A method and system for load rate analysis and evaluation based on multi-source data fusion and deep learning

Abstract

This invention discloses a load rate analysis and evaluation method based on multi-source data fusion and deep learning. The main steps include: first, data preprocessing by collecting the actual and rated load rates of the power system; second, fusing the collected data from multiple data sources to obtain a dataset containing more information; training the dataset using a DBN model to establish a power system load rate prediction model; validating the model using a validation set; predicting the load rate of the power system using the trained model on new data; analyzing the actual load rate of the power system based on the prediction results and external factor data; and updating the model in a timely manner based on the analysis results to improve prediction accuracy and adaptability. Through load rate analysis and evaluation, the safety and stability of the power system can be improved, while ensuring the safe operation of the power system.

Description

Technical Field

[0001] This invention relates to a load rate analysis and evaluation method and system based on multi-source data fusion and deep learning in the field of multi-source data fusion. Background Technology

[0002] Load factor analysis and evaluation are crucial aspects of power system operation and management, primarily used to assess the power load on various parts of the power system. Its purpose is to understand the power system's operating status, predict its operational conditions, improve its operational efficiency and stability, and ensure reliable power supply. The load factor reflects the load condition of the power system and is one of the important indicators for evaluating its economic efficiency and reliability.

[0003] Currently, load rate analysis and evaluation in power systems still face several challenges. Specifically, these challenges include: (1) Diverse data sources and massive data volume: Data in power systems comes from different devices and systems, each with its own data format and storage method. Therefore, load rate analysis and evaluation requires processing multi-source data, which is typically very large and difficult to analyze and process efficiently. (2) Accuracy and precision issues: Due to the complexity and uncertainty of the power system itself, as well as limitations in data quality, the accuracy and precision of current load rate analysis and evaluation still have certain limitations. (3) Insufficient predictive capabilities: Current load rate analysis and evaluation methods are mainly based on historical data for prediction, which is insufficient to cope with sudden events or new changes in the power system.

[0004] To overcome these challenges, researchers are actively exploring and applying new technologies and methods, such as deep learning, big data analytics, and artificial intelligence, to improve the accuracy and efficiency of load rate analysis and evaluation. Simultaneously, a more robust data acquisition and processing system is needed to ensure data accuracy and consistency. A load rate analysis and evaluation method based on multi-source data fusion and deep learning mainly needs to consider two aspects: (1) how to classify large amounts of data into corresponding categories; and (2) how to use input data to complete the overall optimal model building and load rate calculation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a load rate analysis and evaluation method and system based on multi-source data fusion and deep learning, for analyzing and evaluating transformer load rates and predicting transformer load capacity.

[0006] One technical solution to achieve the above objectives is a load rate analysis and evaluation method based on multi-source data fusion and deep learning, comprising the following steps:

[0007] S1: First, collect a large amount of data from the active distribution network, including the actual load data and rated load data of the power system, and obtain external factor data from weather forecasts and holiday schedules. After collecting the data, proceed to step 2.

[0008] S2: Perform data preprocessing on the data from step 1. Data preprocessing includes the following parts:

[0009] S2.1 Data Collection: Collect relevant data of the power system, including generator capacity, load data, line data, and substation data;

[0010] S2.2 Data Cleaning: Cleaning and processing errors, missing values, and outliers in the data to ensure the accuracy and integrity of the data;

[0011] S2.3 Data Conversion: Convert the data into a standard format to facilitate subsequent analysis and processing;

[0012] S2.4 Data Normalization: Normalize the data to eliminate the differences in dimensions between different data, so that different data can be compared and analyzed.

[0013] S2.5 Data Sampling: For large amounts of data, a data sampling method is used to randomly select a portion of the data for analysis and processing, in order to save computing resources and time;

[0014] S2.6: Anonymize the data. To protect data privacy and security, proceed to S3.

[0015] S3: After receiving the anonymous data from step 2, perform correlation verification on the data, and then proceed to S4;

[0016] S4: Classify the data into training set, validation set, and test set, and then proceed to S5;

[0017] S5: Merge multiple data sets from S4 to obtain a dataset containing more information, and proceed to S6;

[0018] S6: Use the data classified in S5 as input to the deep belief network model, and substitute it according to the likelihood function equation:

[0019] Differentiate the above equation again:

[0020] The value of σ is obtained. At this point, σ is the optimal parameter solution of the prediction model. Proceed to S7.

[0021] S7: Validate the model using the validation set, evaluate the model's predictive performance, and proceed to S8;

[0022] S8: The parameters of S7 are corrected by the backpropagation neural network BP algorithm to adjust the network weights and thresholds, and the overall optimal model is completed, and then proceed to S9;

[0023] S9: Based on the data model in S8, use the trained model to predict the new data and predict the ratio of the actual load to the rated load of the power system, i.e., the load factor, and then proceed to S10.

[0024] S10: Based on the forecast results and external factor data, analyze and evaluate the load rate of the power system. This can be displayed using charts and visualization tools to help users better understand and make decisions, and update the model based on the analysis results.

[0025] The load rate analysis and evaluation system based on the above-mentioned multi-source data fusion and deep learning method includes a data processor, a multi-source data fusion unit, a load rate predictor, and an analysis and evaluation processor.

[0026] The data processor collects actual and rated load data of the power system and preprocesses the data, executing steps S1 to S4; the multi-source data fusion unit fuses the collected data to obtain a dataset containing more information, executing step S5; the load rate predictor trains the dataset using a deep belief network model to establish a power system load rate prediction model and predicts the load rate, executing steps S6 to S9; the analysis and evaluation processor analyzes and evaluates the power system load rate based on the prediction results and external factors, updates the model based on the analysis results, and executes step S10.

[0027] Furthermore, the data processor collects actual load data and rated load data of the power system, and obtains external factor data from weather forecasts and holiday schedules. The collected data is preprocessed, including data cleaning, normalization, and feature extraction.

[0028] Furthermore, the multi-source data fusion unit integrates and merges information from different data sources to obtain more accurate and comprehensive load factor analysis and evaluation results: by fusing weather forecast data, it can accurately predict the fluctuation trend of load factor; by fusing historical data, it can conduct a more comprehensive analysis of the long-term trend of load factor; by fusing real-time monitoring data, it can react more quickly to the instantaneous changes in load factor; after multiple analyses, the results are combined and finally input into the deep belief network model.

[0029] Furthermore, the load rate analyzer divides the actual load by the rated load to calculate the load rate of the power system. The closer the load rate is to 1, the higher the utilization efficiency of the power system; conversely, the lower the load rate, the lower the efficiency. Based on the load rate calculation results, combined with the power system's operating conditions and historical data, the factors affecting the load rate are analyzed.

[0030] Furthermore, the analysis and evaluation processor proposes solutions to improve the load factor of the power system based on the analysis results.

[0031] The load rate analysis and evaluation method and system based on multi-source data fusion and deep learning of the present invention can fully mine power grid data information, accurately predict potential hidden dangers in the power system, and ensure the safe and stable operation of the entire power system. Attached Figure Description

[0032] Figure 1 This is a system structure diagram of the load rate analysis and evaluation method;

[0033] Figure 2 This is a data cleaning flowchart;

[0034] Figure 3 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0035] To better understand the technical solution of the present invention, detailed descriptions are provided below through specific embodiments:

[0036] The technical solution of this invention uses a deep learning-based evaluation and prediction algorithm to predict and evaluate the load rate of transformers. Specifically, firstly, the data is preprocessed by collecting the actual and rated load rates of the power system. Secondly, multiple data sources are fused to obtain a dataset containing more information. A DBN model is then used to train the dataset, establishing a power system load rate prediction model. The model is validated using a validation set. The trained model is then used to predict new data, predicting the power system load rate. Based on the prediction results and external factor data, the actual load rate of the power system is analyzed. Based on the analysis results, the model is updated in a timely manner to improve prediction accuracy and adaptability. Through the analysis and evaluation of the load rate, the safety and stability of the power system can be improved, while ensuring the safe operation of the power system.

[0037] Please see Figure 1 The load rate analysis and evaluation system mainly consists of four parts: a data processor, a multi-source data fusion unit, a load rate predictor, and an analysis and evaluation processor. The data processor collects actual and rated load data of the power system and preprocesses the data; the multi-source data fusion unit merges various collected data to obtain a dataset containing more information; the load rate predictor uses a Deep Belief Network (DBN model) to train the dataset, establishes a power system load rate prediction model, and predicts the load rate; the analysis and evaluation processor analyzes and evaluates the power system load rate based on the prediction results and external factors, and updates the model based on the analysis results. A detailed introduction follows:

[0038] Data Processor: Collects actual and rated load data from the power system, and obtains external factor data from multiple sources such as weather forecasts and holiday schedules. Preprocesses the collected data, including data cleaning, normalization, and feature extraction. Since real-time data collected from the power system often contains noise and outliers, these can interfere with load rate analysis and evaluation, leading to errors. Therefore, data preprocessing is necessary before load rate analysis and evaluation to improve data quality and accuracy, resulting in more reliable analysis and evaluation results. Data preprocessing includes the following steps:

[0039] (1) Data collection: First, it is necessary to collect relevant data of the power system, including generator capacity, load data, line data, substation data, etc.

[0040] (2) Data cleaning: Cleaning and processing errors, missing values ​​and outliers in the data to ensure the accuracy and integrity of the data.

[0041] (3) Data conversion: Convert the data into a suitable format to facilitate subsequent analysis and processing. For example, time data can be converted into a standard time format, and power data can be converted into the corresponding units (such as kilowatts and megawatts).

[0042] (4) Data normalization: Normalize the data to eliminate the differences in units between different data, so that different data can be compared and analyzed. For example, methods such as maximum-minimum normalization or standard deviation normalization can be used.

[0043] (5) Data sampling: If the amount of data is large, a data sampling method can be used to randomly select a portion of the data for analysis and processing, in order to save computing resources and time.

[0044] First, all data undergoes anonymous auditing, and all data is analyzed uniformly to evaluate data that does not meet standards or is abnormal. Anonymization is achieved by anonymizing data IDs without altering interrelated data, and different processing methods are applied to different data types. Precise authorization is granted to users accessing the data, preventing unauthorized access. Access is logged, and information on sensitive data access is analyzed. This approach not only allows for the extensive use of non-sensitive data but also maximizes the prevention of data privacy leaks and strengthens data management.

[0045] Data quality assessment is a crucial step in ensuring the accuracy, completeness, consistency, and reliability of data, thereby guaranteeing its suitability for subsequent analysis and decision-making. Offline specialized profiling tools are essential for data quality assessment, enabling full-field, full-record data profiling of data tables to identify data quality issues and provide solutions.

[0046] At the data structure level, data quality assessment primarily analyzes the data's value range and distribution, data type, and data format to identify the data's standardization and accuracy. At the data content level, it automatically analyzes and visually displays data entity and attribute information to identify data completeness and consistency. At the data relationship level, it analyzes important data characteristics such as the data values ​​and distribution, data format and distribution, fill value range, fill value length, fill value dispersion, and null value rate of each field to identify the data's reliability and relevance.

[0047] Taking main transformer data quality analysis as an example, data quality assessment can be conducted on main transformer ledger information tables, transformer oil chromatography tables, and transformer core ammeters to display data quality indicators such as data correlation rate, data non-standardization rate, data outlier rate, and data non-empty value rate, thereby identifying data quality issues. Simultaneously, data quality assessment can also be used to correlate equipment codes, displaying the number of matching devices and the number of matched monitoring devices for main transformers to identify data completeness and consistency. Furthermore, it can display a list of main transformers without matching devices and a list of main transformers without matching monitoring data, providing a basis for data governance.

[0048] Data cleaning is a process of identifying and correcting identifiable errors in data. Specifically, it involves checking for missing data, data validity, and data consistency. Data cleaning primarily deals with incomplete, erroneous, and duplicate data. Incomplete data mainly refers to missing information, such as missing categories, customer names, or company departments. This type of missing data is completed within a specified timeframe before being entered into the database. For erroneous data, such as incorrect date formats or character encoding, these errors are identified by the database using SQL and handed over to the relevant departments for correction before data extraction. Duplicate data is repeatedly filtered and verified according to different filtering rules. After data cleaning, the data is stored in the platform database, and finally, the data is categorized for data mining and utilization.

[0049] The main approaches and processes of data cleaning are as follows: Figure 2 As shown.

[0050] The data cleaning mechanism first outlines the cleaning rules, enabling batch import and export of data. It unifies cleaning rules and centrally manages data, including completing incomplete data, correcting erroneous data, and deleting duplicate data. It provides separate cleaning solutions for different scenarios within the power system, significantly improving data accuracy. Furthermore, it automatically sets the cleaning time, effectively completing data cleaning for all scenarios within the specified timeframe. The cleaned data is clear and readily visible, providing feedback on the overall data cleaning status and better assisting users in classifying data according to different dimensions, enabling faster classification of training, validation, and test sets.

[0051] Multi-source data fusion unit: This unit integrates and merges information from different data sources to obtain more accurate and comprehensive load factor analysis and evaluation results. By fusing weather forecast data, it can more accurately predict load factor fluctuations; by fusing historical data, it can provide a more comprehensive analysis of long-term load factor trends; and by fusing real-time monitoring data, it can react more quickly to instantaneous changes in load factor. After various analyses, the results are combined and finally input into the DBN model.

[0052] Multi-source data fusion modules integrate data from different sensors, devices, and systems to obtain more comprehensive and accurate load factor information. This data may include parameters such as power load, transmission line current and voltage, and weather data. Analyzing this data allows for a better understanding of the power system's operating status, identification of problems and risks, and timely implementation of corrective measures. Specifically, the functions of multi-source data fusion modules include:

[0053] (1) Improve the accuracy of load rate: The multi-source data fusion module can integrate data from different data sources, such as electricity load, weather forecast, generator operation status, etc., thereby improving the accuracy of load rate analysis.

[0054] (2) Real-time monitoring of power system operation status: The multi-source data fusion module can monitor the power system operation status in real time, detect abnormal situations in a timely manner, and help operation and management personnel make corresponding decisions.

[0055] (3) Optimize power system operation: The multi-source data fusion module can analyze the relationship between different elements in the power system, provide optimization suggestions, and help the power system operate more efficiently and reliably.

[0056] (4) Support for intelligent decision-making: The multi-source data fusion module can use artificial intelligence technology to analyze the operation status of the power system, thereby supporting operation and management personnel to make intelligent decisions.

[0057] Multi-source data fusion modules play an important role in load rate analysis and evaluation in power systems. After data preprocessing, the fusion of all data results in a dataset containing more information, which can improve the efficiency and reliability of power system operation.

[0058] Load factor analyzer: The load factor of a power system is calculated by dividing the actual load by the rated load. A load factor closer to 1 indicates higher power system efficiency, while a value further away indicates lower efficiency. Based on the calculated load factor, combined with the power system's operating conditions and historical data, factors affecting the load factor are analyzed. The load forecaster uses a deep learning model to train a dataset to build a power system load factor prediction model. The model is then validated using a validation set to evaluate its predictive performance. Finally, the trained model is used to predict new data, forecasting the ratio of the actual load to the rated load of the power system, i.e., the load factor.

[0059] Analysis and Evaluation Processor: Based on the analysis results, propose solutions to improve the power system load factor. Possible improvements include optimizing load dispatch strategies, increasing power equipment capacity, and encouraging energy conservation and emission reduction. The implementation effectiveness of the improvement solutions is evaluated, the model is updated in a timely manner, and the power system load factor is analyzed and evaluated based on forecast results and external factor data. Charts and visualization tools can be used to present the results, helping users better understand and make decisions, thereby improving forecast accuracy and adaptability. The processor then assesses whether the expected results have been achieved.

[0060] The DBN model is built using the Deep Belief Network (DBN) deep learning model. By employing a layer-by-layer training approach, it addresses the optimization problem of deep neural networks. Layer-by-layer training assigns the entire network relatively good initial weights, allowing the network to reach the optimal solution with only minor adjustments. The Restricted Boltzmann Machine (RBM) plays a crucial role in this layer-by-layer training process.

[0061] DBN models learn feature representations of data by stacking multiple RBM layers. Each RBM layer consists of a visible layer and a hidden layer, where the visible layer receives the raw input data and the hidden layer extracts features. In DBN, the hidden layers of each RBM layer are also called feature detectors because they learn high-level features of the input data.

[0062] The DBN model is trained layer by layer, starting with the bottommost RBM layer and training it with the raw input data. Then, the hidden features of this RBM layer are used as input to the top RBM layer, and training continues. This process is repeated until all RBM layers have been trained. This layer-by-layer training process allows the model to progressively learn complex feature representations of the data and avoids the vanishing gradient problem common in deep neural networks.

[0063] Once a DBN model has been trained or converged to a stable state, it can be used to generate new data. This is because, by learning a probabilistic model of the data, DBN can generate new samples similar to the original data. This capability is extremely useful for addressing the problem of insufficient sample quantity.

[0064] In a Restricted Boltzmann Machine (RBM), training samples are used as input. By changing the parameter σ, the consistency of the probability distribution and the distribution curve of the RBM training data affected by σ are maximized. The maximum likelihood function value is then obtained, and the likelihood function equation is as follows:

[0065]

[0066] Differentiate equation (1):

[0067]

[0068] Setting the derivative to 0, we obtain the value of σ, which is the optimal parameter solution for the prediction model. In practice, the input data is more complex than the theoretical calculations, so a top-down, layer-by-layer training approach is needed. After establishing the model, the threshold and weights are refined using a backpropagation (BP) neural network algorithm to optimize the computation. The model is validated using a validation set to evaluate its predictive performance. The trained model is then used to predict new data, taking into account various external factors, to predict the ratio of the actual load to the rated load of the power system, i.e., the load factor.

[0069] The load rate analysis and evaluation method based on multi-source data fusion and deep learning of the present invention includes the following steps:

[0070] S1: First, collect a large amount of data from the active distribution network, including the actual load data and rated load data of the power system, and obtain external factor data from weather forecasts and holiday schedules. After collecting the data, proceed to step 2.

[0071] S2: Perform data preprocessing on the data from step 1. Data preprocessing includes the following parts:

[0072] S2.1 Data Collection: Collect relevant data of the power system, including generator capacity, load data, line data, and substation data;

[0073] S2.2 Data Cleaning: Cleaning and processing errors, missing values, and outliers in the data to ensure the accuracy and integrity of the data;

[0074] S2.3 Data Conversion: Convert the data into a standard format to facilitate subsequent analysis and processing;

[0075] S2.4 Data Normalization: Normalize the data to eliminate the differences in dimensions between different data, so that different data can be compared and analyzed.

[0076] S2.5 Data Sampling: For large amounts of data, a data sampling method is used to randomly select a portion of the data for analysis and processing, in order to save computing resources and time;

[0077] S2.6: Anonymize the data. To protect data privacy and security, proceed to S3.

[0078] S3: After receiving the anonymous data from step 2, perform correlation verification on the data, and then proceed to S4;

[0079] S4: Classify the data into training set, validation set, and test set, and then proceed to S5;

[0080] S5: Merge multiple data sets from S4 to obtain a dataset containing more information, and proceed to S6;

[0081] S6: Use the data classified in S5 as input to the deep belief network model, and substitute it according to the likelihood function equation:

[0082] Differentiate the above equation again:

[0083] The value of σ is obtained. At this point, σ is the optimal parameter solution of the prediction model. Proceed to S7.

[0084] S7: Validate the model using the validation set, evaluate the model's predictive performance, and proceed to S8;

[0085] S8: The parameters of S7 are corrected by the backpropagation neural network BP algorithm to adjust the network weights and thresholds, and the overall optimal model is completed, and then proceed to S9;

[0086] S9: Based on the data model in S8, use the trained model to predict the new data and predict the ratio of the actual load to the rated load of the power system, i.e., the load factor, and then proceed to S10.

[0087] S10: Based on the forecast results and external factor data, analyze and evaluate the load rate of the power system. This can be displayed using charts and visualization tools to help users better understand and make decisions, and update the model based on the analysis results.

[0088] For the safe operation and control of power systems, the system load rate is the allowed continuous load time of equipment within a selected working period. In other words, the load rate effectively and clearly shows the system's operating time, achieving maximum operating efficiency while maintaining its safety. Many factors influence the load rate, such as weather, holiday schedules, and other external factors. The load rate of the power system can be predicted by establishing a DBN model, and then combined with external factor data analysis to evaluate the load rate, thereby fully utilizing the power system's operational capacity while ensuring safe operation.

[0089] The specific implementation plan is as follows:

[0090] (1) First, collect the actual load data and rated load data of the power system, and obtain external factor data from multiple sources such as weather forecasts and holiday arrangements.

[0091] (2) The collected data is then preprocessed, including data cleaning, normalization, and feature extraction. Multiple data sources are merged to obtain a dataset containing more information. Finally, the data is classified into: training set, validation set, and test set.

[0092] (3) Then, the classified data is put into the DBN model for simulation calculation. Finally, the optimal parameter solution is obtained. The model is validated using the validation set to evaluate the predictive performance of the model. The trained model is then used to predict the new data to predict the ratio of the actual load to the rated load of the power system, i.e., the load factor.

[0093] (4) Finally, based on the forecast results and external factor data, analyze and evaluate the load factor of the power system. Charts and visualization tools can be used for presentation, and the model should be updated in a timely manner based on the analysis results to improve forecast accuracy and adaptability.

[0094] Those skilled in the art should recognize that the above embodiments are merely illustrative of the present invention and are not intended to limit the present invention. Any variations or modifications to the above embodiments that are within the spirit and essence of the present invention will fall within the scope of the claims of the present invention.

Claims

1. A load rate analysis and evaluation method based on multi-source data fusion and deep learning, characterized in that, Includes the following steps: S1: First, collect a large amount of data from the active distribution network, including the actual load data and rated load data of the power system, and obtain external factor data from weather forecasts and holiday schedules. After collecting the data, proceed to step 2. S2: Perform data preprocessing on the data from step 1. Data preprocessing includes the following parts: S2.1 Data Collection: Collect relevant data of the power system, including generator capacity, load data, line data, and substation data; S2.2 Data Cleaning: Cleaning and processing errors, missing values, and outliers in the data to ensure the accuracy and integrity of the data; S2.3 Data Conversion: Convert the data into a standard format to facilitate subsequent analysis and processing; S2.4 Data Normalization: Normalize the data to eliminate the differences in units between different data, so that different data can be compared and analyzed; S2.5 Data Sampling: For large amounts of data, a data sampling method is used to randomly select a portion of the data for analysis and processing, in order to save computing resources and time; S2.6: Anonymize the data. To protect data privacy and security, proceed to S3. S3: After receiving the anonymous data from step 2, perform correlation verification on the data, and then proceed to S4; S4: Classify the data into training set, validation set, and test set, and then proceed to S5; S5: Merge multiple data sets from S4 to obtain a dataset containing more information, and proceed to S6; S6: Use the data classified in S5 as input to the deep belief network model, and substitute it according to the likelihood function equation: Differentiate the above equation again: The value of σ is obtained. At this point, σ is the optimal parameter solution of the prediction model. Proceed to S7. S7: Validate the model using the validation set, evaluate the model's predictive performance, and proceed to S8; S8: The parameters of S7 are corrected by the backpropagation neural network BP algorithm to adjust the network weights and thresholds, and the overall optimal model is completed, and then proceed to S9; S9: Based on the data model in S8, use the trained model to predict the new data and predict the ratio of the actual load to the rated load of the power system, i.e., the load factor, and then proceed to S10. S10: Based on the forecast results and external factor data, analyze and evaluate the load rate of the power system. This can be displayed using charts and visualization tools to help users better understand and make decisions, and update the model based on the analysis results.

2. A load factor analysis and evaluation system based on the load factor analysis and evaluation method as described in claim 1, characterized in that, Includes data processors, multi-source data fusion units, load factor predictors, and analytics and evaluation processors; The data processor collects actual and rated load data of the power system and preprocesses the data, executing steps S1 to S4; the multi-source data fusion unit fuses the collected data to obtain a dataset containing more information, executing step S5; the load rate predictor trains the dataset using a deep belief network model to establish a power system load rate prediction model and predicts the load rate, executing steps S6 to S9; the analysis and evaluation processor analyzes and evaluates the power system load rate based on the prediction results and external factors, updates the model based on the analysis results, and executes step S10.

3. The load factor analysis and evaluation system according to claim 2, characterized in that, The data processor collects actual load data and rated load data of the power system, and obtains external factor data from weather forecasts and holiday schedules. It preprocesses the collected data, including data cleaning, normalization, and feature extraction.

4. The load factor analysis and evaluation system according to claim 2, characterized in that, Multi-source data fusion merging integrates and combines information from different data sources to obtain more accurate and comprehensive load factor analysis and evaluation results: by fusing weather forecast data, it can accurately predict the fluctuation trend of load factor. By integrating historical data, a more comprehensive analysis of the long-term trend of load factor can be conducted; By integrating real-time monitoring data, the system can react more quickly to instantaneous changes in load rate; after various analyses, the results are combined and finally input into the deep belief network model.

5. The load factor analysis and evaluation system according to claim 2, characterized in that, The load rate analyzer calculates the load rate of the power system by dividing the actual load by the rated load. The closer the load rate is to 1, the higher the utilization efficiency of the power system, and vice versa. Based on the load rate calculation results, combined with the power system's operating conditions and historical data, the factors affecting the load rate are analyzed.

6. The load factor analysis and evaluation system according to claim 2, characterized in that, Based on the analysis results, the analysis and evaluation processor proposes solutions to improve the load factor of the power system.

Images