A power system peak load demand assessment system

Abstract

The present invention discloses a power system peak-shaving demand assessment system, comprising a collection layer, a transmission layer, an edge cloud layer and a management platform; the collection layer is used to collect data; the transmission layer is used to transmit the data collected by the collection layer; the edge cloud layer is used to receive and process data, and the processing of data by the edge cloud layer includes preprocessing and analysis processing; the management platform includes a prediction model, an evaluation model and a visualization module; the present invention quickly processes data through the edge cloud layer to reduce the calculation pressure of the management platform; the management platform includes a prediction model, which predicts power generation through the prediction model, and the management platform includes an evaluation model, which evaluates the peak-shaving process through the evaluation model, and the evaluation and prediction can correspond to each other, so as to facilitate timely regulation when the peak-shaving power is insufficient, and maintain the accurate control of the power in the peak-shaving process.

Description

Technical Field

[0001] The present invention relates to the technical field of power system peak regulation, and in particular to a power system peak regulation demand assessment system. Background Art

[0002] With the transformation of the global energy structure, especially the rapid development of renewable energy such as wind power and solar energy, the operation of the power system is facing new challenges. The traditional power dispatching method relies on stable and predictable power generation sources (such as thermal power, hydropower, etc.), but the power generation characteristics of renewable energy are intermittent, volatile and unadjustable, which brings great uncertainty to the stability and reliability of the power system.

[0003] Since the power load is uneven, the power grid is often overloaded during peak power consumption. At this time, it is necessary to put in generator sets outside normal operation to meet the demand. These generator sets are called peak-shaving units. Because they are used to adjust the peak power consumption, they are called peak-shaving units. The requirements for peak-shaving units are that they are easy to start and stop, and that synchronization adjustment is easy when connected to the grid. Generally, peak-shaving units include gas turbine units and pumped storage units, etc.

[0004] Electricity cannot be stored in large quantities. The generation and use of electricity are synchronized, so the power generation department must generate the same amount of electricity as required. The power load in the power system changes frequently. In order to maintain the balance of power consumption and keep the system frequency stable, the power generation department needs to change the output of the generator accordingly to adapt to the changes in power load. This is called peak load regulation.

[0005] With the rapid development of information technology and data analysis technology, especially the widespread application of technologies such as big data, artificial intelligence and the Internet of Things, the dispatching mode and demand assessment of power systems are also undergoing profound changes. Through real-time monitoring and analysis of multi-dimensional information such as power system operation data, meteorological data, and load data, more accurate peak demand assessment and prediction can be achieved.

[0006] In summary, with the rapid development of renewable energy, the peak load demand assessment of power systems has become increasingly complex. Traditional peak load methods and assessment methods face many challenges, and new technologies (such as big data, artificial intelligence, etc.) provide a possible solution to this problem. Developing an efficient and accurate peak load demand assessment system for power systems can not only effectively improve the efficiency of power dispatching, but also help to further optimize the operation of the power system and promote the transformation and sustainable development of the energy structure.

[0007] Although the power system peak demand assessment system based on big data can evaluate power dispatch, it also faces a series of challenges. First, the evaluation results of power dispatch cannot be used again, resulting in the inability to display the corresponding effect of the power dispatch process, and the inability to achieve accurate dispatch standards for power dispatch, resulting in the falsity of power dispatch and the inability to achieve accurate dispatch management.

[0008] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0009] In view of the problems in the related technology, the present invention proposes a power system peak load demand assessment system to overcome the above technical problems existing in the existing related technology.

[0010] To this end, the specific technical solution adopted by the present invention is as follows:

[0011] A power system peak load demand assessment system, comprising a collection layer, a transmission layer, an edge cloud layer and a management platform;

[0012] The collection layer is used to collect data, and the data collected by the collection layer includes historical load data, equipment data, economic activity data and evaluation data for predicting and determining power generation, as well as substation power data, complaint call data and predicted power generation data for evaluating power peak load regulation;

[0013] The transmission layer is used to transmit the data collected by the collection layer. The historical load data is transmitted through the 5G network, the equipment data is transmitted through the Internet of Things, the economic activity data is transmitted through the 5G network, and the substation power data and the complaint call data are transmitted through the 5G network;

[0014] The edge cloud layer is used to receive and process data. The edge cloud layer receives the historical load data, the equipment data, the economic activity data, the substation power data and the complaint call data through the Internet of Things and the 5G network. The processing of data by the edge cloud layer includes preprocessing and analysis processing. The preprocessing is used to clean, convert and fuse the data, and the analysis processing is used to extract features and determine trends of the preprocessed data;

[0015] The management platform includes a prediction model, an evaluation model and a visualization module. The prediction model predicts the peak-shaving power generation by analyzing and processing the historical load data, the equipment data and the economic activity data, and uses the evaluation data calculated by the evaluation model as update data, and inputs the evaluation data into the prediction model to achieve timely adjustment and control of the predicted power generation. The evaluation model evaluates the peak-shaving process by analyzing and processing the substation power data and the complaint telephone data, and uses the predicted power generation data calculated by the prediction model as update data, and inputs the predicted power generation data into the evaluation model. The prediction power generation data provides data support for the evaluation model. The visualization module establishes a visualization interface for power generation equipment through digital twin technology, which is convenient for integrating and displaying the actual power generation data produced by the power generation equipment, the predicted power generation data and the evaluation data.

[0016] As a preferred embodiment, the historical load data is extracted from a data repository, the equipment data is collected and acquired from various types of power generation equipment through sensors, the economic activity data is used to collect data on activities at various commercial or production sites, the substation power data is used to collect data on the power supply voltage of each substation, mainly used to collect the load voltage for peak load demand, and the complaint call data is used to count the number of complaint calls;

[0017] The predicted power generation data is data obtained by calculation and analysis of the prediction model, and the evaluation data is data obtained by calculation and analysis of the evaluation model, and the predicted power generation data and the evaluation data are respectively used as update data of the evaluation model and update data of the prediction model.

[0018] As a preferred embodiment, the transport layer includes the Internet of Things and a 5G network, the Internet of Things is used to upload data collected by various sensors, the 5G network is used to transmit system data to the edge cloud layer, and the transport layer also includes a firewall and a data gateway;

[0019] The firewall is mainly used to protect the network from attacks and unauthorized access, and to implement traffic monitoring and security policies;

[0020] The data gateway is used to exchange, convert and manage data between different systems or applications to ensure correct data transmission and protocol compatibility.

[0021] As a preferred embodiment, the cleaning is used to remove noise from the data, process missing data, remove duplicate data, process outliers and clean up text data;

[0022] The noise removal is used to filter out the noise in the data. The noise removal assigns different weights to different data through weighted sliding average, so that more recent data is considered when smoothing the data;

[0023] The missing data processing is used to fill in the missing values ​​in the data, and the mean is used to fill in the missing data;

[0024] The deduplication is used to remove duplicate data in the data;

[0025] The outlier processing is used to detect outliers in the data, that is, to identify outliers through Z-Score, and to delete the detected outliers, and to fill the deleted outliers with linear interpolation;

[0026] The cleaned text data is used to process part of the text data, remove redundant spaces, punctuation marks and special characters for fields containing text, unify the case, or perform stem extraction and word form restoration.

[0027] As a preferred embodiment, the analysis and processing generates a data line graph from the data, and calculates the slope between two points in the data line graph to obtain slope data, and then generates a slope line graph from the slope data, and determines the direction of the data according to the inclination direction of the slope line graph, and determines the peak and valley points and high peak points of the data according to the inclination direction of the slope line graph, that is, if the slope line graph tilts upward and then tilts downward, the endpoint of the line of the slope line graph is the peak point, and if the slope line graph tilts downward and then tilts upward, the endpoint of the line of the slope line graph is the peak and valley point, and high peak point data and peak and valley point data are obtained respectively.

[0028] As a preferred embodiment, the feature extraction is used to extract feature information from the data, and the extracted features include mean, standard deviation, variance, maximum and minimum values, skewness and kurtosis;

[0029] Then, the feature data extracted from historical load data, equipment data and economic activity data or substation power data and complaint call data are processed by combined feature processing, and multiple features are combined into a new feature, that is, combined features are generated through feature crossing technology, and peak point data and peak valley point data are integrated into the combined features.

[0030] As a preferred embodiment, the mean is calculated as follows:

[0031] ,

[0032] in, is the mean of the previous data, is the mean of the new data, is the amount of data;

[0033] And the calculation formula of the mean is as follows:

[0034] ,

[0035] in, is the calculated mean, is the weight value of the data, is the data value;

[0036] The standard deviation is calculated as follows:

[0037] ,

[0038] in, is the mean, is the data value, is the amount of data, is the calculated standard deviation;

[0039] The variance is calculated as follows:

[0040] ,

[0041] in, is the mean, is the data value, is the amount of data, is the calculated variance;

[0042] The maximum and minimum values ​​are calculated as follows:

[0043] ,

[0044] in, is the new maximum value after calculation, is the previous maximum value, is the new minimum value after calculation, is the previous minimum value, For newly added data;

[0045] The calculation formula of the skewness is as follows:

[0046] ,

[0047] in, It represents the skewness value obtained after calculation. is the amount of data, is the second center distance, which is the variance, is the third central moment, reflecting the skewness of the data distribution. is the new second central moment, is the second center distance before, is the new third central moment, is the previous third center distance;

[0048] The calculation formula of the kurtosis is as follows:

[0049] ,

[0050] in, It represents the kurtosis value obtained after calculation. It is the fourth central moment, which indicates the sharpness of the data distribution and the thickness of the tail. The fourth central moment calculation reflects the thickness of the tail of the data distribution and the sharpness of the distribution. The kurtosis of the standard normal distribution is 3. After subtracting 3, if the kurtosis is greater than 0, it means that the distribution is sharper or the tail is thicker than the normal distribution; if the kurtosis is less than 0, it means that the distribution is flatter than the normal distribution. is the new fourth center distance, It is the fourth center distance before.

[0051] As a preferred embodiment, the feature crossover technology first fuses the mean, standard deviation, variance, maximum and minimum values, skewness, kurtosis, peak point data and peak valley point data in the historical load data, equipment data, economic activity data, substation power data and complaint call data to form a historical load data feature set. , equipment dataset , Economic Activity Dataset , Substation electricity data set and complaint call dataset ;

[0052] The historical load data feature set used to predict power generation , device data , economic activity data Substation power data set fused through feature cross-talk technology and evaluated for power peak load regulation and complaint call dataset ;

[0053] ,

[0054] ,

[0055] The feature sets for power generation prediction are obtained respectively and power peak load evaluation

[0056] The calculation formula of multi-layer perceptron is:

[0057] ,

[0058] in, is the weight matrix of the fully connected layer, is the bias term, is the activation function, It is the intermediate representation of each layer, the cross feature Further learning through the fully connected layer, and finally output the prediction result .

[0059] As a preferred embodiment, the algorithm used in the prediction model is ridge regression, which is used to predict continuous numerical outputs and predict a continuous target variable based on input data;

[0060] ,

[0061] in, represents the loss function in ridge regression, It means the The actual observed value of samples, It means the The predicted value of samples, It means the The squared error of samples, It represents the sum of squared errors of the samples. It represents the regularization parameter, which controls the influence of the regularization term on the model. It represents the square of the L2 norm of the regression coefficient. is the number of input variables in the dataset, It represents the L2 regularization term, which penalizes the size of the regression coefficient to avoid overfitting. represents the regression coefficient vector.

[0062] As a preferred embodiment, the algorithm adopted by the evaluation model is random forest:

[0063] Set each tree to output a prediction value, and the final prediction result is the average of the prediction values ​​of all trees:

[0064] ,

[0065] in, is the prediction result of random forest, is the total number of trees, It is Tree pair samples The prediction is the prediction result of the power peak load output by the multi-layer perceptron. It represents the output of the multi-layer perceptron. The evaluation and prediction results of each;

[0066] Each tree will give a category label, and the final prediction result is the result of voting by all trees, and the classification result is obtained by majority voting:

[0067] ,

[0068] in, is the final category prediction, It is Tree pair samples The prediction mode is the mode operation, that is, taking the category with the most occurrences.

[0069] The beneficial effects of the present invention are:

[0070] 1. When the present invention is used, data is collected through the collection layer, and the data collected by the collection layer is uploaded through the transmission layer, and the data is transmitted to the edge cloud layer, and the data is pre-processed and analyzed by the edge cloud layer. The processing of data by the edge cloud layer can reduce the pressure of data analysis and the complexity of data analysis. The edge cloud layer pre-processes the data including cleaning, conversion and fusion, and the analysis includes feature extraction and trend determination. The data processing is completed, which can reduce the difficulty of data processing by the management platform;

[0071] 2. When the present invention is used, the management platform includes a prediction model, and the prediction model is used to predict the peak-shaving power generation based on the processed historical load data, equipment data and economic activity data. The management platform includes an evaluation model, and the evaluation model is used to evaluate the peak-shaving process based on the analyzed and processed substation power data and complaint call data. That is, the management platform processes data through the prediction model and the evaluation model, which is convenient for realizing the management and control of power peak-shaving, effectively realizing the prediction of peak-shaving power, generating corresponding power for peak-shaving management, and calculating and analyzing the results of peak-shaving power through the collected data, so as to facilitate the corresponding adjustment of the predicted power through the evaluation data;

[0072] 3. When the present invention is used, the evaluation data is used as update data, and the evaluation data is input into the prediction model to realize timely adjustment and control of the predicted power generation, and the predicted power generation data is used as update data, and the predicted power generation data is input into the evaluation model to realize comparison and judgment of the peak-shaving demand and the predicted power generation data, so that the evaluation and the prediction can correspond to each other, which is convenient for timely regulation when the peak-shaving power is insufficient, and maintains the accurate control of the power in the peak-shaving process;

[0073] 4. When the present invention is used, the visualization module in the management platform establishes a visualization interface of the power generation equipment through the digital twin technology, which is convenient for integrating the actual power generation data, predicted power generation data and evaluation data produced by the power generation equipment on the digital twin model for display, so that the peak power information can be visualized, which is convenient for background analysis and processing;

[0074] In summary, the present invention reduces the computing pressure of the management platform by collecting a large amount of data and quickly processing the data through the edge cloud layer; and the edge cloud layer preprocesses the data including cleaning, conversion and fusion, and the analysis and processing includes feature extraction and trend determination to complete the data processing, which can reduce the difficulty of data processing by the management platform; at the same time, the management platform includes a prediction model, which predicts the power generation through the prediction model, and the management platform includes an evaluation model, which evaluates the peak-shaving process through the evaluation model, that is, the management platform processes the data through the prediction model and the evaluation model, which is convenient for realizing the management and control of power peak-shaving, and the evaluation and prediction can correspond to each other, which is convenient for timely regulation when the peak-shaving power is insufficient, and maintains the accuracy of power control during the peak-shaving process; a visual interface of the power generation equipment is established through the digital twin technology, which is convenient for integrating the actual power generation data, predicted power generation data and evaluation data produced by the power generation equipment on the digital twin model for display, so that the information of the peak-shaving power can be visualized, which is convenient for the background to analyze and process. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0076] Figure 1 4 is a schematic diagram of the structure of a power system peak load demand assessment system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0077] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0078] According to an embodiment of the present invention, a power system peak load demand assessment system is provided.

[0079] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, a power system peak load demand assessment system according to an embodiment of the present invention includes a collection layer, a transmission layer, an edge cloud layer, and a management platform;

[0080] The collection layer is used to collect data, and the data collected by the collection layer includes historical load data, equipment data, economic activity data and evaluation data for predicting and determining power generation, as well as substation power data, complaint call data and predicted power generation data for evaluating power peak load regulation;

[0081] The transmission layer is used to transmit the data collected by the collection layer. The historical load data is transmitted through the 5G network, the equipment data is transmitted through the Internet of Things, the economic activity data is transmitted through the 5G network, and the substation power data and the complaint call data are transmitted through the 5G network;

[0082] The edge cloud layer is used to receive and process data. The edge cloud layer receives the historical load data, the equipment data, the economic activity data, the substation power data and the complaint call data through the Internet of Things and the 5G network. The processing of data by the edge cloud layer includes preprocessing and analysis processing. The preprocessing is used to clean, convert and fuse the data, and the analysis processing is used to extract features and determine trends of the preprocessed data;

[0083] The management platform includes a prediction model, an evaluation model and a visualization module. The prediction model predicts the peak-shaving power generation by analyzing and processing the historical load data, the equipment data and the economic activity data, and uses the evaluation data calculated by the evaluation model as update data, and inputs the evaluation data into the prediction model to achieve timely adjustment and control of the predicted power generation. The evaluation model evaluates the peak-shaving process by analyzing and processing the substation power data and the complaint telephone data, and uses the predicted power generation data calculated by the prediction model as update data, and inputs the predicted power generation data into the evaluation model. The prediction power generation data provides data support for the evaluation model. The visualization module establishes a visualization interface for power generation equipment through digital twin technology, which is convenient for integrating and displaying the actual power generation data produced by the power generation equipment, the predicted power generation data and the evaluation data.

[0084] In this embodiment, preferably, the historical load data is extracted from a data repository, the equipment data is collected and acquired from various types of power generation equipment through sensors, the economic activity data is used to collect data on activities of various commercial or production sites, the substation power data is used to collect data on the power supply voltage of each substation, mainly used to collect the load voltage of peak load demand, and the complaint phone data is used to count the number of complaint calls;

[0085] The predicted power generation data is data obtained by calculation and analysis of the prediction model, and the evaluation data is data obtained by calculation and analysis of the evaluation model, and the predicted power generation data and the evaluation data are used as update data of the evaluation model and update data of the prediction model respectively;

[0086] It should be noted that by collecting a variety of data, it is convenient to predict and generate peak-shaving power, and to evaluate the power during the peak-shaving process through the data. In order to enable the interaction between the prediction data and the evaluation data, the predicted power generation data and the evaluation data are input into the evaluation model and the prediction model respectively, so as to complete the linkage between the prediction and the evaluation and realize the corresponding regulation and control of the peak-shaving power.

[0087] In this embodiment, preferably, the transport layer includes the Internet of Things and a 5G network, the Internet of Things is used to upload data collected by various sensors, the 5G network is used to transmit system data to the edge cloud layer, and the transport layer also includes a firewall and a data gateway;

[0088] The firewall is mainly used to protect the network from attacks and unauthorized access, and to implement traffic monitoring and security policies;

[0089] The data gateway is used to exchange, convert and manage data between different systems or applications to ensure correct data transmission and protocol compatibility;

[0090] It should be noted that the Internet of Things, 5G networks, firewalls and data gateways jointly ensure that the system can efficiently, reliably and securely transmit and process data. The Internet of Things is responsible for data collection, the 5G network is responsible for high-speed transmission, the firewall provides security, and the data gateway ensures data compatibility and protocol conversion between systems, which can realize large-scale device access, data transmission and real-time processing.

[0091] In this embodiment, preferably, the cleaning is used to remove noise from the data, process missing data, remove duplicate data, process outliers, and clean up text data;

[0092] The noise removal is used to filter out the noise in the data. The noise removal assigns different weights to different data through weighted sliding average, so that more recent data is considered when smoothing the data;

[0093] The missing data processing is used to fill in the missing values ​​in the data, and the mean is used to fill in the missing data;

[0094] The deduplication is used to remove duplicate data in the data;

[0095] The outlier processing is used to detect outliers in the data, that is, to identify outliers through Z-Score, and to delete the detected outliers, and to fill the deleted outliers with linear interpolation;

[0096] The cleaning text data is used to process part of the text data, remove redundant spaces, punctuation marks and special characters for the fields containing text, unify the capitalization, or perform stem extraction and lemmatization;

[0097] It should be noted that cleaning can effectively improve the quality of data and ensure the accuracy and effectiveness of subsequent analysis or modeling;

[0098] Denoising is used to smooth data information and remove noise. The following calculation formula is used to process the data:

[0099] ,

[0100] in, Represented as the data corresponding to time point t in the data, Represented as a time point in the data The corresponding smoothing value is is a smoothing factor used to assign weights to new data and historical data, and Indicates that the data is at scale and location The transformation structure on is the conjugate of the wavelet transform, is the scale parameter, is the translation parameter;

[0101] The missing data are filled with the mean, and the calculation formula of the mean is as follows:

[0102] ;

[0103] in, is the mean of the calculated data, is the weight of each data, usually the latest data has a larger weight, For the data in time series, is the noise error, is the moving average coefficient of the weight, is the total amount of data;

[0104] Handling outliers For each time point, calculate the standardized deviation of the data from the mean and use the Z-Score formula:

[0105] ,

[0106] in, is the data at the current time point, is the mean value of the data, is the standard deviation of the data, represents the standardized deviation of the data from the mean;

[0107] And when Z exceeds the set threshold, it is marked as an outlier, that is, , and delete the outliers.

[0108] In this embodiment, preferably, the analysis and processing generates a data line graph from the data, and calculates the slope between two points in the data line graph to obtain slope data, and then generates a slope line graph from the slope data, and determines the direction of the data according to the inclination direction of the slope line graph, and determines the peak and valley points and high peak points of the data according to the inclination direction of the slope line graph, that is, if the slope line graph tilts upward and then tilts downward, the endpoint of the line of the slope line graph is the peak point, and if the slope line graph tilts downward and then tilts upward, the endpoint of the line of the slope line graph is the peak and valley point, and the peak point data and peak and valley point data are obtained respectively;

[0109] It should be noted that the combination of data line graph and slope line graph can accurately identify the peak points, valley points and direction of the data, which can help data analysts understand the volatility of the data more clearly and provide key information for subsequent predictions, decisions or further analysis.

[0110] In this embodiment, preferably, the feature extraction is used to extract feature information from the data, and the extracted features include mean, standard deviation, variance, maximum and minimum values, skewness and kurtosis;

[0111] Then, the feature data extracted from the historical load data, equipment data and economic activity data or substation power data and complaint call data are processed by combining features, and multiple features are combined into a new feature, that is, the combined feature is generated by feature crossover technology, and the peak point data and peak valley point data are integrated into the combined feature;

[0112] It should be noted that by extracting the key features of time series data, such as mean, standard deviation, variance, maximum and minimum values, skewness and kurtosis, and combining feature crossover technology to generate new combined features, as well as incorporating peak and valley data, the predictive ability of the data model can be significantly improved. The feature extraction and combined processing methods not only help to deeply understand the inherent laws of the data, but also provide richer information support for subsequent modeling and decision-making.

[0113] In this embodiment, preferably, the mean is calculated as follows:

[0114] ,

[0115] in, is the mean of the previous data, is the mean of the new data, is the amount of data;

[0116] And the calculation formula of the mean is as follows:

[0117] ,

[0118] in, is the calculated mean, is the weight value of the data, is the data value;

[0119] The standard deviation is calculated as follows:

[0120] ,

[0121] in, is the mean, is the data value, is the amount of data, is the calculated standard deviation;

[0122] The variance is calculated as follows:

[0123] ,

[0124] in, is the mean, is the data value, is the amount of data, is the calculated variance;

[0125] The maximum and minimum values ​​are calculated as follows:

[0126] ,

[0127] in, is the new maximum value after calculation, is the previous maximum value, is the new minimum value after calculation, is the previous minimum value, For newly added data;

[0128] The calculation formula of the skewness is as follows:

[0129] ,

[0130] in, It represents the skewness value obtained after calculation. is the amount of data, is the second center distance, which is the variance, is the third central moment, reflecting the skewness of the data distribution. is the new second central moment, is the second center distance before, is the new third central moment, is the previous third center distance;

[0131] The calculation formula of the kurtosis is as follows:

[0132] ,

[0133] in, It represents the kurtosis value obtained after calculation. It is the fourth central moment, which indicates the sharpness of the data distribution and the thickness of the tail. The fourth central moment calculation reflects the thickness of the tail of the data distribution and the sharpness of the distribution. The kurtosis of the standard normal distribution is 3. After subtracting 3, if the kurtosis is greater than 0, it means that the distribution is sharper or the tail is thicker than the normal distribution; if the kurtosis is less than 0, it means that the distribution is flatter than the normal distribution. is the new fourth center distance, is the fourth center distance before;

[0134] It should be noted that feature extraction is the process of machine learning converting data into features that can effectively represent the problem, which is used to extract the information that is most helpful for model learning. The quality of feature extraction directly affects the performance of the machine learning model. By extracting useful features, the complexity of the data can be reduced, the training speed of the model can be increased, and the accuracy of the prediction can be improved. Feature extraction can help remove redundant and irrelevant information, extract more discriminative features, and improve the accuracy of the model. Through feature selection or dimensionality reduction technology, the dimension of the input data can be reduced, thereby reducing computational overhead and improving training efficiency. The extracted features are usually simpler than the data, which can improve the interpretability of the model.

[0135] In this embodiment, preferably, the feature crossover technology first fuses the mean, standard deviation, variance, maximum and minimum values, skewness, kurtosis, peak point data and peak valley point data in the historical load data, equipment data, economic activity data, substation power data and complaint call data to form a historical load data feature set. , equipment dataset , Economic Activity Dataset , Substation electricity data set and complaint call dataset ;

[0136] The historical load data feature set used to predict power generation , device data , economic activity data Substation power data set fused through feature cross-talk technology and evaluated for power peak load regulation and complaint call dataset ;

[0137] ,

[0138] ,

[0139] The feature sets for power generation prediction are obtained respectively and power peak load evaluation

[0140] The calculation formula of multi-layer perceptron is:

[0141] ,

[0142] in, is the weight matrix of the fully connected layer, is the bias term, is the activation function, It is the intermediate representation of each layer, the cross feature Further learning through the fully connected layer, and finally output the prediction result ;

[0143] It should be noted that feature crossover is the fusion combination of the mean, standard deviation, variance, maximum and minimum values, skewness, kurtosis, peak point data and peak valley point data in historical load data, equipment data, economic activity data, substation power data and complaint call data as new features, and then the historical load data, equipment data, economic activity data, substation power data and complaint call data are fused to capture the interactions between them, which helps linear models or other traditional models to better model complex nonlinear relationships.

[0144] In this embodiment, preferably, the algorithm used by the prediction model is ridge regression, which is used to predict continuous numerical outputs and predict a continuous target variable based on input data;

[0145] ,

[0146] in, represents the loss function in ridge regression, It means the The actual observed value of samples, It means the The predicted value of samples, It means the The squared error of samples, It represents the sum of squared errors of the samples. It represents the regularization parameter, which controls the influence of the regularization term on the model. It represents the square of the L2 norm of the regression coefficient. is the number of input variables in the dataset, It represents the L2 regularization term, which penalizes the size of the regression coefficient to avoid overfitting. represents the regression coefficient vector;

[0147] It should be noted that ridge regression is used to minimize the sum of squared errors of the data. The introduction of the L2 regularization term avoids overfitting of the model when there are many features and large noise. The introduction of the regularization term enables the prediction model to balance the fitting accuracy of the data and the complexity of the regression coefficient, thereby improving the generalization ability of the prediction model and being able to effectively predict the peak power generation.

[0148] In this embodiment, preferably, the algorithm used by the evaluation model is random forest:

[0149] Set each tree to output a prediction value, and the final prediction result is the average of the prediction values ​​of all trees:

[0150] ,

[0151] in, is the prediction result of random forest, is the total number of trees, It is Tree pair samples The prediction is the prediction result of the power peak load output by the multi-layer perceptron. It represents the output of the multi-layer perceptron. The evaluation and prediction results of each;

[0152] Each tree will give a category label, and the final prediction result is the result of voting by all trees, and the classification result is obtained by majority voting:

[0153] ,

[0154] in, is the final category prediction, It is Tree pair samples The prediction mode is the mode operation, that is, taking the category with the most occurrences;

[0155] It should be noted that by generating multiple decision trees and combining their prediction results to make the final prediction, in the regression task, the final prediction of the random forest is the average of the prediction values ​​of all decision trees; in the classification task, the final prediction is the voting result of all trees, which can evaluate and process the peak-shaving power according to the substation power data and complaint call data, improve the correlation between prediction and evaluation, and facilitate the precise control of the predicted peak-shaving power.

[0156] The specific operation process of this application:

[0157] When in use, data is collected through the collection layer, and the collected data includes historical load data, equipment data, economic activity data and evaluation data for predicting and determining power generation, as well as substation power data, complaint call data and predicted power generation data for evaluating power peak regulation; and the data collected by the collection layer is transmitted through the transmission layer; and the edge cloud layer receives and processes the historical load data, equipment data, economic activity data, substation power data and complaint call data through the Internet of Things and 5G network. The edge cloud layer processes the data including preprocessing and analysis processing. The preprocessing is used to clean, convert and fuse the data; and the analysis processing is used to extract features and determine trends of the preprocessed data;

[0158] The analysis and processing generates a data line graph from the data, and calculates the slope between two points in the data line graph to obtain slope data, and then generates a slope line graph from the slope data, and determines the direction of the data according to the inclination direction of the slope line graph, and determines the peak and valley points and high peak points of the data according to the inclination direction of the slope line graph, that is, if the slope line graph tilts upward and then tilts downward, the end point of the line of the slope line graph is the peak point, and if the slope line graph tilts downward and then tilts upward, the end point of the line of the slope line graph is the peak and valley point, and the peak point data and peak and valley point data are obtained respectively; it can help data analysts understand the fluctuation characteristics of the data more clearly, and provide key information for subsequent predictions, decisions or further analysis;

[0159] Feature extraction extracts characteristic information from the data, that is, feature extraction of time series data. The extracted features include mean, standard deviation, variance, maximum and minimum values, skewness and kurtosis;

[0160] Then, the feature data extracted from the historical load data, equipment data and economic activity data or substation power data and complaint call data are processed by combining features, and multiple features are combined into a new feature, that is, the combined feature is generated by feature crossover technology, and the peak point data and peak valley point data are integrated into the combined feature;

[0161] The feature crossover technology first integrates the mean, standard deviation, variance, maximum and minimum values, skewness, kurtosis, peak point data and peak and valley point data in historical load data, equipment data, economic activity data, substation power data and complaint call data to form a historical load data feature set. , equipment dataset , Economic Activity Dataset , Substation electricity data set and complaint call dataset ;

[0162] The historical load data feature set used to predict power generation , device data , economic activity data Substation power data set fused through feature cross-talk technology and evaluated for power peak load regulation and complaint call dataset ;

[0163] ,

[0164] ,

[0165] The feature sets for power generation prediction are obtained respectively and power peak load evaluation

[0166] The calculation formula of multi-layer perceptron is:

[0167] ,

[0168] in, is the weight matrix of the fully connected layer, is the bias term, is the activation function, It is the intermediate representation of each layer, the cross feature Further learning through the fully connected layer, and finally output the prediction result ;

[0169] The edge cloud layer transmits the processed feature extraction and trend determination to the management platform, and the management platform includes a prediction model, an evaluation model and a visualization module; the prediction model predicts the peak-shaving power generation by analyzing the processed historical load data, equipment data and economic activity data, and uses the evaluation data as update data and inputs the evaluation data into the prediction model to achieve timely adjustment and control of the predicted power generation; the evaluation model evaluates the peak-shaving process by analyzing the processed substation power data and complaint call data, and uses the predicted power generation data as update data and inputs the predicted power generation data into the evaluation model to achieve comparison and determination of the peak-shaving demand and predicted power generation data;

[0170] The algorithm used in the prediction model is ridge regression, which is used to predict continuous numerical outputs and predict a continuous target variable based on the input data;

[0171] ,

[0172] This improves the generalization ability of the prediction model and can effectively predict the power generation for peak load regulation;

[0173] The algorithm used to evaluate the model is random forest:

[0174] Set each tree to output a prediction value, and the final prediction result is the average of the prediction values ​​of all trees:

[0175] ,

[0176] Each tree will give a category label, and the final prediction result is the result of voting by all trees, and the classification result is obtained by majority voting:

[0177] ;

[0178] By generating multiple decision trees and combining their prediction results to make the final prediction, in the regression task, the final prediction of the random forest is the average of the prediction values ​​of all decision trees; in the classification task, the final prediction is the voting result of all trees, which can evaluate and process the peak load according to the substation power data and complaint call data, improve the correlation between prediction and evaluation, and facilitate the accurate control of the predicted power of peak load;

[0179] The visualization module uses digital twin technology to establish a visualization interface for power generation equipment, which facilitates the integration of actual power generation data, predicted power generation data and evaluation data produced by the power generation equipment on the digital twin model for display, so that the peak power information can be visualized and facilitate background analysis and processing.

[0180] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A power system peak load demand assessment system, characterized in that: It includes collection layer, transmission layer, edge cloud layer and management platform; The collection layer is used to collect data, and the data collected by the collection layer includes historical load data, equipment data, economic activity data and evaluation data for predicting and determining power generation, as well as substation power data, complaint call data and predicted power generation data for evaluating power peak load regulation; The transmission layer is used to transmit the data collected by the collection layer. The historical load data is transmitted through the 5G network, the equipment data is transmitted through the Internet of Things, the economic activity data is transmitted through the 5G network, and the substation power data and the complaint call data are transmitted through the 5G network; The edge cloud layer is used to receive and process data. The edge cloud layer receives the historical load data, the equipment data, the economic activity data, the substation power data and the complaint call data through the Internet of Things and the 5G network. The processing of data by the edge cloud layer includes preprocessing and analysis processing. The preprocessing is used to clean, convert and fuse the data, and the analysis processing is used to extract features and determine trends of the preprocessed data; The analysis and processing generates a data line graph from the data, and calculates the slope between two points in the data line graph to obtain slope data, and then generates a slope line graph from the slope data, and determines the direction of the data according to the inclination direction of the slope line graph, and determines the peak and valley points and high peak points of the data according to the inclination direction of the slope line graph, that is, if the slope line graph tilts upward and then tilts downward, the end point of the line of the slope line graph is the high peak point, and if the slope line graph tilts downward and then tilts upward, the end point of the line of the slope line graph is the peak and valley point, and high peak point data and peak and valley point data are obtained respectively; The feature extraction is used to extract feature information from the data, and the extracted features include mean, standard deviation, variance, maximum and minimum values, skewness and kurtosis; Then, the feature data extracted from the historical load data, equipment data and economic activity data or substation power data and complaint call data are processed by combining features, and multiple features are combined into a new feature, that is, the combined feature is generated by feature crossover technology, and the peak point data and peak valley point data are integrated into the combined feature; The mean is calculated as follows: ， in, is the mean of the previous data, is the mean of the new data, is the amount of data; And the calculation formula of the mean is as follows: ， in, is the calculated mean, is the weight value of the data, is the data value; The standard deviation is calculated as follows: ， in, is the mean, is the data value, is the amount of data, is the calculated standard deviation; The variance is calculated as follows: ， in, is the mean, is the data value, is the amount of data, is the calculated variance; The maximum and minimum values ​​are calculated as follows: ， in, is the new maximum value after calculation, is the previous maximum value, is the new minimum value after calculation, is the previous minimum value, For newly added data; The calculation formula of the skewness is as follows: ， in, It represents the skewness value obtained after calculation. is the amount of data, is the second center distance, which is the variance, is the third central moment, reflecting the skewness of the data distribution. is the new second central moment, is the second center distance before, is the new third central moment, is the previous third center distance; The calculation formula of the kurtosis is as follows: ， in, It represents the kurtosis value obtained after calculation. It is the fourth central moment, which indicates the sharpness of the data distribution and the thickness of the tail. The fourth central moment calculation reflects the thickness of the tail of the data distribution and the sharpness of the distribution. The kurtosis of the standard normal distribution is 3. After subtracting 3, if the kurtosis is greater than 0, it means that the distribution is sharper or the tail is thicker than the normal distribution; if the kurtosis is less than 0, it means that the distribution is flatter than the normal distribution. is the new fourth center distance, is the fourth center distance before; The feature crossover technology first integrates the mean, standard deviation, variance, maximum and minimum values, skewness, kurtosis, peak point data and peak valley point data in historical load data, equipment data, economic activity data, substation power data and complaint call data to form a historical load data feature set. , equipment dataset , Economic Activity Dataset , Substation electricity data set and complaint call dataset ; The historical load data feature set used to predict power generation , device data , economic activity data Substation power data set fused through feature cross-talk technology and evaluated for power peak load regulation and complaint call dataset ; ， ， The feature sets for power generation prediction are obtained respectively and power peak load evaluation The calculation formula of the multi-layer perceptron is: ， in, is the weight matrix of the fully connected layer, is the bias term, is the activation function, It is the intermediate representation of each layer, the cross feature Further learning through the fully connected layer, and finally output the prediction result ; The management platform includes a prediction model, an evaluation model and a visualization module. The prediction model predicts the peak-shaving power generation by analyzing and processing the historical load data, the equipment data and the economic activity data, and uses the evaluation data calculated by the evaluation model as update data, and inputs the evaluation data into the prediction model to achieve timely adjustment and control of the predicted power generation. The evaluation model evaluates the peak-shaving process by analyzing and processing the substation power data and the complaint telephone data, and uses the predicted power generation data calculated by the prediction model as update data, and inputs the predicted power generation data into the evaluation model. The prediction power generation data provides data support for the evaluation model. The visualization module establishes a visualization interface for power generation equipment through digital twin technology, which is convenient for integrating and displaying the actual power generation data produced by the power generation equipment, the predicted power generation data and the evaluation data.

2. A power system peak load demand assessment system according to claim 1, characterized in that: The historical load data is extracted from the data repository, the equipment data is collected from various types of power generation equipment through sensors, the economic activity data is used to collect data on activities at various commercial or production sites, the substation power data is used to collect data on the power supply voltage of each substation, mainly used to collect load voltage for peak load demand, and the complaint call data is used to count the number of complaint calls; The predicted power generation data is data obtained by calculation and analysis of the prediction model, and the evaluation data is data obtained by calculation and analysis of the evaluation model, and the predicted power generation data and the evaluation data are respectively used as update data of the evaluation model and update data of the prediction model.

3. A power system peak load demand assessment system according to claim 1, characterized in that: The transport layer includes the Internet of Things and the 5G network. The Internet of Things is used to upload data collected by various sensors. The 5G network is used to transmit system data to the edge cloud layer. The transport layer also includes a firewall and a data gateway. The firewall is mainly used to protect the network from attacks and unauthorized access, and to implement traffic monitoring and security policies; The data gateway is used to exchange, convert and manage data between different systems or applications to ensure correct data transmission and protocol compatibility.

4. A power system peak load demand assessment system according to claim 1, characterized in that: The cleaning is used to remove noise from the data, process missing data, remove duplicate data, process outliers and clean up text data; The noise removal is used to filter out the noise in the data. The noise removal assigns different weights to different data through weighted sliding average, so that more recent data is considered when smoothing the data; The missing data processing is used to fill in the missing values ​​in the data, and the mean is used to fill in the missing data; The deduplication is used to remove duplicate data in the data; The outlier processing is used to detect outliers in the data, that is, to identify outliers through Z-Score, and to delete the detected outliers, and to fill the deleted outliers with linear interpolation; The cleaned text data is used to process part of the text data, remove redundant spaces, punctuation marks and special characters for fields containing text, unify the case, or perform stem extraction and word form restoration.

5. The power system peak load demand assessment system according to claim 1, characterized in that: The algorithm used in the prediction model is ridge regression, which is used to predict continuous numerical outputs and predict a continuous target variable based on input data; ， in, represents the loss function in ridge regression, It means the The actual observed value of samples, It means the The predicted value of samples, It means the The squared error of samples, It represents the sum of squared errors of the samples. It represents the regularization parameter, which controls the influence of the regularization term on the model. It represents the square of the L2 norm of the regression coefficient. is the number of input variables in the dataset, It represents the L2 regularization term, which penalizes the size of the regression coefficient to avoid overfitting. represents the regression coefficient vector.

6. A power system peak load demand assessment system according to claim 5, characterized in that: The algorithm used in the evaluation model is random forest: Set each tree to output a prediction value, and the final prediction result is the average of the prediction values ​​of all trees: ， in, is the prediction result of random forest, is the total number of trees, It is Tree pair samples The prediction is the prediction result of the power peak load output by the multi-layer perceptron. It represents the output of the multi-layer perceptron. The evaluation and prediction results of each; Each tree will give a category label, and the final prediction result is the result of all trees voting, and the classification result is obtained by majority voting: ， in, is the final category prediction, It is Tree pair samples The prediction mode is the mode operation, that is, taking the category with the most occurrences.

Images