Client awareness and metrication method for power consumption information collection data transmission

Abstract

The application discloses a client perception and quantification method for power utilization information collection data transmission, comprising the following steps: collecting multi-dimensional parameters such as link bandwidth occupation, transmission delay and packet loss rate through an edge end electric collection perception calculation and analysis platform, performing feature mapping and enhancement through an electric collection transmission abnormality perception identification network, calling a transmission link perception diagnosis prediction model to predict abnormal trends, dynamically correcting load deviation by using a client load balancing error correction algorithm, constructing a quantification evaluation system by integrating various data through the platform, and finally outputting results such as client perception quantification values, while performing operations such as calibration feature screening, time sequence decomposition, correction amount calculation and parameter integration. Through multi-technology cooperation and full-process closed-loop design, the application realizes comprehensive perception and accurate quantification of the transmission state of the client, and adapts to the complex scene requirements of power utilization information collection data transmission.

Description

Technical Field

[0001] This invention relates to the field of electricity information acquisition and data management technology, and in particular to a client-side perception and measurement method for electricity consumption information acquisition and data transmission. Background Technology

[0002] With the large-scale deployment of electricity consumption information collection systems, the transmission scenarios between clients and edge devices are becoming increasingly complex. Issues such as link status fluctuations, uneven load distribution, and data transmission anomalies directly affect the integrity and timeliness of the collected data. Traditional collection and transmission methods often rely on single-parameter monitoring, which is insufficient to cover multiple influencing factors such as link bandwidth, transmission latency, packet loss rate, and node load. Furthermore, the collaborative perception capabilities between edge devices and clients are inadequate, resulting in a lack of systematic and accurate control over the transmission status. To meet the practical needs of comprehensive client status perception, early anomaly prediction, and scientific quantification in electricity consumption information transmission, it is urgent to construct a complete solution that integrates multi-source parameters, merges core algorithms, and a dedicated platform. This solution should achieve closed-loop management of the entire process from parameter acquisition and feature processing to quantification output, providing a reliable basis for transmission optimization.

[0003] Existing technologies have two key shortcomings: First, multi-parameter processing lacks an effective integration mechanism, failing to systematically integrate different types of parameters such as link status, load conditions, and transmission protocol adaptation. It only analyzes single-dimensional data in isolation, resulting in a one-sided portrayal of the transmission status and failing to reflect the synergistic effects between various factors. Second, the accuracy and dynamic adaptability of perception and quantification are insufficient. There is a lack of targeted anomaly prediction models and load balancing error correction schemes. The prediction of transmission anomalies is lagging, and load adjustment does not take into account real-time transmission status and client characteristics, resulting in a large deviation between the quantification results and the actual transmission situation, making it difficult to support refined transmission optimization decisions. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of existing technologies, this invention provides a client-side perception and quantification method for electricity consumption information collection and data transmission.

[0005] The technical solution adopted in this invention is a client-side perception and quantification method for electricity information collection and data transmission, comprising the following steps: S1, obtaining the link bandwidth occupancy coefficient, data frame transmission delay parameters, data packet loss rate, client access node load, transmission protocol adaptation parameters, and electricity data priority identifier during the client-side electricity information collection and transmission process through an edge-end electricity collection perception and analysis platform; S2, performing multi-dimensional feature mapping on the parameters based on the electricity collection transmission anomaly perception and identification network, constructing a client transmission status perception matrix, and performing nonlinear transformation and feature enhancement on the client transmission status perception matrix through network hidden layer nodes; S3, calling the transmission link perception diagnosis and prediction model to... The enhanced client transmission status perception matrix predicts transmission anomaly trends and outputs link stability correlation parameters and anomaly probability distribution; S4, the client load balancing error correction algorithm dynamically corrects the load distribution deviation parameters in the prediction results, generating load balancing adjustment coefficients and data transmission path optimization strategies; S5, the feature mapping results, anomaly prediction parameters, and error correction data are integrated through the edge-end power collection and perception computing analysis platform to construct a client transmission status metric evaluation system; S6, based on this evaluation system, the client perception metric value, transmission link health index, and load balancing adaptation parameters are output, forming client perception and metric results for power consumption information collection and data transmission.

[0006] Furthermore, the feature mapping expression of the power acquisition transmission anomaly sensing and identification network is as follows: ,

[0007] in, Transmit the state-aware matrix to the client. For network activation functions, This is the network weight matrix. For bias terms, For parameter adjustment coefficients, This is the link bandwidth utilization factor. For data frame transmission delay parameters, This refers to the packet loss rate metric. For client access node load, For transmission protocol adaptation parameters, This is used as a priority identifier for electrical data collection.

[0008] Furthermore, the abnormal trend prediction expression of the transmission link sensing diagnostic prediction model is as follows: ,

[0009] in, For abnormal trend prediction results, The number of feature dimensions. Transmit the state-aware matrix to the client dimensional elements, These are the dimension weight coefficients. These are trend adjustment parameters.

[0010] Furthermore, the error correction expression for the client-side load balancing error correction algorithm is as follows: ,in, This is the load balancing error correction amount. This represents the actual load. For the expected load, For the number of clients, For the first Client transmission rate For the first Client cache capacity, To correct the algorithm parameters.

[0011] Furthermore, the parameter integration expression of the edge-end power acquisition and sensing computing analysis platform is as follows: ,in, The integrated parameter matrix, Element-wise addition of matrices. Element-wise multiplication of matrices. To integrate the weighting coefficients.

[0012] Furthermore, the quantitative output expression of the client transmission state metric evaluation system is as follows: ,in, For the client to perceive metric values, The first calibration dimension element of the composite parameter matrix. This is the second calibration dimension element of the composite parameter matrix. This is the third calibration dimension element of the comprehensive parameter matrix. This is the fourth calibration dimension element of the comprehensive parameter matrix. This is the fifth calibration dimension element of the comprehensive parameter matrix. This is the quantization coefficient.

[0013] Further, step S3 includes the following sub-steps: S31, extracting link stability correlation features from the client transmission status perception matrix through the transmission link perception and diagnosis prediction model, matching these features with historical transmission anomaly data, and filtering out a subset of calibrated features with anomaly prediction value; S32, constructing a multi-dimensional prediction vector based on the calibrated feature subset, and performing time series decomposition on the vector through the time series prediction module in the model to separate trend components, periodic components, and random components; S33, substituting the decomposed components into the prediction expression of the transmission link perception and diagnosis prediction model, and generating a preliminary anomaly occurrence probability distribution through matrix operations; S34, performing feature fusion processing on the preliminary probability distribution, removing probability values ​​corresponding to invalid noise data, and outputting the purified link stability correlation parameters and anomaly occurrence probability distribution.

[0014] Further, step S4 includes the following sub-steps: S41, obtaining the anomaly probability distribution output by the transmission link perception diagnosis prediction model and the load data of the current access node of the client, and determining the calibration influence factor of the load distribution deviation; S42, substituting the calibration influence factor into the error correction expression of the client load balancing error correction algorithm, and calculating the initial load balancing error correction amount; S43, dynamically adjusting the initial correction amount based on the real-time transmission status parameters fed back by the edge power acquisition perception calculation analysis platform, and optimizing the dimension weight coefficient in the correction algorithm; S44, generating the corresponding load balancing adjustment coefficient according to the adjusted correction amount, and formulating a data transmission path optimization strategy in combination with the characteristics of the client transmission path.

[0015] Further, S5 includes the following sub-steps: S51, collecting the feature mapping results output by the power acquisition transmission anomaly perception and identification network through the parameter receiving module of the edge-end power acquisition perception and analysis platform, and standardizing the data format; S52, receiving the anomaly prediction parameters output by the transmission link perception diagnosis and prediction model, aligning them with the feature mapping results in dimensions, and constructing a preliminary integrated dataset; S53, importing the error correction data generated by the client load balancing error correction algorithm, and performing matrix operations corresponding to the parameter integration expression through the platform's integrated operation module; S54, constructing a client transmission status metric evaluation system including perception features, prediction indicators, and correction parameters based on the operation results, and clarifying the weight allocation rules for each evaluation dimension.

[0016] A client-side perception and quantification method for electricity consumption information data collection and transmission is proposed. This method is implemented through different units, including: a multi-dimensional electricity collection parameter acquisition unit, which establishes a data transmission connection with an edge-end electricity collection perception and analysis platform to collect link bandwidth occupancy coefficients and data frame transmission delay parameter calibration parameters during the client's electricity consumption information collection and transmission process, and transmits the collected data to a feature mapping processing unit in real time; a feature mapping processing unit, which communicates bidirectionally with the multi-dimensional electricity collection parameter acquisition unit and the transmission anomaly prediction unit, performs nonlinear transformation and feature enhancement on the collected parameters through an electricity transmission anomaly perception and identification network, and outputs a client transmission status perception matrix to the transmission anomaly prediction unit; and a transmission anomaly prediction unit, which is connected to both the feature mapping processing unit and the load balancing correction unit, and calls upon the transmission link... The perception and diagnostic prediction model predicts abnormal trends in the client transmission status perception matrix and sends the prediction results to the load balancing correction unit. The load balancing correction unit establishes signal interaction with the transmission anomaly prediction unit and the parameter integration and evaluation unit, and uses the client load balancing error correction algorithm to dynamically correct the load distribution deviation, outputting the corrected data to the parameter integration and evaluation unit. The parameter integration and evaluation unit communicates with the load balancing correction unit and the quantification result output unit, integrates various parameters through the edge-end power acquisition perception calculation and analysis platform, constructs a quantified evaluation system, and transmits the evaluation data to the quantification result output unit. The quantification result output unit is unidirectionally connected to the parameter integration and evaluation unit, and outputs the client perception quantification value, transmission link health index, and load balancing adaptation parameters based on the evaluation system.

[0017] Beneficial Effects: This invention proposes a client-side perception and quantification method for electricity information collection and data transmission. By constructing a complete technical system encompassing "multi-parameter acquisition - feature mapping - anomaly prediction - error correction - parameter integration - quantification output," it effectively overcomes the shortcomings of existing technologies, such as isolated processing of multiple parameters and insufficient accuracy in perception quantification. This invention utilizes an edge-end electricity data collection and analysis platform to comprehensively collect multi-dimensional parameters such as link bandwidth occupancy, transmission latency, and packet loss rate. Through an electricity transmission anomaly perception and identification network, it achieves deep fusion and feature enhancement of multi-source parameters, solving the problem of insufficient parameter integration in traditional methods. The transmission link perception diagnosis and prediction model predicts anomaly trends in advance, and the client-side load balancing error correction algorithm dynamically optimizes load allocation, overcoming the shortcomings of existing technologies such as delayed anomaly prediction and lack of targeted load adjustment. The core algorithms and platform work collaboratively, achieving organic integration of different types of parameters through standardized processing and dimensional adaptation. The constructed quantification evaluation system can output accurate client-side perception quantification values ​​and link health indicators. The supporting six-unit system realizes closed-loop linkage of parameter acquisition, feature processing, anomaly prediction, load correction, parameter integration and quantification output, ensuring the comprehensiveness, real-time and accuracy of perception and quantification, providing reliable support for the optimization decision-making of electricity information acquisition and data transmission, and significantly improving transmission stability and data transmission quality. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0019] Figure 2 This is a flowchart of method step S3 of the present invention;

[0020] Figure 3 This is a flowchart of method step S4 of the present invention;

[0021] Figure 4 This is a flowchart of step S5 of the method of the present invention;

[0022] Figure 5 This is a diagram showing the unit composition for implementing the method of the present invention. Detailed Implementation

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] like Figure 1 As shown, the client-side sensing and quantification method for electricity consumption information collection and data transmission includes the following steps:

[0025] S1, through the edge-end power collection and perception computing analysis platform, obtains the link bandwidth occupancy coefficient, data frame transmission delay parameters, data packet loss rate index, client access node load, transmission protocol adaptation parameters and power collection data priority identifier during the power consumption information collection and transmission process of the client.

[0026] Specifically, the edge-end power consumption data collection and analysis platform establishes a real-time communication link with the client's power consumption information collection terminal. It acquires six core parameters during the power consumption information transmission process at a collection frequency of once every 10 milliseconds. These parameters include: a link bandwidth occupancy coefficient, monitored in real-time by the platform's built-in bandwidth detection module as the proportion of effective data occupying the total bandwidth in the transmission channel, with a value ranging from 0.1 to 0.9; a data frame transmission delay parameter, recorded to the microsecond level, detailing the time difference between data frame transmission from the client and reception completion at the edge node; and a packet loss rate, calculated by the difference between the total number of data packets sent and the number of successfully received data packets per unit time. The ratio ranges from 0 to 0.3; the client access node load is calculated by statistically analyzing the number of concurrent transmission tasks currently being processed by the access node, the proportion of memory resources occupied, and the CPU utilization rate, with a value range of 10 to 500; the transmission protocol adaptation parameter generates an adaptation coefficient of 0.2 to 0.8 based on the transmission protocol type used by the client, such as TCP or UDP; the electricity data priority identifier is divided into five levels from 1 to 5 according to the importance of the electricity information, each corresponding to a different transmission priority weight. The platform synchronously collects and stores the above six parameters through a multi-interface data aggregation module, providing complete data support for subsequent processing.

[0027] S2, based on the power acquisition transmission anomaly perception and identification network, performs multi-dimensional feature mapping on parameters, constructs the client transmission status perception matrix, and performs nonlinear transformation and feature enhancement on the client transmission status perception matrix through network hidden layer nodes;

[0028] Specifically, after receiving six core parameters transmitted from the edge-end power acquisition and sensing computing analysis platform, the power acquisition transmission anomaly perception and identification network first inputs each parameter into the network's input layer nodes. The input layer consists of 12 neurons, each corresponding to a different feature dimension of a parameter. Subsequently, the parameters are multi-dimensionally feature-mapped through the network's first hidden layer. This hidden layer uses 32 neurons and performs feature transformation on each parameter through a non-linear activation function. It correlates the link bandwidth occupancy coefficient with the packet loss rate index, couples the data frame transmission delay parameter with the transmission protocol adaptation parameter, and constructs a multi-dimensional feature space by combining the client access node load and the power acquisition data priority identifier. This forms a 16-row, 32-column client transmission status perception matrix. This matrix is ​​then input into the network's second hidden layer, which consists of 24 neurons. Through matrix multiplication and feature recombination, the client transmission status perception matrix undergoes non-linear transformation, strengthening the correlation features and anomaly-sensitive features between different parameters, filtering out invalid and redundant information, and improving the client transmission status perception matrix's ability to represent the transmission status, thus providing high-quality feature data for subsequent anomaly prediction.

[0029] S3, call the transmission link perception diagnosis and prediction model to predict the transmission anomaly trend of the enhanced client transmission status perception matrix, and output the link stability correlation parameters and anomaly occurrence probability distribution;

[0030] Specifically, the transmission link-aware diagnostic prediction model calls the client transmission status awareness matrix after feature enhancement. First, the model's feature input module performs dimensionality verification and format conversion on the matrix to ensure that the matrix dimensions are consistent with the model input requirements. Then, the model's time-series analysis module divides the client transmission status awareness matrix into time series segments, using 50 data points as an analysis window to extract the feature change trend within each window. Combining this with the model's built-in historical transmission anomaly feature library, it matches and identifies anomalous sensitive features in the current client transmission status awareness matrix. Through similarity calculation, feature vectors with a similarity higher than 0.7 with historical anomaly features are selected. Then... The model's probability prediction module employs a multi-dimensional probability calculation method to extrapolate the probability of anomalies from the selected feature vectors. It considers various anomaly scenarios, such as sudden changes in link bandwidth occupancy coefficients, continuous increases in data frame transmission delay parameters, and packet loss rates exceeding thresholds. The output includes 10 dimensions of link stability-related parameters, with each dimension corresponding to a transmission stability index. Simultaneously, it generates a continuous anomaly probability distribution between 0 and 1. The interval with a probability value higher than 0.6 is marked as a high-risk anomaly area, the interval between 0.3 and 0.6 is a medium-risk area, and the interval below 0.3 is a low-risk area, providing a clear predictive basis for subsequent error correction.

[0031] S4. The client load balancing error correction algorithm is used to dynamically correct the load distribution deviation parameter in the prediction result, and generate load balancing adjustment coefficient and data transmission path optimization strategy.

[0032] Specifically, the client-side load balancing error correction algorithm first receives the anomaly probability distribution and client access node load data output by the transmission link sensing diagnostic prediction model. Through the algorithm's impact factor analysis module, it determines the core influencing factors of load distribution deviation, including load parameters corresponding to dimensions with high anomaly probabilities, differences in client access node load distribution, and the degree of adaptation to different client transmission rates. Subsequently, based on these core influencing factors and a preset expected load range, the algorithm calculates the deviation between the actual load and the expected load. The expected load is determined by considering the number of clients, the upper limit of the transmission link bandwidth, and the priority identifier of the data collected. The algorithm sets the expected load range for a single client to be 20 to 80. It then adjusts the deviation value using a dynamic correction module, introducing the square root of the probability of anomalies as the correction weight. This is combined with a weighted sum of the transmission rates and cache capacities of all clients for normalization. Simultaneously, a sigmoid function is used to apply non-linear constraints to the correction amount, preventing over- or under-correction. Finally, a load balancing adjustment coefficient between 0.05 and 0.5 is generated, with each client corresponding to a unique adjustment coefficient. Furthermore, based on the adjustment coefficient and parameters such as bandwidth and latency of the transmission link, a differentiated data transmission path optimization strategy is formulated to allocate the optimal transmission path for data collected from different priorities.

[0033] S5 integrates feature mapping results, anomaly prediction parameters, and error correction data through the edge-end power acquisition and sensing computing analysis platform to construct a quantified evaluation system for client transmission status.

[0034] Specifically, the edge-end power acquisition sensing and analysis platform initiates a parameter integration process. First, it receives the client transmission status perception matrix output by the power acquisition transmission anomaly sensing and identification network through the platform's parameter receiving module. This module supports multi-format data reception and converts the client transmission status perception matrix into a unified data format for the platform. Then, it receives the link stability correlation parameters and anomaly probability distribution output by the transmission link sensing diagnosis and prediction model. Using a dimension alignment algorithm, the dimensions of the two types of parameters are unified to 16 dimensions, constructing a preliminary integrated dataset. This dataset includes 32 feature values ​​of the client transmission status perception matrix, 10 dimension values ​​of the link stability correlation parameters, and 5 interval values ​​of the anomaly probability distribution. Afterwards, the platform imports the client... The load balancing adjustment coefficients and error correction data generated by the client-side load balancing error correction algorithm are integrated by the platform's integration and calculation module using a combination of matrix addition and element-wise multiplication to weight and integrate the three types of data. The weight coefficients for feature mapping results, anomaly prediction parameters, and error correction data are all set to 0.3. During the integration process, the data is normalized to ensure that all parameter values ​​are within the range of 0 to 1. Finally, a client transmission status metric evaluation system is constructed, including the perception feature dimension, anomaly prediction dimension, and load correction dimension. The specific indicators and calculation rules for each evaluation dimension are clearly defined, providing a standardized evaluation framework for quantitative output.

[0035] S6, based on the evaluation system, outputs client perception quantification values, transmission link health indicators, and load balancing adaptation parameters to form client perception and quantification results for electricity information collection and data transmission.

[0036] Specifically, based on the constructed client transmission status metric evaluation system, firstly, 16 core indicators of the perception feature dimension, 8 key parameters of the anomaly prediction dimension, and 12 feature values ​​of the load correction dimension are extracted from the evaluation system. The client perception metric value is calculated by weighted summation, where the weight of the perception feature dimension is 0.35, the anomaly prediction dimension is 0.35, and the load correction dimension is 0.3. The quantification value ranges from 0 to 100, with higher values ​​indicating better client perception. Subsequently, the transmission link health index is calculated. This index is derived by comprehensively considering the 6 core dimensions of the link stability-related parameters, the proportion of risk intervals in the anomaly probability distribution, and the rationality of the load balancing adjustment coefficient. The data is divided into ten levels, from 1 to 10, with higher levels indicating better link health. Simultaneously, load balancing adaptation parameters are output, including the optimal load allocation value, transmission path selection coefficient, and protocol adaptation adjustment value for each client. The optimal load allocation value is calculated based on the client access node load, transmission rate, and buffer capacity. The transmission path selection coefficient is determined by combining the link bandwidth occupancy coefficient and data frame transmission delay parameters. The protocol adaptation adjustment value is generated based on the transmission protocol adaptation parameters and the priority identifier of the electricity data collection. The combined output of these three quantitative results forms a complete client-side perception and quantification result for electricity information collection and data transmission, providing comprehensive data support for optimizing electricity information transmission.

[0037] Preferably, the feature mapping expression of the power transmission anomaly sensing and identification network is: ,in, Transmit the state-aware matrix to the client. For network activation functions, This is the network weight matrix. For bias terms, For parameter adjustment coefficients, This is the link bandwidth utilization factor. For data frame transmission delay parameters, This refers to the packet loss rate metric. For client access node load, For transmission protocol adaptation parameters, This is used as a priority identifier for electrical data collection.

[0038] Specifically, the feature mapping expression of the power transmission anomaly perception and identification network is based on the requirement of multi-parameter collaborative feature extraction. It divides six core parameters into two groups: link status and client attributes, assigns different weight matrices to each group, performs linear transformation, introduces an adjustment coefficient to balance the contribution of each group of parameters, and finally achieves non-linear feature mapping through an activation function. This formula is based on the fact that parameters such as link bandwidth occupancy and packet loss rate have different correlation strengths with transmission anomalies, requiring weight allocation to strengthen key features. The adjustment coefficient ranges from 0.1 to 0.9, determined based on parameter sensitivity test results under different power consumption scenarios. The matrix is ​​obtained through iterative optimization of training samples, with bias terms ranging from -0.5 to 0.5. In the implementation process, the six parameters are first substituted into two sets of linear combination operations, and then the combination result is nonlinearly transformed through an activation function to generate the client transmission status awareness matrix. This formula can effectively integrate multi-dimensional parameter information, highlight anomaly sensitive features, and suppress irrelevant interference information, so that the output client transmission status awareness matrix more accurately reflects the client transmission status, providing a highly recognizable feature basis for subsequent anomaly prediction. Its rationality is verified by the feature extraction accuracy under a large number of simulated transmission scenarios, ensuring that effective features can be stably output under different load and link conditions.

[0039] Preferably, the abnormal trend prediction expression of the transmission link sensing diagnosis prediction model is: ,in, For abnormal trend prediction results, The number of feature dimensions. Transmit the state-aware matrix to the client dimensional elements, These are the dimension weight coefficients. These are trend adjustment parameters.

[0040] Specifically, the anomaly trend prediction expression of the transmission link perception diagnosis prediction model follows the idea of ​​combining time-series features and probability distributions. It performs weighted product operations on the elements of each dimension of the client transmission status perception matrix, incorporates the proportion coefficients of each dimension to strengthen the influence of the main features, and introduces a logarithmic function to correct the influence of extreme values. The formula is based on the fact that the occurrence of transmission anomalies is related to the synergistic changes of multiple feature dimensions. A single-dimensional feature cannot fully reflect the anomaly trend, so multiplication operations are needed to reflect the synergistic effect of multiple dimensions. The dimension weight coefficients are determined based on the contribution analysis of each dimension in historical anomaly data, with a value range of 0.05 to 0.3. The trend adjustment parameter has a value of 0.2 to 0.8, which was obtained through orthogonal experiment optimization. In implementation, the feature dimensions are first divided and the proportion of each dimension is calculated. Then, the product operation and logarithmic correction term are substituted to generate the anomaly trend prediction result. This formula can comprehensively consider the dynamic changes of multi-dimensional features, accurately depict the probability distribution of anomaly occurrence, and distinguish the anomaly intervals of different risk levels. Its effectiveness has been verified by comparative experiments. It can accurately output the link stability correlation parameters and anomaly probability distribution under various anomaly scenarios, providing reliable predictive input for load balancing error correction.

[0041] Preferably, the error correction expression of the client load balancing error correction algorithm is: ,in, This is the load balancing error correction amount. This represents the actual load. For the expected load, For the number of clients, For the first Client transmission rate For the first Client cache capacity, To correct the algorithm parameters.

[0042] Specifically, the error correction expression of the client-side load balancing error correction algorithm is based on the dynamic adjustment logic of deviation. It uses the difference between the actual load and the expected load as the core, introduces the square root of the probability of anomalies as a risk weight, normalizes it through a weighted sum of client transmission rate and cache capacity, and then uses an S-shaped function to constrain the range of the correction amount. The formula is based on the fact that load distribution deviation is affected by both the degree of anomaly risk and the client's hardware capabilities; the higher the anomaly probability, the greater the correction force needs to be. At the same time, it is necessary to consider the client's transmission rate and cache capacity to avoid the correction exceeding the hardware's carrying capacity. In the correction algorithm parameters, λ is set to 0. The values ​​of η and μ range from 0.3 to 1.2, κ range from 1 to 5, and τ range from 0.4 to 0.6. These values ​​were determined through load balancing simulation experiments. During implementation, the load deviation value was first calculated, and then the risk weight, normalization processing, and nonlinear constraint calculation were performed sequentially to generate the load balancing error correction amount. This formula can dynamically adjust the correction intensity according to abnormal risks, taking into account both load balancing and hardware compatibility. It avoids transmission interruption caused by over-correction or load imbalance caused by under-correction. Its feasibility was verified through multi-client concurrent transmission tests to ensure that the corrected load distribution meets the transmission stability requirements.

[0043] Preferably, the parameter integration expression of the edge-end power acquisition and sensing computing analysis platform is: ,in, The integrated parameter matrix, Element-wise addition of matrices. Element-wise multiplication of matrices. To integrate the weighting coefficients.

[0044] Specifically, the parameter integration expression of the edge-end power acquisition and sensing computing analysis platform is based on the principle of multi-source data fusion. It adopts a matrix operation method combining element-wise addition and multiplication, assigning different integration weights to feature mapping results, anomaly prediction parameters, and error correction data. The formula is established based on the different roles of the three types of data in metric evaluation: feature mapping results are basic data, anomaly prediction parameters are trend indicators, and error correction data are the basis for adjustment. The importance of each needs to be reflected through weight allocation. The integration weight coefficient ranges from 0.2 to 0.5 and is determined by the analytic hierarchy process. During implementation, the three types of data are first converted into a matrix form of a unified dimension, and then element-wise addition and multiplication operations are performed to generate a comprehensive parameter matrix. This formula can effectively integrate multi-source heterogeneous data, eliminate fusion errors caused by differences in data format, and strengthen the correlation information between different data. Its rationality is verified by data fusion consistency testing, ensuring that the integrated comprehensive parameter matrix can comprehensively include the core information of the client's transmission status, providing complete data support for the construction of the metric evaluation system. At the same time, the calculation process is simple and efficient, adapting to the computing resource limitations of the edge platform.

[0045] Preferably, the quantization output expression of the client transmission state metric evaluation system is: ,in, For the client to perceive metric values, The first calibration dimension element of the composite parameter matrix. This is the second calibration dimension element of the composite parameter matrix. This is the third calibration dimension element of the comprehensive parameter matrix. This is the fourth calibration dimension element of the comprehensive parameter matrix. This is the fifth calibration dimension element of the comprehensive parameter matrix. This is the quantization coefficient.

[0046] Specifically, the quantitative output expression of the client-side perception quantification evaluation system is based on a multi-dimensional weighted evaluation approach. It integrates the core dimensions of the comprehensive parameter matrix through square root operations and introduces the arctangent function to balance the numerical differences of different dimensions. The formula is established based on the fact that client-side perception needs to comprehensively reflect the performance of feature mapping, anomaly prediction, and load correction. The importance of the core dimensions is reflected by quantification coefficients, which range from 0.1 to 0.4 and are determined according to the priority ranking of quantification evaluation indicators. In the implementation process, the core dimension elements of the comprehensive parameter matrix are first extracted and substituted into the square root operation and arctangent correction operation to generate the client-side perception quantification value. This formula can transform multi-dimensional comprehensive parameters into a single quantitative indicator. At the same time, the arctangent function limits the range of numerical fluctuations to ensure the stability and comparability of the quantification results. Its effectiveness is verified by quantification accuracy tests in actual power consumption information transmission scenarios. The quantification value can accurately distinguish the client-side perception effect under different transmission states, providing an intuitive evaluation basis for transmission optimization. At the same time, the formula structure is simple and the computational complexity is low, making it suitable for the real-time quantification calculation needs of edge platforms.

[0047] Preferred, such as Figure 2 As shown, step S3 includes the following sub-steps: S31, extracting link stability correlation features from the client transmission status perception matrix through the transmission link perception and diagnosis prediction model, matching these features with historical transmission anomaly data, and filtering out a subset of calibrated features with anomaly prediction value; S32, constructing a multi-dimensional prediction vector based on the calibrated feature subset, and performing time series decomposition on the vector through the time series prediction module in the model to separate trend components, periodic components, and random components; S33, substituting the decomposed components into the prediction expression of the transmission link perception and diagnosis prediction model, and generating a preliminary anomaly occurrence probability distribution through matrix operations; S34, performing feature fusion processing on the preliminary probability distribution, removing probability values ​​corresponding to invalid noise data, and outputting the purified link stability correlation parameters and anomaly occurrence probability distribution.

[0048] Specifically, step S3 uses the transmission link awareness diagnosis and prediction model as its core. S31 first uses the model's built-in feature extraction module to select 12 core features directly related to link stability from the client's transmission status awareness matrix. These include parameters such as the link bandwidth occupancy coefficient change rate, data frame transmission delay fluctuation value, and cumulative packet loss rate. Then, it calls the historical transmission anomaly data knowledge base and calculates the matching degree between the current feature and historical anomaly features using a cosine similarity algorithm. A similarity threshold of 0.7 is set, and a subset of features with a matching degree higher than this threshold is selected, ensuring that the subset includes at least 8 valid features. S32, based on the selected core feature subset, a 16-dimensional multi-dimensional prediction vector is constructed. Each dimension of the vector corresponds to the time-series change data of one feature. Then, the model's time-series prediction module uses a sliding window method to decompose the vector into a time series component. The window size is set to 50 data points, decomposing the vector into trend components, periodic components, and random components. The trend component reflects the long-term change pattern of the feature, while the periodic component reflects the long-term change pattern of the feature. The periodic component reflects periodic fluctuations, while the random component represents accidental interference factors. S33 substitutes the three decomposed components into the prediction calculation process of the transmission link sensing diagnosis prediction model with a weight ratio of 4:3:3. A preliminary anomaly probability distribution is generated through matrix multiplication and accumulation operations. This distribution includes continuous values ​​between 0 and 1, corresponding to the probability of anomalies occurring within the next 30 time units. S34 optimizes the preliminary probability distribution using a feature fusion algorithm, eliminating probability values ​​corresponding to invalid noise data caused by the random component, and retaining valid probability data supported by the trend and periodic components. Finally, a purified 10-dimensional link stability correlation parameter and anomaly probability distributions in 5 intervals are output. The parameter values ​​range from 0.1 to 0.9, and the probability distribution intervals are divided by 0.3 and 0.6, clearly defining high, medium, and low risk areas. This provides accurate predictive data support for subsequent load balancing error correction. Throughout the process, data processing latency is controlled within 20 milliseconds to ensure the real-time nature of transmission status prediction.

[0049] Preferred, such as Figure 3 As shown, step S4 includes the following sub-steps: S41, obtaining the anomaly probability distribution output by the transmission link perception diagnosis prediction model and the load data of the current access node of the client, and determining the calibration influence factor of the load distribution deviation; S42, substituting the calibration influence factor into the error correction expression of the client load balancing error correction algorithm, and calculating the initial load balancing error correction amount; S43, dynamically adjusting the initial correction amount based on the real-time transmission status parameters fed back by the edge power acquisition perception calculation analysis platform, and optimizing the dimension weight coefficient in the correction algorithm; S44, generating the corresponding load balancing adjustment coefficient according to the adjusted correction amount, and formulating a data transmission path optimization strategy in combination with the characteristics of the client transmission path.

[0050] Specifically, step S4 applies the client-side load balancing error correction algorithm. S41 first obtains the anomaly probability distribution output by the transmission link perception diagnosis prediction model and the current client access node load data through the algorithm's data receiving module. The load data includes three core indicators for each access node: concurrent tasks, memory usage ratio, and CPU utilization. Principal component analysis is used to determine the core influencing factors of load distribution deviation, resulting in six key factors, including load parameters corresponding to high-dimensional anomaly probabilities and load differences between nodes. S42 sorts the six core influencing factors by weight and substitutes them into the core calculation formula of the client-side load balancing error correction algorithm. Combined with a preset expected load range (20 to 80 per client), the deviation between the actual load and the expected load is calculated. The deviation is calculated to three decimal places to ensure the accuracy of the initial correction calculation. S43 uses edge terminal... The data acquisition and analysis platform collects transmission status parameters in real time, including real-time link bandwidth, data transmission success rate, and client response speed. The data is updated every 5 milliseconds. Based on the updated parameters, the platform dynamically adjusts the dimension weight coefficients in the correction algorithm, keeping the adjustment range within ±0.05 to avoid instability caused by sudden weight changes. The S44 algorithm generates a load balancing adjustment coefficient between 0.05 and 0.5 based on the adjusted value. Each client has a unique adjustment coefficient. Simultaneously, it combines link bandwidth occupancy coefficients, data frame transmission latency parameters, and other link characteristics to allocate dedicated transmission paths for data acquisition of different priorities. The path selection weights for priorities 1 to 5 are 0.5, 0.3, 0.1, 0.07, and 0.03, respectively, ensuring that high-priority data receives better transmission resources. Throughout the entire step-by-step implementation process, the load adjustment response time does not exceed 15 milliseconds, guaranteeing the real-time performance and effectiveness of load balancing.

[0051] Preferred, such as Figure 4 As shown, S5 includes the following sub-steps: S51, collecting the feature mapping results output by the power acquisition transmission anomaly perception and identification network through the parameter receiving module of the edge-end power acquisition perception and analysis platform, and standardizing the data format; S52, receiving the anomaly prediction parameters output by the transmission link perception diagnosis and prediction model, aligning them with the feature mapping results in dimensions, and constructing a preliminary integrated dataset; S53, importing the error correction data generated by the client load balancing error correction algorithm, and performing matrix operations corresponding to the parameter integration expression through the platform's integrated operation module; S54, constructing a client transmission status metric evaluation system including perception features, prediction indicators, and correction parameters based on the operation results, and clarifying the weight allocation rules for each evaluation dimension.

[0052] Specifically, step S5 utilizes the edge-end power acquisition sensing and analysis platform to complete the parameter integration and evaluation system construction in steps S51 to S54. S51 first receives the feature mapping results output by the power acquisition transmission anomaly sensing and identification network through the platform's parameter receiving module. This module supports multi-port parallel reception with a data transmission rate of no less than 1000 megabits per second. After reception, the data format of the client transmission status sensing matrix is ​​standardized, converting all parameter values ​​into normalized values ​​between 0 and 1 to ensure data format uniformity. S52 receives the 10-dimensional link stability correlation parameters and the anomaly probability distribution of 5 intervals output by the transmission link sensing diagnosis prediction model. The dimension of the anomaly prediction parameters is adjusted to 16 dimensions using a dimension alignment algorithm to match the dimension of the feature mapping results. Then, the two types of data are concatenated row by row to construct a preliminary integrated dataset including feature mapping values, stability parameters, and probability distribution values, with a dataset size of 16 rows and 32 columns. S53 imports the load balancing adjustment coefficient generated by the client load balancing error correction algorithm and... Error correction data includes eight items such as adjustment coefficients and load deviation correction values ​​for each client. The platform's integrated computation module performs matrix addition and element-wise multiplication, setting the weights of feature mapping results (0.4), anomaly prediction parameters (0.3), and error correction data (0.3). Fixed-point arithmetic is used to improve computational efficiency, ensuring integration latency is controlled within 10 milliseconds. Based on the integrated parameter matrix, S54 constructs a quantified evaluation system for client transmission status, including perception feature dimensions, anomaly prediction dimensions, and load correction dimensions. The perception feature dimension includes 16 indicators, the anomaly prediction dimension includes 8 indicators, and the load correction dimension includes 12 indicators. The calculation rules and weight allocation for each evaluation dimension are clearly defined: perception feature dimension accounts for 0.35%, anomaly prediction dimension accounts for 0.35%, and load correction dimension accounts for 0.3%, forming a standardized evaluation framework that provides a unified calculation basis for subsequent quantitative output. The entire step-by-step implementation process ensures the integrity of data integration and the scientific nature of the evaluation system.

[0053] like Figure 5As shown, a client-side perception and quantification method for electricity consumption information collection and data transmission is implemented through different units, including: a multi-dimensional electricity collection parameter acquisition unit, which establishes a data transmission connection with the edge-end electricity collection perception and analysis platform to collect link bandwidth occupancy coefficients, data frame transmission delay parameters, and calibration parameters during the client-side electricity consumption information collection and transmission process, and transmits the collected data to the feature mapping processing unit in real time; a feature mapping processing unit, which communicates bidirectionally with the multi-dimensional electricity collection parameter acquisition unit and the transmission anomaly prediction unit, performs nonlinear transformation and feature enhancement on the collected parameters through the electricity collection transmission anomaly perception and identification network, and outputs the client transmission status perception matrix to the transmission anomaly prediction unit; and a transmission anomaly prediction unit, which is connected to the feature mapping processing unit and the load balancing correction unit respectively, and calls the transmission... The link-aware diagnostic prediction model predicts abnormal trends in the client's transmission status perception matrix and sends the prediction results to the load balancing correction unit. The load balancing correction unit interacts with the transmission anomaly prediction unit and the parameter integration and evaluation unit, using a client-side load balancing error correction algorithm to dynamically correct load distribution deviations and outputs the corrected data to the parameter integration and evaluation unit. The parameter integration and evaluation unit communicates with the load balancing correction unit and the quantification result output unit, integrating various parameters through the edge-end power acquisition and perception computing analysis platform to construct a quantified evaluation system, which is then transmitted to the quantification result output unit. The quantification result output unit is unidirectionally connected to the parameter integration and evaluation unit, outputting client-side perception quantified values, transmission link health indicators, and load balancing adaptation parameters based on the evaluation system.

[0054] The formula in this invention integrates different scalar and vector parameters for unified calculation. Its core lies in eliminating parameter type differences through standardization and dimension adaptation mechanisms, while simultaneously building a collaborative correlation logic based on weight allocation and operational rules. Scalar parameters, such as link bandwidth occupancy coefficients and packet loss rate indicators, although single values, are converted into elements matching vector dimensions by multiplying by adaptation coefficients and integrated into matrix operations. Vector parameters, such as client transmission status perception matrices and multi-dimensional prediction vectors, have elements in each dimension corresponding to different transmission status indicators, forming a one-to-one correspondence with scalars through element-wise operations. For example, the scalar link bandwidth occupancy coefficient, in the feature mapping formula of the power acquisition transmission anomaly perception and identification network, participates in linear combination and nonlinear transformation as a matrix element after multiplying with an adjustment coefficient. The vector client transmission status perception matrix, on the other hand, achieves dimension calibration through a weight matrix, ensuring that the elements converted from scalars perfectly match the vector dimensions, guaranteeing compatibility of different parameter types within the same operational framework. Meanwhile, the weighting coefficients and adjustment parameters in the formula are determined based on parameter correlation analysis. For example, the vector data of link stability correlation parameters and anomaly probability distribution are used to form an ordered operation with the load balancing adjustment coefficient in the integrated formula by setting a reasonable weight ratio. This preserves the core information of each parameter and achieves consistency in data fusion.

[0055] Furthermore, the formula design employs multi-level operations to process different types of parameters hierarchically. Through diverse operational forms such as linear transformations, nonlinear constraints, and matrix operations, it adapts to the single-point characteristics of scalars and the multidimensional characteristics of vectors. Simultaneously, normalization maps all parameters to a unified numerical range, eliminating computational conflicts caused by dimensional differences. Taking the prediction formula of the transmission link sensing diagnostic prediction model as an example, the scalar trend adjustment parameter is integrated into the product operation of the vector client transmission status sensing matrix through logarithmic operations. This utilizes the multidimensional characteristics of vectors to comprehensively characterize the transmission status while correcting the range of the calculation results through scalar parameters. In the client load balancing error correction algorithm, the scalar load deviation value and the vector client transmission rate and buffer capacity data form a collaborative operational relationship through weighted and normalized processing. The scalar serves as the core adjustment factor, while the vector provides multidimensional constraints, ultimately achieving the organic integration of different types of parameters. This design ensures the integrity of parameter information and, through the adaptability of operational rules, enables the formula to have unified computational capabilities, meeting the needs of client sensing and quantification for the fusion processing of multi-source heterogeneous parameters in electricity information collection and data transmission.

[0056] A client-side perception and measurement method for electricity consumption information collection and data transmission has been developed, constructing a closed-loop system with multi-technology collaboration. Through the deep integration of dedicated networks, models, algorithms, and platforms, a comprehensive upgrade of client-side perception and measurement for electricity consumption information transmission has been achieved. It relies on a dedicated edge platform to comprehensively collect multi-dimensional core parameters such as link, load, and protocol. Then, an anomaly perception and identification network completes feature mapping and enhancement of multi-source parameters, solving the problem of isolated parameter processing in traditional methods. The transmission link diagnostic prediction model predicts anomaly trends in advance, and the load balancing error correction algorithm dynamically optimizes the allocation of deviations, overcoming the limitations of existing technologies such as delayed anomaly response and lack of targeted adjustments. Each stage achieves organic integration of different types of parameters through standardized processing and dimensional adaptation mechanisms. The constructed measurement and evaluation system can output accurate perception quantification values ​​and link health indicators, ensuring the comprehensiveness and accuracy of transmission status characterization.

[0057] This method employs a multi-parameter systematic acquisition and fusion mechanism, integrating dispersed parameters such as link bandwidth occupancy, transmission latency, and node load into a unified client transmission status perception matrix. This completely changes the one-sidedness of traditional single-parameter monitoring, comprehensively reflecting the integrated characteristics of transmission status. It innovatively introduces an anomaly prediction and dynamic correction linkage scheme, proactively identifying transmission anomaly trends and optimizing load allocation based on client characteristics, solving the problems of passive anomaly handling and unreasonable load adjustment in existing technologies. Relying on a dedicated edge platform, it achieves efficient end-to-end computation, ensuring real-time parameter acquisition, feature processing, and quantization output, adapting to the dynamic changes in complex transmission scenarios. The entire solution achieves closed-loop linkage of each link through a modular system architecture, significantly improving the stability of electricity information transmission and the accuracy of perception quantification, providing scientific and reliable decision support for transmission optimization.

[0058] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A client-side sensing and quantification method for electricity consumption information collection and data transmission, characterized in that, Includes the following steps: S1, through the edge-end power collection and perception computing analysis platform, obtains the link bandwidth occupancy coefficient, data frame transmission delay parameters, data packet loss rate index, client access node load, transmission protocol adaptation parameters and power collection data priority identification data during the power consumption information collection and transmission process of the client. S2, based on the power acquisition and transmission anomaly perception and identification network, performs multi-dimensional feature mapping on the data collected and transmitted in S1, constructs the client transmission status perception matrix, and performs nonlinear transformation and feature enhancement on the client transmission status perception matrix through network hidden layer nodes. S3, call the transmission link perception diagnosis and prediction model to predict the transmission anomaly trend of the enhanced client transmission status perception matrix, and output the link stability correlation parameters and anomaly occurrence probability distribution; S4. The client load balancing error correction algorithm is used to dynamically correct the load distribution deviation parameter in the prediction result, and generate load balancing adjustment coefficient and data transmission path optimization strategy. S5 integrates feature mapping results, anomaly prediction parameters, and error correction data through the edge-end power acquisition and sensing computing analysis platform to construct a quantified evaluation system for client transmission status. S6, based on the evaluation system, outputs client perception quantification values, transmission link health indicators, and load balancing adaptation parameters to form client perception and quantification results for electricity information collection and data transmission.

2. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 1, characterized in that, The feature mapping expression of the power acquisition transmission anomaly sensing and identification network is: ， in, Transmit the state-aware matrix to the client. For network activation functions, This is the network weight matrix. For bias terms, For parameter adjustment coefficients, This is the link bandwidth utilization factor. For data frame transmission delay parameters, This refers to the packet loss rate metric. For client access node load, For transmission protocol adaptation parameters, This is used as a priority identifier for electrical data collection.

3. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 2, characterized in that, The abnormal trend prediction expression of the transmission link sensing diagnosis prediction model is as follows: ， in, For abnormal trend prediction results, The number of feature dimensions. Transmit the state-aware matrix to the client dimensional elements, These are the dimension weight coefficients. These are trend adjustment parameters.

4. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 3, characterized in that, The error correction expression for the client-side load balancing error correction algorithm is as follows: ,in, This is the load balancing error correction amount. This represents the actual load. For the expected load, For the number of clients, For the first Client transmission rate For the first Client cache capacity, To correct the algorithm parameters.

5. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 4, characterized in that, The parameter integration expression of the edge-end power acquisition and sensing computing analysis platform is as follows: ， in, The integrated parameter matrix, Element-wise addition of matrices. Element-wise multiplication of matrices. To integrate the weighting coefficients.

6. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 5, characterized in that, The quantitative output expression of the client transmission state metric evaluation system is as follows: ， in, For the client to perceive metric values, The first calibration dimension element of the composite parameter matrix. This is the second calibration dimension element of the composite parameter matrix. This is the third calibration dimension element of the comprehensive parameter matrix. This is the fourth calibration dimension element of the comprehensive parameter matrix. This is the fifth calibration dimension element of the comprehensive parameter matrix. This is the quantization coefficient.

7. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 1, characterized in that, The S3 includes the following sub-steps: S31, extracting link stability correlation features from the client transmission status perception matrix through the transmission link perception diagnosis prediction model, matching the correlation of the features with historical transmission anomaly data, and filtering out a subset of calibration features with anomaly prediction value; S32, construct a multi-dimensional prediction vector based on the calibrated feature subset, and perform time series decomposition on the vector through the time series prediction module in the model to separate trend components, periodic components and random components; S33, substitute the decomposed components into the prediction expression of the transmission link sensing diagnosis prediction model, and generate a preliminary anomaly occurrence probability distribution through matrix operations; S34, perform feature fusion processing on the preliminary probability distribution, remove the probability values ​​corresponding to invalid noise data, and output the purified link stability correlation parameters and anomaly occurrence probability distribution.

8. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 1, characterized in that, S4 includes the following sub-steps: S41, obtaining the anomaly occurrence probability distribution and the current access node load data of the client output by the transmission link perception diagnosis prediction model, and determining the calibration influence factor of the load distribution deviation; S42, substituting the calibration influence factor into the error correction expression of the client load balancing error correction algorithm, and calculating the initial load balancing error correction amount. S43, Based on the real-time transmission status parameters fed back by the edge power acquisition and sensing computing analysis platform, the initial correction amount is dynamically adjusted, and the dimension weight coefficients in the correction algorithm are optimized; S44, Based on the adjusted correction amount, the corresponding load balancing adjustment coefficient is generated, and a data transmission path optimization strategy is formulated in combination with the characteristics of the client transmission path.

9. The client-side sensing and quantification method for electricity consumption information collection and data transmission according to claim 1, characterized in that, S5 includes the following steps: S51, collecting the feature mapping results output by the power acquisition transmission anomaly perception and identification network through the parameter receiving module of the edge-end power acquisition perception and analysis platform, and standardizing the data format; S52, receiving the anomaly prediction parameters output by the transmission link perception diagnosis and prediction model, aligning them with the feature mapping results in dimensions, and constructing a preliminary integrated dataset; S53, importing the error correction data generated by the client load balancing error correction algorithm, and performing matrix operations corresponding to the parameter integration expression through the platform's integrated operation module; S54, constructing a client transmission status metric evaluation system including perception features, prediction indicators, and correction parameters based on the operation results, and clarifying the weight allocation rules for each evaluation dimension.

Images