Power marketing dynamic data acquisition method based on load adaptation

Abstract

The invention belongs to the technical field of power marketing data acquisition, and particularly relates to a power marketing dynamic data acquisition method based on load adaptation. Load self-adaption is achieved by counting the number of data collection tasks in real time and evaluating the system resource utilization rate. In a high-load state, the system is automatically switched to a sampling mode of zero crossing point feature windowing, and a dual-channel verification mechanism is started. And different data transmission mechanisms are selected according to the load state. And after receiving the data, the platform performs digital simulation comparison, and triggers phase calibration and resampling when the deviation exceeds the standard. When a plurality of nodes are continuously in a high-load state, the system starts a zero crossing point characteristic window shunting technology and dynamically adjusts a transmission priority according to an equivalent loss value. Intelligent and efficient power marketing data acquisition is realized, and remarkable economic benefits and technical advancement are achieved.

Description

Technical Field

[0001] This invention belongs to the field of power marketing data acquisition technology, specifically relating to a dynamic power marketing data acquisition method based on load adaptation. Background Technology

[0002] In the field of power information acquisition technology, with the deepening of smart grid construction and the increasing demand for refined power market management, power marketing data acquisition faces multi-dimensional challenges. On the one hand, data acquisition in the distribution network environment is easily affected by factors such as load fluctuations and signal interference, resulting in a large amount of duplicate, erroneous, or invalid information in the raw data. Traditional data cleaning methods are unable to accurately identify abnormal data in complex scenarios, affecting the accuracy of subsequent fault analysis. On the other hand, existing technologies for judging power consumption faults are mostly based on single-dimensional data, lacking comprehensive analysis of power consumption efficiency and dynamic changes in circuit parameters, leading to ambiguous fault location and an inability to quickly pinpoint the specific fault type. In addition, when the system is under high load, traditional acquisition methods often cause resource contention due to fixed sampling frequency and full data transmission mode, resulting in data transmission delays or losses, making it difficult to meet real-time monitoring requirements.

[0003] For example, CN118445553B provides a method and system for collecting electricity marketing data. Although it constructs a framework for collecting electricity marketing data through data cleaning, electricity efficiency analysis and parameter verification, it has significant shortcomings in terms of adaptability to complex scenarios, multi-dimensional fault analysis and dynamic load management. Summary of the Invention

[0004] To address the aforementioned shortcomings, this invention proposes a load-adaptive dynamic data acquisition method for electricity marketing. By dynamically assessing the system load status and intelligently adjusting the sampling mode, transmission mechanism, and anomaly detection strategy, the efficiency, accuracy, and system resource utilization of electricity marketing data acquisition are improved, effectively solving the deficiencies of traditional methods under load fluctuations.

[0005] The specific technical solution of the present invention is as follows:

[0006] One of the technical solutions of this invention is to provide a method for dynamic data collection in electricity marketing based on load adaptation, comprising:

[0007] Deploy load monitoring modules at the distribution network terminals to count the number of data acquisition tasks in real time and dynamically assess the load status in conjunction with system resource utilization.

[0008] When the system is under high load, the system automatically switches to the sampling mode of zero-crossing feature windowing;

[0009] Build a power consumption anomaly feature library; when the system is under high load, if an anomaly pattern matching the power consumption anomaly feature library is detected in the zero-crossing feature window, the system automatically increases the sampling frequency of that node to the millisecond level and starts a dual-channel verification mechanism.

[0010] The data transmission mechanism is selected based on the load status determined by the system. Under high load, active pull transmission is used, and the terminal only uploads the feature window compressed data and the equivalent loss value. Under standard load or low load, broadcast full transmission is switched, but duplicate data is avoided by deduplication through flag bits.

[0011] After receiving the data uploaded by the terminal, the platform starts the digital simulation comparison process; if the deviation continues to exceed the standard, it sends an instruction to the terminal to trigger the phase calibration of the zero-crossing feature window and resamples.

[0012] When the system detects that multiple nodes are under high load, it automatically enables zero-crossing feature window splitting technology. At the same time, the transmission priority is dynamically adjusted by the equivalent loss value.

[0013] As a further option of the present invention, the dynamic evaluation of load status includes:

[0014] Constructing a mathematical model for the load adaptability index LAI:

[0015] LAI = ω1·T + ω2·R + ω3·N;

[0016] Where N represents the number of currently active tasks, R represents the system resource utilization rate, T represents the network latency or task response time, and ω1, ω2, and ω3 represent weighting coefficients.

[0017] The load adaptive trigger threshold (ATT) is dynamically adjusted using a sliding time window algorithm, and the ATT value is increased during peak electricity consumption periods.

[0018] The normalization methods for the current number of active tasks, system resource utilization, and network latency or task response time are as follows:

[0019] Task quantity normalization: Let the maximum number of tasks be N. max Then the normalized number of tasks is N′=N / N max ;

[0020] Resource utilization rate normalization: Let the upper limit of resource utilization rate be R. max Then the normalized resource utilization rate is R′=R / R max ;

[0021] Network latency or task response time normalization: Let the maximum allowable latency be T. max Then the normalized network delay is T′=T / Tmax .

[0022] As a further option of the present invention, the dynamic evaluation of load status also includes: comparing the calculated LAI with ATT to classify the load status;

[0023] The logic for classifying load states is as follows:

[0024] Compare the calculated LAI with ATT to determine the current load status:

[0025] If LAI < ATT HLS This indicates a high load condition.

[0026] If ATT SLS ≤LAI≤ATT HLS This is determined to be a standard load condition;

[0027] If LAI < ATT SLS This indicates a low load state.

[0028] Among them, ATT HLS ATT adapts to the trigger threshold for high-load conditions. SLS Low load state load adaptive trigger threshold.

[0029] As a further option of the present invention, the sampling mode of the zero-crossing feature windowing includes:

[0030] The signal is preprocessed using a digital filter, and then combined with a zero-crossing detection algorithm or a hardware comparator circuit to achieve microsecond-level zero-crossing detection.

[0031] Based on the timing optimization strategy, the sampling window is limited to the abnormal feature range of ±1ms to ±2ms from the zero crossing point;

[0032] Power quality anomaly features within a window can be extracted using wavelet transform, Hilbert-Huang transform, or convolutional neural networks.

[0033] As a further option of the present invention, the sampling mode of the zero-crossing feature windowing also includes:

[0034] The power quality feature windows are dynamically prioritized and sorted based on a weighted score of the intensity, duration and scope of the anomaly.

[0035] As a further option of the present invention, the construction of the power consumption anomaly feature library includes: the feature library is designed as a three-layer structure, with the top layer being the anomaly category, the middle layer being the specific anomaly mode, and the bottom layer being the feature parameters;

[0036] The specific anomaly patterns are classified into templates using clustering algorithms, and feature parameters are extracted from historical data using Fourier transform or wavelet transform.

[0037] The matching adopts a hierarchical matching strategy: primary matching calculates similarity through dynamic time warping, and when the similarity is >85%, it enters advanced matching; advanced matching uses a deep feature matching network to extract abstract features and calculate cosine similarity; in the decision mechanism, a comprehensive similarity threshold of 0.92 is set, and if the threshold is exceeded, it is determined to be a valid match.

[0038] As a further option of the present invention, the dual-channel verification mechanism includes:

[0039] Channel 1: Employs high-precision ADC and GPS time synchronization technology to capture raw instantaneous current / voltage values;

[0040] Channel 2: Equivalent loss value calculated based on thermal path model:

[0041] The formula for the thermal circuit model is:

[0042]

[0043] Among them, P loss K represents the equivalent loss value. T Represents the temperature compensation coefficient, I(t) represents the instantaneous current value, and R... th denoted by , V(t) represents the instantaneous voltage value, tanδ represents the dielectric loss factor, and t0 and t1 represent the start and end times of the zero-crossing characteristic window.

[0044] The dual-channel verification mechanism is implemented as follows:

[0045] When the dual-channel verification results show that the equivalent loss value of a node exceeds the preset threshold, the system needs to mark it as a high-risk node and upload data first.

[0046] As a further option of the present invention, the data transmission mechanism satisfies:

[0047] When using active pull-based transmission, the upload priority is set according to the equivalent loss value.

[0048] During broadcast full transmission, unique identifiers are used to deduplicate flag bits.

[0049] As a further option of the present invention, the digital simulation comparison process includes:

[0050] (a) Reconstructing the power grid topology, load distribution, and multiphysics coupling scenario using a finite element simulation model;

[0051] (b) Calculate the equivalent loss value P uploaded by the terminal. lossThe theoretical equivalent loss value P′ calculated by the simulation model loss The deviation; the calculation formula is:

[0052]

[0053] Where D represents the percentage deviation, P loss P′ represents the equivalent loss value uploaded by the terminal. loss This represents the theoretical equivalent loss value calculated by the simulation model.

[0054] (c) Based on the percentage of deviation, it is divided into different levels;

[0055] (d) If the deviation is severe and persistent, a command is sent to the terminal to trigger phase calibration with zero-crossing feature windowing and to resample;

[0056] The phase calibration algorithm is implemented using the least squares method or a Kalman filter, and the system calculates the calibration value using the following formula:

[0057]

[0058] Where, Δ t t represents the sampling window offset. i t represents the actual sampling time of the terminal device. i ′ represents the theoretical sampling time, and N represents the number of sampling points;

[0059] Among them, after bit calibration, resampled data is uploaded through an incremental transmission strategy.

[0060] As a further option of the present invention, the zero-crossing feature window splitting technique includes:

[0061] The discretized sampling window offset is assigned to consecutive high-load nodes, and the calculation formula is as follows:

[0062] T offset,i =ΔT·(i-1);

[0063] Among them, T offset,i ΔT represents the sampling window offset of the i-th node, ΔT represents the minimum time interval between adjacent nodes, and i represents the node number;

[0064] Transmission priority is classified according to equivalent loss value:

[0065] High priority: Equivalent loss value > 1.5 times the normal range;

[0066] Medium priority: equivalent loss value ∈ [1.2, 1.5] times the normal range;

[0067] Low priority: equivalent loss value < 1.2 times the normal range.

[0068] The beneficial effects of the technical solutions provided in this application include at least the following:

[0069] Through real-time load monitoring and dynamic adjustment strategies, the system automatically switches to windowed sampling mode under high load, reducing task processing pressure, reducing resource waste by 20-30%, and significantly improving the operational stability of terminal devices.

[0070] By employing the "zero-crossing feature windowing" technique, sampling is concentrated in the high-variability range of ±1ms to ±2ms near the zero-crossing point. Combined with timing optimization strategies, redundant data collection is reduced by 40-50%, while ensuring that the capture rate of power quality anomaly features reaches over 95%.

[0071] A three-tiered power consumption anomaly feature database is constructed. Through a hierarchical matching strategy and dual-channel verification, the anomaly identification accuracy is improved to 95%, effectively eliminating the risk of missed detections. Based on a dynamic transmission selection mechanism according to load status and combined with priority ranking of equivalent loss values, network bandwidth usage is reduced by 35%-40%.

[0072] When multiple nodes are under high load, window splitting technology is enabled, along with dynamic priority based on equivalent loss value, which increases system throughput by 25% and reduces resource contention latency by 50%.

[0073] In summary, this method achieves three major breakthroughs: a leap in resource efficiency, a breakthrough in decision-making accuracy, and enhanced system scalability. With zero-crossing feature windowing as its core, combined with a dual-channel verification and simulation calibration mechanism for the thermal circuit model, it provides a highly robust dynamic data acquisition framework for the power marketing system, effectively supporting key operations such as anti-theft and equipment health management, and is estimated to reduce overall operation and maintenance costs by 20%-30%. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the overall process of a load-adaptive dynamic data collection method for electricity marketing.

[0075] Figure 2 A detailed flowchart of the S100 steps of the load-adaptive dynamic data acquisition method for electricity marketing;

[0076] Figure 3 A detailed flowchart of the S200 steps of the load-adaptive dynamic data acquisition method for electricity marketing;

[0077] Figure 4 A detailed flowchart of the S300 steps for the load-adaptive dynamic data acquisition method for electricity marketing;

[0078] Figure 5 A detailed flowchart of the steps in the S4500 method for dynamic data acquisition in power marketing based on load adaptation;

[0079] Figure 6 A detailed flowchart of the steps in the load-adaptive dynamic data acquisition method for electricity marketing, S600. Detailed Implementation

[0080] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0081] In the field of power marketing, distribution network data acquisition faces challenges such as large load fluctuations and complex data volumes. Traditional data acquisition methods typically employ fixed sampling frequencies and full data transmission modes. Under high system loads, this easily leads to resource contention and data transmission delays, severely impacting the efficiency and accuracy of detecting abnormal electricity consumption. Furthermore, traditional methods cannot dynamically optimize sampling and transmission strategies based on system load conditions, resulting in wasted system resources and low data processing efficiency. To address these issues, please refer to [link to relevant documentation]. Figure 1 This illustrates a load-adaptive-based dynamic data acquisition method for electricity marketing provided by an embodiment of the present invention, the method comprising:

[0082] S100: Deploy a load monitoring module at the distribution network terminal to count the number of data acquisition tasks currently in operation in real time, and dynamically assess the load status in conjunction with the system resource utilization rate.

[0083] S200: When the system is under high load, the system automatically switches to the zero-crossing feature windowing sampling mode.

[0084] S300: Construct an abnormal power consumption feature library; when the system is under high load, if an abnormal pattern matching the abnormal power consumption feature library is detected in the zero-crossing feature window, the system automatically increases the sampling frequency of the node to the millisecond level and starts a dual-channel verification mechanism.

[0085] S400: Selects the data transmission mechanism based on the load status determined by the system; when under high load, it adopts active pull transmission, and the terminal only uploads the feature window compressed data and the equivalent loss value; when under standard load or low load, it switches to broadcast full transmission, but avoids duplicate data by deduplication through flag bits.

[0086] S500: After receiving the data uploaded by the terminal, the platform starts the digital simulation comparison process; if the deviation continues to exceed the standard, it sends an instruction to the terminal to trigger the phase calibration of the zero-crossing feature window and resample.

[0087] S600: When the system detects that multiple nodes are under high load, it automatically enables zero-crossing feature window splitting technology. At the same time, the transmission priority is dynamically adjusted by the equivalent loss value.

[0088] The S100-S600 series achieves intelligent and efficient power marketing data acquisition through a load-adaptive dynamic data acquisition strategy, resulting in significant economic benefits and technological advancements. It is suitable for power marketing data acquisition and monitoring scenarios in various distribution network environments.

[0089] The specific plan is as follows:

[0090] In the load-adaptive dynamic data acquisition method for power marketing, the S100 deploys a load monitoring module at the distribution network terminal to achieve real-time statistics on the number of data acquisition tasks currently in operation, and performs dynamic evaluation in conjunction with system resource utilization. This process includes hardware deployment and initialization, real-time task count statistics, system resource utilization monitoring, calculation of load adaptability indicators, and comparison with preset thresholds to determine the load status. This provides a scientific basis for subsequently adjusting the sampling mode and data transmission mechanism according to the load status, ensuring that the system can operate efficiently and optimize resource allocation under different load conditions.

[0091] Please refer to Figure 2 The diagram illustrates a flowchart of an exemplary load-adaptive dynamic data acquisition method for electricity marketing, S100, which includes:

[0092] S110: Design the hardware architecture of the distribution network terminal load monitoring module to count the number of data acquisition tasks in real time.

[0093] Deploying load monitoring modules at the distribution network terminals is the first step in implementing a load-adaptive data acquisition method. The module's hardware architecture comprehensively considers real-time performance, reliability, and scalability to ensure accurate capture of dynamic changes in the power system. The core components of the load monitoring module include a high-performance processor, multi-channel sensor interfaces, a high-precision clock synchronization unit, and a communication module.

[0094] In one possible implementation, the high-performance processor employs a multi-core embedded processor, the multi-channel sensor interface supports the access of various types of sensors, the high-precision clock synchronization unit achieves microsecond-level time synchronization via GPS or PTP, and the communication module integrates wired and wireless communication interfaces to support efficient data transmission and remote management.

[0095] In order to dynamically assess the system load, it is necessary to perform real-time statistics on the number of data acquisition tasks currently in operation.

[0096] In one possible implementation, the method for counting the number of data acquisition tasks currently in operation includes: storing all data acquisition tasks in a priority queue, with each task containing a task ID, task type, task priority, task start time, and resource usage information. The task queue is periodically polled to count the number of currently active tasks.

[0097] S120: Dynamically evaluate system resource utilization.

[0098] System resource utilization is a crucial indicator for assessing load status, including CPU utilization, memory utilization, network bandwidth utilization, and storage space utilization. Dynamic evaluation of these indicators helps understand the system's operational pressure.

[0099] In one possible implementation, CPU utilization monitoring assesses CPU utilization by periodically sampling the CPU core's operating frequency and load ratio. Memory utilization is measured by monitoring the memory allocation and deallocation process. Network bandwidth utilization is calculated by statistically analyzing the amount of data sent and received per unit time. Storage space utilization monitoring uses file system tools to obtain the total capacity and used capacity of the storage device to calculate storage space utilization.

[0100] S130: Construct a mathematical model to transform resource usage data from various dimensions into a unified load adaptability metric.

[0101] The Load Adaptability Index (LAI) is a core parameter for measuring system load status, and its calculation formula is as follows:

[0102] LAI = ω1·T + ω2·R + ω3·N;

[0103] Where N represents the number of currently active tasks, R represents the system resource utilization rate, which is a weighted average considering CPU, memory, network and storage, T represents network latency or task response time, and ω1, ω2, ω3 represent weighting coefficients, which are adjusted according to the actual application scenario.

[0104] Based on the load adaptability index calculation formula, specifically, during the calculation:

[0105] Task quantity normalization: The number of tasks N is normalized to the interval [0,1] for easier comparison with other metrics. In one possible implementation, the task quantity normalization is calculated as follows: Let the maximum number of tasks be N. max Then the normalized number of tasks is N′=N / N max .

[0106] Resource utilization rate normalization: Normalize the resource utilization rate R to the interval [0,1]. In one possible implementation, the resource utilization rate normalization is calculated as follows: Let the upper limit of resource utilization rate be R. max Then the normalized resource utilization rate is R′=R / R max .

[0107] Network latency normalization: Normalizing the network latency T to the interval [0,1]. In one possible implementation, network latency normalization is calculated as follows: Let the maximum allowable latency be T. max Then the normalized network delay is T′=T / T max .

[0108] S140: Compare the calculated load adaptability index with the preset load adaptability trigger threshold to determine the current system load status.

[0109] The Adaptive Trigger Threshold (ATT) is a key parameter for determining the system load status. Based on a comparison between the load adaptability index and the preset adaptive trigger threshold, the load status is divided into High Load State (HLS), Standard Load State (SLS), and Low Load State (LLS).

[0110] In one possible implementation, ATT sets the load adaptability index threshold range for high load, standard load, and low load states based on historical data and system performance test results. For example, for high load states, ATT is set to greater than 0.8; for standard load states, ATT is set to 0.5-0.8; and for low load states, ATT is set to less than 0.5.

[0111] In one possible implementation, ATT is not a fixed value, but is dynamically adjusted based on historical data and real-time operating conditions. For example, during holidays or peak electricity consumption periods, the ATT value under high load conditions can be appropriately increased to cope with sudden increases in load.

[0112] Compare the calculated LAI with ATT to determine the current load status:

[0113] If LAI < ATT HLS This indicates a high load condition.

[0114] If ATT SLS ≤LAI≤ATT HLS This is determined to be a standard load condition;

[0115] If LAI < ATT SLS It is determined to be in a low load state.

[0116] In the load-adaptive dynamic data acquisition method for power marketing, when the system is under high load, in order to effectively reduce resource consumption and ensure the integrity of key data, step S200 proposes a sampling mode based on "zero-crossing feature windowing". This mode integrates the idea of ​​zero-crossing detection timing optimization with sensor load judgment logic, and only captures data from the power quality feature window near the zero-crossing point, thereby avoiding stable periodic data and reducing unnecessary data acquisition tasks.

[0117] Please refer to Figure 3 The diagram illustrates a flowchart of an exemplary load-adaptive dynamic data acquisition method for electricity marketing, S200, which includes:

[0118] S210: Identifies and enters a high-load state, and the system automatically switches to the "zero-crossing feature windowing" sampling mode.

[0119] The system needs to monitor the current load status in real time and adjust the load based on LAI and ATT. HLS A comparison is performed to determine whether a high-load state has been entered. When the Load Adaptability Index (LAI) exceeds the preset high-load state trigger threshold (ATT_HLS), the system automatically enters a high-load state. This is a prerequisite for switching to the "zero-crossing feature windowing" sampling mode when determining whether a high-load state has been entered.

[0120] Under high load conditions, the core of the "zero-crossing feature windowing" sampling mode is to accurately detect the zero-crossing points of voltage and current signals and combine them with timing optimization strategies to capture data only for key power quality feature windows.

[0121] The zero-crossing detection principle refers to the moment when a voltage or current signal transitions from the positive half-cycle to the negative half-cycle or vice versa.

[0122] In one possible implementation, a digital filter is used to preprocess the original signal to remove noise interference. Then, a zero-crossing detection algorithm is used to identify the specific time of the zero-crossing.

[0123] As another embodiment of this step, in order to improve detection accuracy, hardware auxiliary circuits, such as comparator circuits, are introduced in conjunction with the zero-crossing detection algorithm to ensure that the time resolution of zero-crossing detection reaches the microsecond level.

[0124] The timing optimization strategy is based on zero-crossing detection results, limiting the sampling time window to a specific range near the zero-crossing point. The goal of timing optimization is to reduce the collection of unnecessary data, thereby reducing system resource consumption. For example, during a steady period, voltage and current signals typically change little and do not require frequent sampling; however, near the zero-crossing point, signal changes are drastic and may exhibit abnormal characteristics, thus requiring focused sampling.

[0125] S220: Define and divide the power quality feature window in the "zero-crossing feature windowing" sampling mode.

[0126] The power quality characteristic window refers to the time interval in which significant abnormal characteristics exist near the zero-crossing point. These include, but are not limited to, voltage distortion, harmonic abrupt changes, and frequency shifts.

[0127] In one possible implementation, the system identifies anomalous components in the signal through spectral analysis and quantifies the intensity of the anomalous features using statistical methods. For example, if the harmonic content within a certain time period exceeds a threshold, it is marked as a feature window.

[0128] The feature window is dynamically adjusted based on the distribution pattern of abnormal features. For example, for voltage distortion, the system divides the window into several sub-windows of ±1ms, ±2ms, etc., corresponding to different anomaly levels.

[0129] In one possible implementation, the system uses a sliding window technique to continuously monitor the signal and updates the boundary of the feature window in real time using an anomaly detection algorithm.

[0130] As another embodiment of this step, in the "zero-crossing feature windowing" sampling mode, different feature windows have different importance, so they need to be prioritized. For example, windows containing harmonic abrupt changes have higher priority than windows with only slight voltage fluctuations.

[0131] In one possible implementation, priority ranking is achieved through a weighted scoring mechanism, with scoring factors including the intensity, duration, and scope of influence of the anomalous features.

[0132] S230: Capture abnormal electricity consumption characteristic data within the feature window.

[0133] Within the feature window, the system needs to capture abnormal power consumption feature data to ensure that the collected information reflects the actual power quality problems.

[0134] The system employs multiple algorithms to analyze the data within the feature window and extract key anomaly features.

[0135] In one possible implementation, the signal is decomposed by wavelet transform to identify local anomalies and extract key anomaly features; the nonlinear signal is analyzed by Hilbert-Huang transform to extract key anomaly features.

[0136] As another embodiment of this step, the system combines deep learning models, such as convolutional neural networks, to classify and identify data within the feature window, thereby improving the accuracy of anomaly detection.

[0137] In the load-adaptive dynamic data acquisition method for electricity marketing, S300 identifies and processes power quality anomalies under high load conditions by constructing a power consumption anomaly feature library, and improves the accuracy of data acquisition and transmission by combining a dual-channel verification mechanism.

[0138] Please refer to Figure 4 The diagram illustrates a flowchart of an exemplary load-adaptive dynamic data acquisition method for electricity marketing, S300, which includes:

[0139] S310: Construct a database of abnormal electricity consumption characteristics.

[0140] In power systems, abnormal electricity consumption typically manifests as patterns such as electricity theft waveforms, equipment aging harmonics, and voltage distortion. To accurately detect abnormal electricity consumption, a database of abnormal electricity consumption features is constructed.

[0141] Feature library design principles include:

[0142] The top layer represents anomaly categories, and the feature library covers a variety of typical anomaly categories, such as electricity theft waveforms and equipment aging harmonics.

[0143] The intermediate layer represents specific anomaly patterns, further refined to the anomaly category, such as electricity theft waveforms that bypass metering devices, low-frequency harmonics, etc.

[0144] The bottom layer contains feature parameters, which store the specific feature parameters for each abnormal mode, including waveform shape, frequency distribution, amplitude variation, etc.

[0145] In one possible implementation, the underlying feature parameters employ signal processing techniques, such as Fourier transform and wavelet transform, to extract features of various anomaly patterns from historical data. For each specific anomaly pattern, its key parameters, such as harmonic content, voltage amplitude variation range, and frequency offset amplitude, are recorded and stored as feature vectors.

[0146] As another embodiment of this step, in order to improve matching efficiency, specific abnormal patterns in the feature library can be classified by clustering algorithms to form templates for different abnormal categories.

[0147] S320: Under high load conditions, the system initiates real-time stream matching on the captured zero-crossing feature window.

[0148] Under high load conditions, the system needs to monitor the data within the zero-point feature window in real time and match it with templates in the power consumption anomaly feature library to identify potential anomaly patterns.

[0149] In one possible implementation, a hierarchical matching strategy is adopted: first, similar waveforms are quickly screened through dynamic time warping (DTW), then a precise comparison is performed based on a deep feature matching network, and finally, a similarity score with the feature library template is output.

[0150] Specifically, the tiered matching process includes:

[0151] Primary matching: Calculate the DTW distance between the current window data and the baseline template in the feature library. When the similarity is >85%, proceed to the fine matching stage.

[0152] Advanced matching: Extracts abstract features using a pre-trained deep feature matching network, calculates cosine similarity, and weights and fuses the DTW results.

[0153] Decision-making mechanism: Set a comprehensive similarity threshold of 0.92. If the similarity exceeds the threshold, it is determined to be a valid match and triggers subsequent actions.

[0154] When an abnormal pattern matching the power consumption anomaly feature library is detected within the zero-crossing feature window, the system needs to automatically increase the sampling frequency of that node to the millisecond level in order to capture more detailed abnormal features.

[0155] For example, under high load, the default sampling frequency is 100ms / time to reduce resource consumption. However, when an abnormal pattern is detected, the system needs to dynamically adjust the sampling frequency to the millisecond level. The adjustment strategy is implemented in stages according to the severity of the abnormal pattern. For minor anomalies, such as small voltage fluctuations, the sampling frequency is increased to 10ms / time; for severe anomalies, such as a significant increase in harmonic abrupt changes, the sampling frequency is increased to 1ms / time.

[0156] S330: Start dual-channel high-precision acquisition for successfully matched nodes: Channel 1 directly captures the raw current / voltage instantaneous values, and Channel 2 inversely calculates the equivalent loss value based on the thermal circuit model.

[0157] To further verify the accuracy of the anomaly detection results, the system activated a dual-channel verification mechanism, cross-validating the results by acquiring the original data and inferring the equivalent loss value from the thermal path model.

[0158] First channel: Raw data acquisition:

[0159] The first channel is responsible for acquiring the raw instantaneous current / voltage values, which serve as the primary basis for anomaly detection.

[0160] During data acquisition, ensure sampling accuracy and time synchronization to avoid errors caused by noise or delay.

[0161] In one possible implementation, the system employs high-precision ADC (analog-to-digital converter) and GPS time synchronization technology to ensure data accuracy and consistency.

[0162] Second channel: Equivalent loss value derived from thermal circuit model:

[0163] The second channel uses the thermal circuit model principle to inversely deduce the equivalent loss value, which is used to assess the impact of abnormal modes on the system.

[0164] The core idea of ​​the thermal circuit model is to map the energy loss of electrical equipment to a thermal network, and derive the equivalent loss value by calculating parameters such as thermal resistance and heat capacity. Specifically, the formula for the thermal circuit model is:

[0165]

[0166] Among them, P loss K represents the equivalent loss value. T Represents the temperature compensation coefficient, I(t) represents the instantaneous current value, and R... th denoted by , V(t) represents the equivalent thermal resistance, V(t) represents the instantaneous voltage value, tanδ represents the dielectric loss factor, and t0 and t1 represent the start and end times of the zero-crossing characteristic window.

[0167] The dual-channel verification mechanism is implemented as follows:

[0168] When the dual-channel verification results show that the equivalent loss value of a node exceeds the preset threshold, the system needs to mark it as a high-risk node and upload relevant data first.

[0169] In the load-adaptive dynamic data acquisition method for electricity marketing, the S400 is used to select the appropriate data transmission mechanism under different load conditions. The selection of the data transmission mechanism is a core element in ensuring efficient utilization of system resources and priority delivery of critical data. Depending on the system load state, the system needs to dynamically adjust the data transmission strategy to balance resource consumption and data integrity.

[0170] The choice of data transmission mode is directly determined by the system's current load status. By monitoring the LAI in real time and comparing it with ATT, the system automatically switches to the data transmission mode suitable for the current load status.

[0171] In one possible implementation, under high load conditions, the terminal devices face significant resource pressure, and network bandwidth may approach saturation. To reduce redundant data transmission and optimize resource allocation, the system adopts an active pull-based transmission mode.

[0172] Specifically, the proactive pull-based transmission mode involves the platform sending request commands to the terminal based on demand. The terminal only uploads compressed data and its equivalent loss value within the feature window. The upload priority mechanism is based on the magnitude of the equivalent loss value. The proactive pull-based transmission mode avoids network congestion caused by uploading the entire amount of data, while ensuring that critical data is delivered in a timely manner.

[0173] In one possible implementation, under standard or low load conditions, system resources are relatively abundant, thus allowing a switch to broadcast full transmission mode.

[0174] Specifically, broadcast full transmission allows terminal devices to upload the complete dataset, including feature window data, non-feature window data, and background data stream. However, to prevent duplicate data transmission, the system introduces a flag-based deduplication mechanism. This mechanism avoids uploading duplicate data by adding a unique identifier to each data segment, ensuring that the same data is uploaded only once.

[0175] In the load-adaptive dynamic data acquisition method for electricity marketing, the S500, after receiving data uploaded by terminal devices, reconstructs the electricity consumption scenario using digital simulation technology and performs a deviation analysis between the equivalent loss values ​​reported by the terminals and the simulation results. This step aims to verify the accuracy and consistency of the data.

[0176] Please refer to Figure 5 The diagram illustrates a flowchart of an exemplary load-adaptive dynamic data acquisition method for electricity marketing, S500, which includes:

[0177] S510: After receiving the data uploaded by the terminal, the platform uses a finite element simulation model to reconstruct the power consumption scenario.

[0178] After receiving the data uploaded by the terminal, the platform first needs to preprocess the received data to ensure the accuracy of subsequent simulations. In order to reconstruct the power consumption scenario and perform deviation analysis, the platform uses a finite element simulation model to simulate the power grid topology, load distribution, and power quality characteristics, providing theoretical reference values.

[0179] In one possible implementation, the core components of the finite element simulation model include a power grid topology module, a load distribution module, and a power quality analysis module.

[0180] Specifically, the power grid topology module generates a mathematical model of the power grid based on the actual power grid structure. The load distribution module combines historical electricity consumption data and real-time monitoring results to simulate the current load distribution in the power grid. The power quality analysis module uses spectrum analysis and signal processing techniques to evaluate power quality issues in the simulation scenario.

[0181] A finite element method (FEM) simulation model is used to simulate actual power consumption scenarios, generating simulation results corresponding to the data uploaded by the terminal. In one possible implementation, the specific process is as follows:

[0182] a. Initial condition setting: Before the simulation starts, determine the initial conditions, including: the initial voltage and current values ​​of each node; the operating status of the equipment; and external disturbance sources.

[0183] b. Time step selection: To ensure simulation accuracy, select a time step. For high-frequency signals, use a time step in the microsecond range; for low-frequency signals, use a time step in the second range.

[0184] c. Multiphysics coupling: The influence of coupling phenomena of multiple physical fields is comprehensively considered during the simulation process to improve the realism of the simulation results.

[0185] d. Output format: After the simulation is completed, the simulation results are output.

[0186] S520: Based on the simulation output results, perform deviation analysis with the terminal data.

[0187] Deviation analysis compares the data uploaded from the terminal with the simulation results, enabling the system to quickly identify anomalies and determine whether a calibration mechanism needs to be triggered. The core of deviation analysis lies in quantifying the difference between the terminal data and the simulation results. The specific formula is as follows:

[0188]

[0189] Where D represents the percentage deviation, P loss P′ represents the equivalent loss value uploaded by the terminal. loss This represents the theoretical equivalent loss value calculated by the simulation model.

[0190] Based on the percentage deviation, it is divided into different levels. In one possible implementation, the criteria for dividing the different levels are as follows:

[0191] Slight deviation: A deviation of less than 5% indicates that the data are basically consistent;

[0192] Moderate bias: Bias ranges from 5% to 15%, requiring further investigation into the cause;

[0193] Severe deviation: Deviation greater than 15% may indicate equipment failure or abnormal data acquisition, requiring immediate triggering of the calibration mechanism.

[0194] S530: If the deviation continues to be severe, a command is sent to the terminal to trigger phase calibration with zero-crossing feature windowing and resampling.

[0195] The core objective of phase calibration is to eliminate deviations caused by sampling window offsets or noise interference. For example, if the sampling window of the terminal device deviates from the theoretical zero-crossing position, it is adjusted to the correct position through calibration.

[0196] In one possible implementation, the phase calibration algorithm is implemented using the least squares method or a Kalman filter to accurately calculate and compensate for the offset of the sampling window.

[0197] For example, for the sampling window offset of a certain node, the system uses the following formula to calculate the calibration value:

[0198]

[0199] Where, Δ t t represents the sampling window offset. i t represents the actual sampling time of the terminal device. i ′ represents the theoretical sampling time, and N represents the number of sampling points.

[0200] After phase calibration is completed, the system triggers a resampling process to capture more accurate power quality characteristics. The resampling data needs to be uploaded to the platform in real time for subsequent simulation comparisons. To reduce transmission latency, the system employs an incremental transmission strategy, uploading only newly added anomaly characteristic data.

[0201] In the load-adaptive dynamic data acquisition method for power marketing, the "zero-crossing feature window diversion technology" is a key element in solving the resource contention problem caused by simultaneous sampling by multiple nodes. By staggering the feature sampling windows of different nodes in time and combining them with a dynamic adjustment mechanism for transmission priority, the S600 system can effectively alleviate resource pressure while ensuring that high-risk data is uploaded to the platform first.

[0202] Please refer to Figure 6 The diagram illustrates a flowchart of an exemplary load-adaptive dynamic data acquisition method for electricity marketing, S600, which includes:

[0203] S610: Real-time monitoring and determination of continuous high load conditions.

[0204] In order to enable the "zero-crossing feature window diversion technology" in a timely manner, the system monitors the load status of multiple nodes in real time and determines whether the conditions for continuous high load are met.

[0205] In one possible implementation, the determination of a continuous high load state satisfies the following condition:

[0206] Three or more nodes are simultaneously under high load;

[0207] The high load state lasts for more than a certain time threshold.

[0208] For example, the system uses a sliding time window algorithm to statistically analyze the load distribution of each node over a recent period. If multiple nodes are shown to be in a high-load state for five consecutive time windows, it is determined to be a continuous high-load state.

[0209] S620: Design a timing allocation strategy for "zero-crossing feature window splitting technology".

[0210] The core of the traffic splitting technology lies in staggering the feature sampling windows of different nodes in time to avoid resource contention caused by multiple nodes sampling simultaneously.

[0211] In one possible implementation, the basic principle of timing allocation is that timing allocation should follow the principles of uniform distribution and dynamic adjustment to ensure that the sampling windows of each node do not overlap, while taking into account the actual needs of power grid operation.

[0212] For example, for two nodes A and B, the system sets the sampling window for node A to 5ms after the voltage zero crossing and the sampling window for node B to 10ms.

[0213] In one possible implementation, the system employs a discrete-time scheduling algorithm to achieve time-series allocation, dynamically allocating sampling windows based on the number of nodes and load pressure. The specific formula is as follows:

[0214] T offset,i =ΔT·(i-1);

[0215] Among them, T offset,i ΔT represents the sampling window offset of the i-th node, ΔT represents the minimum time interval between adjacent nodes, and i represents the node number.

[0216] For example, if four nodes are simultaneously under high load, their sampling window offsets are 0ms, 5ms, 10ms, and 15ms, respectively.

[0217] S630: Transmission priority is dynamically adjusted by the equivalent loss value, and high-risk data with high equivalent loss values ​​are uploaded to the platform first.

[0218] In addition to reducing resource contention between nodes through staggered time allocation, the system also dynamically adjusts data transmission priority based on equivalent loss values. This ensures that data containing important information is uploaded to the platform first, thereby improving the overall system's response speed and decision-making accuracy.

[0219] Specifically, the system assigns different transmission priorities to each node based on the magnitude of the equivalent loss value captured. In one possible implementation, the data transmission priority is dynamically determined by the equivalent loss value, and is divided into three levels: high, medium, and low.

[0220] High priority: Equivalent loss value > 1.5 times the normal range;

[0221] Medium priority: equivalent loss value ∈ [1.2, 1.5] times the normal range;

[0222] Low priority: equivalent loss value < 1.2 times the normal range.

[0223] For example, if a high-risk node has an equivalent loss value that is twice the normal range, then its priority is high.

[0224] This application's method involves deploying load monitoring modules at the distribution network terminal to statistically analyze the number of tasks in real time and assess load status in conjunction with system resource utilization. The method includes hardware architecture design, dynamic assessment of system resource utilization, construction of a load adaptive index model, and selection of different sampling modes and data transmission mechanisms based on load status. Under high load conditions, a "zero-crossing feature windowing" sampling mode is employed, defining and dividing power quality feature windows to capture abnormal data, while simultaneously constructing a power consumption anomaly feature database for matching analysis. Furthermore, it involves technologies such as dual-channel verification, active pull-type data transmission, digital simulation comparison, and phase calibration to optimize resource allocation and improve data acquisition accuracy, ensuring the efficient and stable operation of the power system.

[0225] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0226] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0227] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0228] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features of the invention herein.

[0229] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications or equivalent substitutions made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for dynamic data collection in electricity marketing based on load adaptation, characterized in that, include: Deploy load monitoring modules at the distribution network terminals to count the number of data acquisition tasks in real time and dynamically assess the load status in conjunction with system resource utilization. When the system is under high load, the system automatically switches to the sampling mode of zero-crossing feature windowing; Build a power consumption anomaly feature library; when the system is under high load, if an anomaly pattern matching the power consumption anomaly feature library is detected in the zero-crossing feature window, the system automatically increases the sampling frequency of that node to the millisecond level and starts a dual-channel verification mechanism. The data transmission mechanism is selected based on the load status determined by the system. Under high load, active pull transmission is used, and the terminal only uploads the feature window compressed data and the equivalent loss value. Under standard load or low load, broadcast full transmission is switched, but duplicate data is avoided by deduplication through flag bits. After receiving the data uploaded by the terminal, the platform starts the digital simulation comparison process; if the deviation continues to exceed the standard, it sends an instruction to the terminal to trigger the phase calibration of the zero-crossing feature window and resamples. When the system detects that multiple nodes are under high load, it automatically enables zero-crossing feature window splitting technology. At the same time, the transmission priority is dynamically adjusted by the equivalent loss value.

2. The method for dynamic data collection of electricity marketing based on load adaptation according to claim 1, characterized in that, The dynamic assessment of load status includes: Constructing a mathematical model for the load adaptability index LAI: LAI = ω1·T + ω2·R + ω3·N; Where N represents the number of currently active tasks, R represents the system resource utilization rate, T represents the network latency or task response time, and ω1, ω2, and ω3 represent weighting coefficients. The load adaptive trigger threshold (ATT) is dynamically adjusted using a sliding time window algorithm, and the ATT value is increased during peak electricity consumption periods. The normalization methods for the current number of active tasks, system resource utilization, and network latency or task response time are as follows: Task quantity normalization: Let the maximum number of tasks be N. max Then the normalized number of tasks is N′=N / N max ; Resource utilization rate normalization: Let the upper limit of resource utilization rate be R. max Then the normalized resource utilization rate is R′=R / R max ; Network latency or task response time normalization: Let the maximum allowable latency be T. max Then the normalized network delay is T′=T / T max .

3. The method for dynamic data acquisition in electricity marketing based on load adaptation according to claim 1 or 2, characterized in that, The dynamic assessment of load status also includes: comparing the calculated LAI with ATT to classify the load status; The logic for classifying load states is as follows: Compare the calculated LAI with ATT to determine the current load status: If LAI < ATT HLS This indicates a high load condition. If ATT SLS ≤LAI≤ATT HLS This is determined to be a standard load condition; If LAI < ATT SLS This indicates a low load state. Among them, ATT HLS ATT adapts to the trigger threshold for high-load conditions. SLS Low load state load adaptive trigger threshold.

4. The method for dynamic data collection of electricity marketing based on load adaptation according to claim 1, characterized in that, The sampling modes for zero-crossing feature windowing include: The signal is preprocessed using a digital filter, and then combined with a zero-crossing detection algorithm or a hardware comparator circuit to achieve microsecond-level zero-crossing detection. Based on the timing optimization strategy, the sampling window is limited to the abnormal feature range of ±1ms to ±2ms from the zero crossing point; Power quality anomaly features within a window can be extracted using wavelet transform, Hilbert-Huang transform, or convolutional neural networks.

5. The method for dynamic data acquisition in electricity marketing based on load adaptation according to claim 4, characterized in that, The sampling mode for zero-crossing feature windowing also includes: The power quality feature windows are dynamically prioritized and sorted based on a weighted score of the intensity, duration and scope of the anomaly.

6. The method for dynamic data collection of electricity marketing based on load adaptation according to claim 1, characterized in that, The construction of the electricity consumption anomaly feature library includes: the feature library is designed with a three-layer structure, the top layer is the anomaly category, the middle layer is the specific anomaly mode, and the bottom layer is the feature parameters; The specific anomaly patterns are classified into templates using clustering algorithms, and feature parameters are extracted from historical data using Fourier transform or wavelet transform. The matching adopts a hierarchical matching strategy: primary matching calculates similarity through dynamic time warping, and when the similarity is >85%, it enters advanced matching; advanced matching uses a deep feature matching network to extract abstract features and calculate cosine similarity; in the decision mechanism, a comprehensive similarity threshold of 0.92 is set, and if the threshold is exceeded, it is determined to be a valid match.

7. The method for dynamic data acquisition in electricity marketing based on load adaptation according to claim 1, characterized in that, The dual-channel verification mechanism includes: Channel 1: Employs high-precision ADC and GPS time synchronization technology to capture raw instantaneous current / voltage values; Channel 2: Equivalent loss value calculated based on thermal path model: The formula for the thermal circuit model is: Among them, P loss K represents the equivalent loss value. T Represents the temperature compensation coefficient, I(t) represents the instantaneous current value, and R... th denoted by , V(t) represents the instantaneous voltage value, tanδ represents the dielectric loss factor, and t0 and t1 represent the start and end times of the zero-crossing characteristic window. The dual-channel verification mechanism is implemented as follows: When the dual-channel verification results show that the equivalent loss value of a node exceeds the preset threshold, the system needs to mark it as a high-risk node and upload data first.

8. The method for dynamic data collection of electricity marketing based on load adaptation according to claim 1, characterized in that, The data transmission mechanism satisfies: When using active pull-based transmission, the upload priority is set according to the equivalent loss value. During broadcast full transmission, unique identifiers are used to deduplicate flag bits.

9. The method for dynamic data acquisition in electricity marketing based on load adaptation according to claim 1, characterized in that, The digital simulation comparison process includes: (a) Reconstructing the power grid topology, load distribution, and multiphysics coupling scenario using a finite element simulation model; (b) Calculate the equivalent loss value P uploaded by the terminal. loss The theoretical equivalent loss value P′ calculated by the simulation model loss The deviation; the calculation formula is: Where D represents the percentage deviation, P loss P′ represents the equivalent loss value uploaded by the terminal. loss This represents the theoretical equivalent loss value calculated by the simulation model; (c) Based on the percentage of deviation, it is divided into different levels; (d) If the deviation is severe and persistent, a command is sent to the terminal to trigger phase calibration with zero-crossing feature windowing and to resample; The phase calibration algorithm is implemented using the least squares method or a Kalman filter, and the system calculates the calibration value using the following formula: Where Δt represents the sampling window offset, t i t represents the actual sampling time of the terminal device. i ′ represents the theoretical sampling time, and N represents the number of sampling points; Among them, after bit calibration, resampled data is uploaded through an incremental transmission strategy.

10. The method for dynamic data collection of electricity marketing based on load adaptation according to claim 1, characterized in that, The zero-crossing feature window splitting technique includes: The discretized sampling window offset is assigned to consecutive high-load nodes, and the calculation formula is as follows: T offset,i =ΔT·(i-1); Among them, T offset,i ΔT represents the sampling window offset of the i-th node, ΔT represents the minimum time interval between adjacent nodes, and i represents the node number; Transmission priority is classified according to equivalent loss value: High priority: Equivalent loss value > 1.5 times the normal range; Medium priority: equivalent loss value ∈ [1.2, 1.5] times the normal range; Low priority: equivalent loss value < 1.2 times the normal range.

Images