform deviation analysis on the component proportion model and output the deviation value; Compare the deviation value with the preset limit to determine whether iterative optimization is triggered; If iterative optimization is triggered, the optimization algorithm is used to iteratively process the parameter values ​​until the deviation value converges within the preset limit, then training stops, the component proportion model is updated, and the optimized component proportion is output.

8. The method for dynamic aggregation of power component data based on non-intrusive monitoring according to claim 1, characterized in that: The generation of final stable data includes: Multi-dimensional electrical energy breakdown data is input into the verification module and compared with historical accurate data to obtain the comparison results; If the comparison results meet the preset verification rules, the multi-dimensional electrical energy sub-item data are deemed valid and are regarded as the final stable data output. If the comparison results do not meet the preset verification rules, the multi-dimensional energy component data will be corrected, and the corrected multi-dimensional energy component data will be output as the final stable data.

9. The system for dynamic aggregation of power component data based on non-intrusive monitoring as described in claim 1, characterized in that: include: The time series analysis module acquires the total power consumption data and historical time series records of electrical equipment, extracts fluctuation feature data from the total power consumption data, and constructs the potential coupling patterns between various electrical equipment by combining the historical time series records. The relation matrix construction module constructs a real-time relation matrix based on potential coupling patterns, analyzes the element values ​​in the real-time relation matrix, and generates a set of strongly coupled links. The clustering and grouping module extracts dynamic change indicators from the set of strongly coupled links, groups the dynamic change indicators, and generates change path clusters. The weight adjustment module calculates the proportional weights within each cluster for the changing path clusters and generates the adjusted weight distribution. The real-time input tracking module generates a real-time input sequence based on the current total electricity consumption data and the adjusted weight distribution; The real-time input fusion module integrates and processes the real-time input sequence according to the preset business classification attributes to determine the fused input sequence; The component proportion update module updates the component proportion model based on the fused input sequence and generates the optimized component proportion. The item ratio association module associates the optimized item ratios with preset business verification attributes to generate the associated item ratios. The data output and verification module obtains multi-dimensional power energy sub-item data based on the associated sub-item ratios. The verification module compares the multi-dimensional power energy sub-item data with historical accurate data to generate the final stable data.

Images