Virtual power plant multi-dimensional adjustable potential assessment method of comprehensive intelligent model

Abstract

The invention provides a virtual power plant multi-dimensional adjustable potential evaluation method based on a comprehensive intelligent model, and relates to the technical field of power generation potential evaluation, and the method comprises the steps: obtaining a first power supply node of a virtual power plant; performing power generation influence factor analysis according to the power generation type identifier to obtain an associated influence factor; based on the unit bit number identifier and the associated influence factor, collecting historical operation cross section data to train a plurality of generating capacity prediction models; performing preset time zone power generation potential evaluation on the first power supply node to obtain a predicted power supply amount of the first power supply node; and performing power generation potential visual identification on the first power supply node in the virtual power plant in the preset time zone. According to the method and the device, the technical problem that the refinement degree is insufficient due to the fact that power generation capacity evaluation lacks analysis of the power generation units with bit numbers as units in the prior art can be solved, the bit number of each power generation unit is analyzed, the exclusive power generation capacity prediction model is trained for potential evaluation, and the refinement degree of evaluation is improved.

Description

Technical Field

[0001] The present application relates to the technical field of power generation potential assessment, and in particular to a multi-dimensional adjustable potential assessment method for a virtual power plant of a comprehensive intelligent model. Background Art

[0002] Virtual power plants use advanced information and communication technologies and intelligent control technologies to aggregate distributed power sources (such as renewable energy generation equipment such as solar energy and wind energy), energy storage systems, and controllable loads to form a virtual power plant that can be uniformly coordinated and controlled and optimized. Virtual power plants contain various types of resources with different characteristics and adjustable capabilities. In order to achieve optimal resource allocation and improve overall energy efficiency and economic benefits, it is necessary to evaluate its multi-dimensional adjustable potential. By accurately evaluating the multi-dimensional adjustable potential of virtual power plants, it can be ensured that when the power system faces sudden failures or extreme situations (such as failure of some power generation equipment, grid transmission congestion, etc.), it can be reasonably planned and dispatched to avoid grid operation problems caused by the disorderly access of distributed energy, thereby promoting large-scale and orderly access to distributed energy. However, the existing power generation capacity evaluation methods usually focus on the overall performance of the entire power plant or large power generation equipment, and do not go deep into the analysis of power generation units by bit number, resulting in insufficient evaluation of power generation capacity, which affects the overall power generation efficiency and reliability.

[0003] In summary, the prior art has a technical problem of insufficient refinement due to the lack of analysis of power generation units in units of bit numbers in power generation capacity assessment. Summary of the invention

[0004] The purpose of this application is to provide a multi-dimensional adjustable potential assessment method for a virtual power plant based on a comprehensive intelligent model, so as to solve the technical problem in the prior art that the power generation capacity assessment lacks analysis of the power generation units in units of bit numbers, resulting in insufficient refinement.

[0005] In view of the above problems, the present application provides a multi-dimensional adjustable potential assessment method for a virtual power plant of a comprehensive intelligent model, wherein the multi-dimensional adjustable potential assessment method for a virtual power plant of a comprehensive intelligent model includes: obtaining a first power supply node of a virtual power plant, wherein the first power supply node has a plurality of power generation units, and any one of the power generation units has a power generation type identifier and a unit position number identifier; performing power generation influencing factor analysis according to the power generation type identifier to obtain associated influencing factors; based on the unit position number identifier and the associated influencing factors, collecting historical operating cross-sectional data to train a plurality of power generation prediction models, wherein the plurality of power generation prediction models correspond one-to-one to the plurality of power generation units; performing a preset time zone power generation potential assessment on the first power supply node according to the plurality of power generation prediction models to obtain a predicted power supply of the first power supply node; performing a preset time zone power generation potential visualization identification of the first power supply node in the virtual power plant according to the predicted power supply of the first power supply node.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: By obtaining the first power supply node of the virtual power plant, wherein the first power supply node has several power generation units, and any power generation unit has a power generation type identifier and a unit position number identifier; performing power generation influencing factor analysis according to the power generation type identifier to obtain associated influencing factors; based on the unit position number identifier and the associated influencing factors, collecting historical operation cross-sectional data to train several power generation prediction models, wherein the several power generation prediction models correspond to the several power generation units one by one; according to the several power generation prediction models, the power generation potential of the first power supply node is evaluated in a preset time zone to obtain the predicted power supply of the first power supply node; according to the predicted power supply of the first power supply node, the power generation potential of the first power supply node in the virtual power plant is visually identified in the preset time zone. In other words, by identifying the key influencing factors of different types of power generation units according to the power generation type identifier; combining the position number identifier and influencing factor of each power generation unit, collecting the historical operation data of the equipment, modeling each power generation unit separately, and training multiple more targeted prediction models; according to the power supply capacity of the node predicted in the preset time zone, accurate management of each power generation unit is achieved, and the refinement of power generation potential evaluation is improved.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0009] Figure 1 Schematic diagram of the process flow of the multi-dimensional adjustable potential assessment method of the virtual power plant based on the comprehensive intelligent model of this application.

[0010] Figure 2 A schematic diagram of the process of obtaining associated influencing factors in the multi-dimensional adjustable potential assessment method of a virtual power plant of the comprehensive intelligent model of this application. DETAILED DESCRIPTION

[0011] This application solves the technical problem in the prior art that the power generation capacity assessment lacks the analysis of the power generation unit in units of bit numbers, resulting in insufficient refinement, by providing a multi-dimensional adjustable potential assessment method for virtual power plants with a comprehensive intelligent model. The key influencing factors of different types of power generation units are identified based on the power generation type identification; the historical operation data of the equipment is collected in combination with the bit number identification and influencing factors of each power generation unit, and each power generation unit is modeled separately to train multiple more targeted prediction models; the power supply capacity of the node is predicted according to the preset time zone, which realizes the precise management of each power generation unit and improves the refinement of the power generation potential assessment.

[0012] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0013] For examples, please see the attached Figure 1 The present application provides a method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model, wherein the method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model specifically comprises the following steps: Step 1: Obtain a first power supply node of a virtual power plant, wherein the first power supply node has a plurality of power generation units, and any power generation unit has a power generation type identifier and a unit position number identifier.

[0014] Specifically, a virtual power plant refers to a system that integrates decentralized power generation resources (such as solar photovoltaic, wind power, energy storage systems, etc.) and adjustable loads through advanced communication technologies and software platforms to achieve unified scheduling and optimized management. It provides power services to the outside world like a traditional power plant, but there is actually no physical power plant entity. In a virtual power plant, the first power supply node is identified and determined, which is one of the key power supply points in the virtual power plant. It usually contains multiple power generation units, such as wind turbines, solar panels, energy storage systems and other single power generation equipment in the virtual power plant. Each power generation unit is given two important identifiers: the power generation type identifier and the unit position identifier. The power generation type identifier is used to distinguish different types of power generation units, such as photovoltaic, wind power, energy storage, etc. The unit position identifier is a number that uniquely identifies each power generation unit, which is used to distinguish each power generation unit in the power supply node in the virtual power plant, similar to the ID card of each unit.

[0015] Step 2: Analyze the power generation influencing factors according to the power generation type identification to obtain the associated influencing factors.

[0016] Specifically, the power generation type identification is sent to the user end and feedback information is received, that is, the factors that may affect the performance of the power generation unit proposed by the user based on actual experience and needs, including the initial influencing factors, are obtained. And the influencing factors are analyzed for correlation according to the power generation type identification, and the correlation between each influencing factor and the power generation is determined, that is, the factors that mainly affect the power generation are screened out, and the influencing factors with high correlation are used as the associated influencing factors. The initial influencing factors with a correlation greater than or equal to the correlation threshold are screened out from the initial influencing factors, and they are used as the associated influencing factors. The power generation influencing factor analysis refers to identifying and screening the main influencing factors related to the type of power generation according to the power generation type identification, which are called associated influencing factors. A variety of statistical and analytical methods (such as correlation analysis or gray correlation analysis) are used to evaluate the relationship between each influencing factor and the power generation. Exclude or reduce the factors that have little impact on the power generation, and retain those factors with high correlation as the associated influencing factors for subsequent model training. By focusing on the main influencing factors, the input variables predicted by the model are more accurate and the noise caused by irrelevant variables is reduced.

[0017] Step three: Based on the unit position identification and the associated influencing factors, historical operation cross-sectional data are collected to train a number of power generation prediction models, wherein the several power generation prediction models correspond one-to-one to the several power generation units.

[0018] Specifically, the associated influencing factors are used as mapping independent variables, the power generation is used as mapping dependent variables, and the first historical operation cross-sectional data of the same type and scale as the power generation unit is collected. The feedforward neural network is trained multiple times, and the deviation between the predicted value and the true value is calculated for each training, and finally a prediction model with a smaller deviation is obtained as the initial power generation prediction model. The associated influencing factors are used as mapping independent variables again, and the power generation is used as the mapping dependent variable. The second historical operation cross-sectional data with the unit position number identification is collected, and the initial power generation prediction model is retrained to obtain a dedicated power generation prediction model. Each model will be personalized according to its corresponding unit position number identification and associated influencing factor data. Finally, these prediction models establish a one-to-one correspondence with their respective power generation units to ensure that each power generation unit can obtain an exclusive and optimized power generation prediction model. By training an independent model for each power generation unit, the power generation characteristics of different units under similar conditions can be captured more accurately, and the historical data and related influencing factors of a specific power generation unit can be combined to improve the accuracy and reliability of the prediction model.

[0019] Step 4: Based on the several power generation prediction models, the power generation potential of the first power supply node in the preset time zone is evaluated to obtain the predicted power supply of the first power supply node.

[0020] Specifically, several power generation prediction models are used to collect relevant data that affect the power generation of the first power supply node, including environmental conditions, equipment operating efficiency, etc. A time zone is pre-set, such as one day, one month, etc. Using the previously trained power generation prediction model, the collected associated influencing factors are used as input, and each power generation unit is predicted separately using the power generation prediction model to obtain the expected power generation of each unit in the preset time zone. The prediction results of all power generation units are summarized to calculate the overall predicted power supply of the first power supply node, and the final result is the predicted power supply of the first power supply node in the preset time zone. By accurately evaluating the power generation potential of the first power supply node, mastering the prediction ability of each power supply node, configuring and scheduling resources, and improving the overall operating efficiency of the virtual power plant. Through the prediction of power supply, it is possible to identify the possible risk of insufficient power supply in advance, and take measures to adjust and respond.

[0021] Step 5: Visually mark the power generation potential of the first power supply node in a preset time zone in the virtual power plant based on the predicted power supply of the first power supply node.

[0022] Specifically, the predicted power supply of the first power supply node, including the predicted power generation value and the preset time zone, is displayed in a visual form so that relevant personnel (such as power dispatchers and managers) can intuitively understand the power generation potential and trend of the node. Select appropriate visualization tools and methods, such as line charts, bar charts or area charts, and data visualization tools such as Matplotlib, Tableau, Power BI, etc. Input the predicted power supply of the first power supply node and the corresponding time zone data into the visualization tool, and draw the corresponding graph according to the predicted power generation, using time as the x-axis and the predicted power supply as the y-axis, to ensure that the graph accurately reflects the prediction results. Display the generated visualization results to relevant decision makers or team members, such as through a real-time monitoring system. Real-time visualization of power generation potential helps to quickly understand complex data, making the prediction results more intuitive and convenient for decision makers to evaluate.

[0023] Further, as attached Figure 2 As shown, step 2 of this application includes: The power generation type identifier is sent to the user end, and user feedback information is received, wherein the user feedback information includes an initial influencing factor; a correlation analysis is performed on the initial influencing factor according to the power generation type identifier to obtain an initial influencing factor correlation degree; and the initial influencing factor whose initial influencing factor correlation degree is greater than or equal to a correlation degree threshold is extracted and set as the correlation influencing factor.

[0024] Specifically, the power generation type identification is sent to the user end, that is, the user interface of the virtual power plant, such as a computer, mobile device or any other device that can interact with the virtual power plant, so that the user can understand the type of power generation unit currently being analyzed (such as wind power, photovoltaic, etc.). Users (such as power plant operators, technical experts or other stakeholders related to power generation) provide feedback information based on the power generation type identification, which is usually based on the user's own experience and needs. The characteristic data that is believed to have an impact on the performance of the power generation unit, such as temperature, wind speed, sunshine intensity, etc. are proposed. For example, for solar power generation, the user may provide initial influencing factors such as solar radiation intensity, cleanliness of photovoltaic panels, and surrounding obstructions; for wind power generation, initial influencing factors such as wind speed, wind direction stability, and degree of wear of wind turbine blades may be provided.

[0025] The correlation analysis of the initial influencing factors fed back by users is carried out to evaluate the correlation or dependence between different influencing factors and power generation, and the correlation degree, that is, the strength of the relationship between the two variables, is obtained. The correlation or degree of influence between the initial influencing factors and the performance of a specific type of power generation unit is determined by statistical analysis or machine learning methods (such as correlation analysis, principal component analysis, etc.). The higher the correlation degree, the greater the influence of the factor on the performance of the power generation unit. The correlation degree of each factor and the power generation is added to the initial influencing factor correlation degree. A correlation threshold is set in advance to determine whether the correlation degree is high enough to determine whether a factor is a correlation influencing factor. The initial influencing factors greater than or equal to the correlation threshold in the initial influencing factor correlation degree are extracted, that is, the influencing factors highly correlated with the power generation, and these factors are used as correlation influencing factors. The initial influencing factors highly correlated with the power generation type are extracted from the user's feedback and used as correlation influencing factors, accurately identifying the factors that have a significant impact on the performance of the power generation unit, removing irrelevant or low-correlation factors, reducing the computational complexity of the model, and optimizing the utilization of system resources.

[0026] Furthermore, the present application also includes the following steps: With the power generation type identifier as a constraint, a power generation record data set is collected, wherein the power generation record data set includes an initial influencing factor record value set and a power generation record value set; the power generation record value set is normalized to generate a first data sequence; the first attribute initial influencing factor record value set of the initial influencing factor record value set is normalized to obtain a second data sequence; until the Nth attribute initial influencing factor record value set of the initial influencing factor record value set is normalized to obtain an Nth data sequence; with the first data sequence as a reference sequence and the second data sequence to the Nth data sequence as a comparison sequence, a grey correlation analysis is performed to obtain the first attribute initial influencing factor correlation degree to the Nth attribute initial influencing factor correlation degree, and add it into the initial influencing factor correlation degree.

[0027] Specifically, the power generation type identification is used as a constraint, and the historical power generation record data related to these power generation types are extracted from the database of the virtual power plant, including the record value set of the initial influencing factors (such as weather, temperature, wind speed, etc.) and the corresponding actual power generation record value set. The power generation record value set is normalized to eliminate the dimension difference and data range inconsistency problems, so that the data can be compared at the same scale. The normalized data forms the first data sequence, and each data point in the data sequence is in a specific interval (such as [0,1] or other set intervals), which serves as the benchmark sequence in the subsequent grey correlation analysis. The normalization process is to convert data of different magnitudes into a unified and comparable range. The data in the power generation record value set may have large numerical differences due to factors such as the scale and operating time of the power generation unit.

[0028] Normalize each attribute initial influencing factor record value set (such as temperature, wind speed, sunshine, etc.) in the initial influencing factor record set, and generate an independent data sequence for each attribute initial influencing factor record value set, that is, the second data sequence to the Nth data sequence, to ensure that all factors that may affect the power generation have undergone data preprocessing. The normalized first data sequence is used as the benchmark sequence, and the second data sequence to the Nth data sequence are used as the comparison sequence, and the grey correlation degree of each comparison sequence relative to the benchmark sequence is calculated in turn.

[0029] Grey correlation analysis measures the similarity between each comparison sequence (influencing factor) and the benchmark sequence (power generation), that is, calculates the correlation coefficient between the benchmark sequence and each comparison sequence at each moment, takes the average of the correlation coefficient of each comparison sequence, and obtains the grey correlation of the sequence relative to the benchmark sequence, thereby measuring the degree of correlation of each initial influencing factor with power generation. The value range of grey correlation is usually between 0 and 1. The closer it is to 1, the closer the relationship between the comparison sequence and the benchmark sequence, that is, the closer the connection between the influencing factor and power generation. According to the results of grey correlation analysis, the correlation of each initial influencing factor is added to the initial influencing factor correlation. Through normalization and grey correlation analysis, the influencing factor with the closest relationship to power generation is identified.

[0030] Further, step three of this application includes: Taking the associated influencing factor as the mapping independent variable and the power generation as the mapping dependent variable, collect the first historical operation section data of the same type and scale as the power generation unit; train the feedforward neural network according to the first historical operation section data to obtain an initial power generation prediction model; taking the associated influencing factor as the mapping independent variable and the power generation as the mapping dependent variable, collect the second historical operation section data with the unit position number identification; train the initial power generation prediction model according to the second historical operation section data to generate a power generation prediction model, wherein the power generation prediction model belongs to the several power generation prediction models.

[0031] Specifically, the associated influencing factor is used as the mapping independent variable, the power generation is used as the mapping dependent variable, and the corresponding first associated influencing factor record value and the first power generation record value are collected. The first associated influencing factor record value is grouped using a clustering algorithm, and similar record values ​​are divided together to obtain multiple groups of first associated influencing factor record values. The first power generation record values ​​are grouped according to multiple groups of first associated influencing factor record values ​​to obtain multiple groups of first power generation record values. Each group of power generation record values ​​is traversed, abnormal data is eliminated, and then each group of data is fitted to obtain multiple first power generation fitting values. Each group of associated influencing factor record values ​​is traversed to randomly select values ​​to obtain the first associated influencing factor sorting record value. Representative samples are obtained by random sampling, which is convenient for covering different ranges of influencing factor characteristics and improving the accuracy of the model's prediction of power generation under different conditions. The first power generation fitting value of each group and the corresponding first associated influencing factor sorting record value are combined and added to the first historical operation section data. The first historical operation section data refers to historical data points with the same type and scale as the current power generation unit.

[0032] According to the first historical operation section data, the feedforward neural network is trained. Each time the training is performed, the deviation between the predicted value and the true value is calculated, and the prediction accuracy is determined according to the ratio of the deviation being less than or equal to the preset power generation deviation threshold. When the prediction accuracy is greater than or equal to the accuracy threshold, the initial power generation prediction model is obtained. The data collection process of the second historical operation section data is the same as the first historical operation section data collection process. The second historical operation section data with a specific unit position number identification is collected, including the associated influencing factor sorting record value and the power generation fitting value with a specific unit position number identification. The initial power generation prediction model is trained according to the second historical operation section data to obtain a prediction model that is more suitable for the specific unit. After two stages of training, the generated power generation prediction model not only adapts to the characteristics of the specific power generation unit, but also accurately reflects the role of the associated influencing factors in the power generation prediction. In the end, each specific power generation unit corresponds to an exclusive prediction model, which is used as part of multiple power generation prediction models in the virtual power plant for multi-dimensional regulation and power generation potential assessment.

[0033] Furthermore, the present application also includes the following steps: Taking the associated influencing factor as the mapping independent variable and the power generation as the mapping dependent variable, collecting the first associated influencing factor record value and the first power generation record value; performing cluster analysis on the first associated influencing factor record value to obtain multiple groups of first associated influencing factor record values; grouping the first power generation record value according to the multiple groups of first associated influencing factor record values ​​to obtain multiple groups of first power generation record values; traversing the multiple groups of first power generation record values ​​to perform power generation outlier fitting to obtain multiple first power generation fitting values; traversing the multiple groups of first associated influencing factor record values ​​to perform random value sampling to obtain multiple first associated influencing factor sorting record values; adding the multiple first power generation fitting values ​​and the multiple first associated influencing factor sorting record values ​​into the first historical operation section data.

[0034] Furthermore, the present application also includes the following steps: The data collection process of the second historical operation cross-sectional data is the same as the data collection process of the first historical operation cross-sectional data.

[0035] Specifically, the associated influencing factors are used as mapping independent variables, and the power generation is used as the mapping dependent variable. The historical records of the associated influencing factors related to the power generation forecast and the corresponding actual power generation record values ​​are extracted from the database of the virtual power plant to obtain the first associated influencing factor record values ​​and the first power generation record values. Cluster analysis is performed on the first associated influencing factor record values, such as K-means or hierarchical clustering, and similar associated influencing factors are grouped into the same group to obtain multiple groups of first associated influencing factor record values, each group including associated influencing factors with high similarity. Cluster analysis is an unsupervised learning method that groups objects in a data set into multiple classes consisting of similar objects, called clusters.

[0036] A threshold is set in advance to measure the degree of difference between the associated influencing factor record values. Two record values ​​are randomly selected from the first associated influencing factor record value, and the deviation between the two record values ​​is calculated. The proportion of attributes whose deviation is greater than the associated influencing factor deviation threshold, that is, the proportion of associated factors, is statistically calculated and used as the first deviation distance. The mean of all attribute deviations in the first and second associated influencing factor record values ​​is calculated and used as the second deviation distance. Two deviation distance thresholds are set respectively. Only when the first deviation distance is less than or equal to the corresponding deviation distance threshold, and the second deviation distance is also less than or equal to the corresponding deviation distance threshold, the two associated influencing factor record values ​​are divided into the same group. And so on, until the comparison of all associated influencing factor record values ​​in the first associated influencing factor record value is completed, and multiple groups of first associated influencing factor record values ​​are obtained.

[0037] The first power generation record values ​​are grouped according to multiple groups of first associated influencing factor record values, that is, the first power generation record values ​​are classified according to the grouping results of the associated influencing factors related thereto, and the power generation record values ​​in each group will correspond to the corresponding associated influencing factor record values, so that the power generation can be analyzed and predicted more accurately. Each group of first power generation record values ​​contains power generation records under the same conditions (for example, the same environmental influencing factor conditions). Each group of power generation record values ​​is traversed, and power generation outlier processing is performed to obtain multiple first power generation fitting values ​​without abnormal interference. For example, using the 3σ principle, the mean a and standard deviation b of each group of data are calculated, and the range is determined such as (a-3b, a+3b), and the data points falling outside the range are regarded as abnormal data points and eliminated. Fit the processed data, such as linear regression method or high-order polynomial fitting.

[0038] Random sampling is performed on multiple groups of first associated influencing factor record values ​​to obtain multiple associated influencing factor sorting record values, that is, random sample data representing each group of characteristics, covering each group of influencing factor characteristics. The power generation record values ​​obtained by removing outliers and the associated influencing factor sorting record values ​​obtained by random sampling are combined to form the final first historical operation section data. The collection process of the second historical operation section data is the same as that of the first historical operation section data, that is, the same as above, and will not be repeated here. The accuracy of the data is improved by clustering and fitting to remove outliers, making the model more adaptable. The data generated by random sampling and cluster analysis covers the changes in power generation under different conditions, and improves the model's ability to respond to multiple input features.

[0039] Furthermore, the present application also includes the following steps: Configure an associated impact factor deviation threshold for an associated impact factor record value; extract the first associated impact factor record value and the second associated impact factor record value of the first associated impact factor record value; count the proportion of attributes whose associated impact factor record value deviation between the first associated impact factor record value and the second associated impact factor record value is greater than or equal to the associated impact factor deviation threshold, and set it as a first deviation distance; count the average of the associated impact factor record value deviations between the first associated impact factor record value and the second associated impact factor record value, and set it as a second deviation distance; when the first deviation distance is less than or equal to the first deviation distance threshold, and the second deviation distance is less than or equal to the second deviation threshold, add the first associated impact factor record value and the second associated impact factor record value as the same group, otherwise, set them as different groups.

[0040] Specifically, a threshold, namely the correlation influence factor deviation threshold, is set to measure the degree of difference between the correlation influence factor record values. It is used to determine whether the difference between two correlation influence factor record values ​​is large enough to be regarded as different groups. Two record values ​​are randomly selected from the first correlation influence factor record value, and are recorded as the first correlation influence factor record value and the second correlation influence factor record value respectively. The deviation between the first correlation influence factor record value and the second correlation influence factor record value is calculated, and the proportion of attributes whose deviation is greater than or equal to the correlation influence factor deviation threshold is counted, which is recorded as the first deviation distance. The attribute proportion refers to the ratio between the number of attributes whose deviation between the first correlation influence factor record value and the second correlation influence factor record value is greater than or equal to the correlation influence factor deviation threshold and the total number of all attributes. For example, assuming that there are four correlation influence factors in each of the two record values, the calculated deviation values ​​are: 0.5, 1.5, 0.3, and 0.7, respectively, and the correlation influence factor deviation threshold is set to 1.0. Only the second deviation value is greater than the correlation influence factor deviation threshold, then the first deviation distance is 1 / 4=0.25.

[0041] Calculate the average deviation of the correlation factor record value between the first correlation impact factor record value and the second correlation impact factor record value, and use it as the second deviation distance. For example, assuming that the deviation values ​​of the four correlation impact factors are: 0.5, 1.5, 0.3, 0.7, the second deviation distance is (0.5+1.5+0.3+0.7) / 4=0.75. If the first deviation distance is less than or equal to the preset first deviation distance threshold, and the second deviation distance is less than or equal to the preset second deviation threshold, it is determined that the deviation of the two record values ​​is within an acceptable range, and the two correlation impact factor record values ​​are added to the same group. The first deviation distance threshold and the second deviation threshold are two preset parameters used to determine whether the difference between the two correlation impact factor record values ​​is small enough to consider that these record values ​​belong to the same group. The first deviation distance threshold is used to control the allowed proportion of large deviation attributes to ensure that most attributes are close; the second deviation threshold is used to determine whether the deviation mean between the two record values ​​is less than or equal to the threshold. If any one of them is not satisfied, such as the first deviation distance is greater than the first deviation distance threshold, or the second deviation distance is greater than the second deviation threshold, the two associated influencing factor record values ​​are divided into different groups. The dual deviation standard (proportion and mean) ensures that the records in the same group remain consistent in multiple dimensions, reducing the error caused by fluctuations in individual attributes.

[0042] Furthermore, the present application also includes the following steps: The feedforward neural network is trained according to the first historical operation cross-sectional data, and Q power generation deviations are obtained each time the training is performed Q times; when the Q power generation deviations are less than or equal to the proportion of the power generation deviation threshold, a prediction accuracy is generated; when the prediction accuracy is greater than or equal to the accuracy threshold, the initial power generation prediction model is obtained.

[0043] Specifically, the feedforward neural network is trained according to the first historical operation section data, with the associated influencing factors as input variables and the power generation as the target output. The training process optimizes the network parameters so that the model gradually adjusts the prediction effect after each training. After each Q training, the predicted power generation deviation in these Q times is recorded, that is, the error between the predicted power generation and the actual power generation in each training result. The power generation deviation is obtained by calculating the absolute value of the predicted power generation minus the actual power generation using measurement indicators such as mean square error (MSE) or absolute error (MAE). The power generation deviation of the Q training results is counted, and the proportion of the training times with a deviation less than or equal to the preset power generation deviation threshold is calculated, which is defined as the prediction accuracy. The power generation deviation threshold is an accuracy threshold defined in advance. Only the predicted power generation less than or equal to this threshold is accurate. Define an accuracy threshold to screen models that meet the requirements and ensure that the prediction effect of the model meets the actual application requirements. This standard is used to filter out models with excessive deviations, and finally obtain a stable and high-precision prediction model. When the prediction accuracy is greater than or equal to the accuracy threshold, the current feedforward neural network is considered to have been trained well enough and has reached the accuracy standard for the initial power generation prediction, and it is saved as the initial power generation prediction model. If it has not been reached, continue to adjust or retrain the model. Through multiple training and deviation statistics, random errors in a single training are avoided, the consistency of the model in multiple rounds of prediction is enhanced, and the prediction accuracy threshold is set so that the power generation prediction results of the model meet the set accuracy requirements. It is suitable for power generation prediction scenarios with high accuracy requirements.

[0044] In summary, the multi-dimensional adjustable potential assessment method of a virtual power plant based on a comprehensive intelligent model provided in this application has the following technical effects: By obtaining the first power supply node of the virtual power plant, wherein the first power supply node has several power generation units, and any power generation unit has a power generation type identifier and a unit position number identifier; performing power generation influencing factor analysis according to the power generation type identifier to obtain associated influencing factors; based on the unit position number identifier and the associated influencing factors, collecting historical operation cross-sectional data to train several power generation prediction models, wherein the several power generation prediction models correspond to the several power generation units one by one; according to the several power generation prediction models, the power generation potential of the first power supply node is evaluated in a preset time zone to obtain the predicted power supply of the first power supply node; according to the predicted power supply of the first power supply node, the power generation potential of the first power supply node in the virtual power plant is visually identified in the preset time zone. In other words, by identifying the key influencing factors of different types of power generation units according to the power generation type identifier; combining the position number identifier and influencing factor of each power generation unit, collecting the historical operation data of the equipment, modeling each power generation unit separately, and training multiple more targeted prediction models; according to the power supply capacity of the node predicted in the preset time zone, accurate management of each power generation unit is achieved, and the refinement of power generation potential evaluation is improved.

[0045] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0046] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on an integrated intelligent model, characterized in that: include: Obtaining a first power supply node of a virtual power plant, wherein the first power supply node has a plurality of power generation units, and any power generation unit has a power generation type identifier and a unit position number identifier; Perform power generation influencing factor analysis according to the power generation type identifier to obtain associated influencing factors; Based on the unit position identification and the associated influencing factor, historical operation cross-sectional data are collected to train a plurality of power generation prediction models, wherein the plurality of power generation prediction models correspond one-to-one to the plurality of power generation units; According to the plurality of power generation prediction models, the power generation potential of the first power supply node in the preset time zone is evaluated to obtain the predicted power supply of the first power supply node; The power generation potential of the first power supply node in a preset time zone is visually identified in the virtual power plant according to the predicted power supply of the first power supply node.

2. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 1, characterized in that: The power generation influencing factors are analyzed according to the power generation type identification to obtain associated influencing factors, including: Sending the power generation type identifier to a user terminal and receiving user feedback information, wherein the user feedback information includes an initial impact factor; Performing a correlation analysis on the initial impact factor according to the power generation type identifier to obtain a correlation degree of the initial impact factor; The initial impact factors whose correlation degree of the initial impact factors is greater than or equal to a correlation degree threshold are extracted and set as the correlation impact factors.

3. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 2, characterized in that: Performing a correlation analysis on the initial impact factor according to the power generation type identifier to obtain the correlation degree of the initial impact factor includes: Taking the power generation type identifier as a constraint, collecting a power generation record data set, wherein the power generation record data set includes an initial impact factor record value set and a power generation record value set; Normalizing the power generation record value set to generate a first data sequence; Normalizing the first attribute initial impact factor record value set of the initial impact factor record value set to obtain a second data sequence; Until the Nth attribute initial impact factor record value set of the initial impact factor record value set is normalized to obtain the Nth data sequence; Taking the first data sequence as the reference sequence and the second data sequence to the Nth data sequence as the comparison sequence, grey correlation analysis is performed to obtain the correlation of the initial influencing factors of the first attribute to the Nth attribute, and add them into the initial influencing factor correlation.

4. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 1, characterized in that: Based on the unit position identification and the associated influencing factors, historical operation cross-sectional data are collected to train several power generation prediction models, including: Taking the associated influencing factor as the mapping independent variable and the power generation as the mapping dependent variable, collecting the first historical operation cross-section data of the same type and scale as the power generation unit; Training a feedforward neural network according to the first historical operation cross-sectional data to obtain an initial power generation prediction model; Taking the associated influencing factor as a mapping independent variable and the power generation as a mapping dependent variable, collecting the second historical operation section data with the unit bit number identification; The initial power generation prediction model is trained according to the second historical operation cross-section data to generate a power generation prediction model, wherein the power generation prediction model belongs to the plurality of power generation prediction models.

5. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 4, characterized in that: Taking the associated influencing factor as the mapping independent variable and the power generation as the mapping dependent variable, collecting the first historical operation cross-section data of the same type and scale as the power generation unit, including: Taking the associated influencing factor as a mapping independent variable and the power generation as a mapping dependent variable, collecting a first associated influencing factor record value and a first power generation record value; Performing cluster analysis on the first correlation impact factor record values ​​to obtain multiple groups of first correlation impact factor record values; Grouping the first power generation record values ​​according to the multiple groups of first associated influencing factor record values ​​to obtain multiple groups of first power generation record values; Traversing the plurality of groups of first power generation record values ​​to perform power generation outlier removal fitting to obtain a plurality of first power generation fitting values; Traversing the multiple groups of first correlation impact factor record values ​​to randomly select values, and obtaining multiple first correlation impact factor sorting record values; The plurality of first power generation fitting values ​​and the plurality of first associated influencing factor sorting record values ​​are added to the first historical operation cross-sectional data.

6. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 5, characterized in that: Taking the associated influencing factor as a mapping independent variable and the power generation as a mapping dependent variable, collecting the second historical operation section data with the unit bit number identification, including: The data collection process of the second historical operation cross-sectional data is the same as the data collection process of the first historical operation cross-sectional data.

7. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 5, characterized in that: Performing cluster analysis on the first correlation impact factor record values ​​to obtain multiple groups of first correlation impact factor record values, including: Configure the correlation impact factor deviation threshold of the correlation impact factor record value; Extracting a first associated impact factor record value and a second associated impact factor record value of the first associated impact factor record value; Counting the proportion of attributes whose associated impact factor record value deviation between the first associated impact factor record value and the second associated impact factor record value is greater than or equal to the associated impact factor deviation threshold, and setting it as a first deviation distance; Counting the deviation average of the associated impact factor record value between the first associated impact factor record value and the second associated impact factor record value, and setting it as a second deviation distance; When the first deviation distance is less than or equal to a first deviation distance threshold, and the second deviation distance is less than or equal to a second deviation threshold, the first associated impact factor record value and the second associated impact factor record value are added as the same group, otherwise, they are set as different groups.

8. The method for evaluating the multi-dimensional adjustable potential of a virtual power plant based on a comprehensive intelligent model according to claim 4, characterized in that: According to the first historical operation section data, a feedforward neural network is trained to obtain an initial power generation prediction model, including: The feedforward neural network is trained according to the first historical operation cross-sectional data, and Q power generation deviations are obtained each time the training is performed Q times; When the Q times of power generation deviation is less than or equal to the power generation deviation threshold, a prediction accuracy rate is generated; When the prediction accuracy is greater than or equal to the accuracy threshold, the initial power generation prediction model is obtained.

Images