A high-precision learning power auxiliary decision-making method and system

Abstract

The present invention discloses a high-precision learning power auxiliary decision-making method and system, which relate to the technical field of power decision-making. The method comprises: performing feature recognition on first user information to obtain first user features; obtaining a power consumption record library of users in the same group; building a first power consumption prediction model; collecting power consumption records of the first user to obtain a second power consumption prediction model; performing power consumption prediction to obtain the first user's predicted power consumption; obtaining a set of power trading varieties; performing power auxiliary decision-making analysis based on the power trading variety set based on the first user's predicted power consumption to obtain a first recommended power trading decision, and sending the first recommended power trading decision to the first user. The present application can solve the problems in the prior art of being unable to perform adaptive power trading variety recommendations for users and the low accuracy of power trading auxiliary decision-making, thereby achieving the effect of improving the accuracy of predicted power consumption and thus improving the accuracy of power trading auxiliary decision-making.

Description

Technical Field

[0001] The present disclosure relates to the technical field of electric power decision-making, and in particular to a high-precision learning electric power auxiliary decision-making method and system. Background Art

[0002] At present, the electricity trading market is in a stage of rapid development. As the scale of the electricity trading market continues to expand, trading rules continue to improve, and trading complexity increases, it is of great significance to assist market players participating in electricity trading to quickly formulate reasonable trading strategies.

[0003] In summary, the existing technology has the problems of being unable to adaptively recommend electricity trading products to users and low accuracy of auxiliary decision-making in electricity trading. Summary of the invention

[0004] The present invention provides a high-precision learning power auxiliary decision-making method and system to solve the technical problems in the prior art of being unable to adaptively recommend power trading products to users and having low accuracy in power trading auxiliary decision-making.

[0005] According to a first aspect of the present disclosure, a high-precision learning method for power auxiliary decision-making is provided, including: collecting basic information of a first user to obtain the first user information, and performing feature recognition on the first user information to obtain the first user feature; based on the first user feature, obtaining users in the same group, and collecting power consumption records based on the users in the same group to obtain a power consumption record library of users in the same group; performing high-precision learning based on the power consumption record library of users in the same group to build a first power consumption prediction model; collecting the power consumption record of the first user to obtain a first power consumption record library, and performing power consumption prediction sensitivity learning on the first power consumption prediction model based on the first power consumption record library to obtain a second power consumption prediction model; based on a first prediction time zone, performing power consumption prediction according to the second power consumption prediction model to obtain the predicted power consumption of the first user; obtaining a power trading variety set, wherein the power trading variety set includes M power trading varieties, wherein M is a positive integer greater than 1; based on the predicted power consumption of the first user, performing power auxiliary decision-making analysis according to the power trading variety set to obtain a first recommended power trading decision, and sending the first recommended power trading decision to the first user.

[0006] According to a second aspect of the present disclosure, a high-precision learning power auxiliary decision-making system is provided, including: a first user feature acquisition module, the first user feature acquisition module is used to collect basic information of the first user, obtain the first user information, and perform feature recognition on the first user information to obtain the first user feature; a module for obtaining a library of electricity consumption records of users in the same group, the module for obtaining a library of electricity consumption records of users in the same group is used to obtain users in the same group based on the first user feature, and collect electricity consumption records based on the users in the same group to obtain a library of electricity consumption records of users in the same group; a first electricity consumption prediction model acquisition module, the first electricity consumption prediction model acquisition module is used to perform high-precision learning based on the library of electricity consumption records of users in the same group to build a first electricity consumption prediction model; a second electricity consumption prediction model acquisition module, the second electricity consumption prediction model acquisition module is used to collect the electricity consumption record of the first user, obtain the first electricity consumption record library , and based on the first electricity consumption record library, the first electricity consumption prediction model is subjected to electricity consumption prediction sensitivity learning to obtain a second electricity consumption prediction model; a first user predicted electricity consumption acquisition module, the first user predicted electricity consumption acquisition module is used to perform electricity consumption prediction based on the first prediction time zone and according to the second electricity consumption prediction model to obtain the first user predicted electricity consumption; a power transaction variety set acquisition module, the power transaction variety set acquisition module is used to obtain a power transaction variety set, wherein the power transaction variety set includes M power transaction varieties, wherein M is a positive integer greater than 1; a first recommended power transaction decision acquisition module, the first recommended power transaction decision acquisition module is used to perform power auxiliary decision analysis based on the first user predicted electricity consumption and according to the power transaction variety set to obtain a first recommended power transaction decision, and send the first recommended power transaction decision to the first user.

[0007] According to a third aspect of the present disclosure, a computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program: collecting basic information of a first user to obtain the first user information, and performing feature recognition on the first user information to obtain the first user feature; based on the first user feature, obtaining users in the same group, and collecting electricity consumption records based on the users in the same group to obtain an electricity consumption record library of users in the same group; performing high-precision learning based on the electricity consumption record library of users in the same group to build a first electricity consumption prediction model; collecting the electricity consumption record of the first user to obtain a first electricity consumption record library, and performing electricity consumption prediction sensitivity learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model; based on a first prediction time zone, performing electricity consumption prediction according to the second electricity consumption prediction model to obtain the predicted electricity consumption of the first user; obtaining a set of electricity trading varieties, wherein the set of electricity trading varieties includes M electricity trading varieties, wherein M is a positive integer greater than 1; based on the predicted electricity consumption of the first user, performing electricity auxiliary decision analysis according to the electricity trading variety set to obtain a first recommended electricity trading decision, and sending the first recommended electricity trading decision to the first user.

[0008] According to a fourth aspect of the present disclosure, a computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the following steps are implemented: collecting basic information of a first user, obtaining the first user information, and performing feature recognition on the first user information to obtain first user features; based on the first user features, obtaining users in the same group, and collecting electricity consumption records based on the users in the same group to obtain an electricity consumption record library of users in the same group; performing high-precision learning based on the electricity consumption record library of users in the same group to build a first electricity consumption prediction model; collecting electricity consumption records of the first user to obtain a first electricity consumption record library, and performing electricity consumption prediction sensitivity learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model; based on a first prediction time zone, performing electricity consumption prediction according to the second electricity consumption prediction model to obtain the predicted electricity consumption of the first user; obtaining a set of electricity trading varieties, wherein the set of electricity trading varieties includes M electricity trading varieties, wherein M is a positive integer greater than 1; based on the predicted electricity consumption of the first user, performing electricity auxiliary decision analysis according to the set of electricity trading varieties to obtain a first recommended electricity trading decision, and sending the first recommended electricity trading decision to the first user.

[0009] One or more technical solutions provided in the present application have at least the following technical effects or advantages: according to the method adopted in the present invention, basic information of the first user is collected to obtain the first user information, and feature recognition is performed on the first user information to obtain the first user feature; based on the first user feature, users in the same group are obtained, and electricity consumption records are collected based on the users in the same group to obtain an electricity consumption record library of users in the same group; high-precision learning is performed based on the electricity consumption record library of users in the same group to build a first electricity consumption prediction model; electricity consumption records of the first user are collected to obtain a first electricity consumption record library, and electricity consumption prediction sensitivity learning is performed on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model; based on the first electricity consumption record library, the electricity consumption prediction sensitivity of the first electricity consumption prediction model is learned to obtain a second electricity consumption prediction model. A prediction time zone is provided, and electricity consumption is predicted according to the second electricity consumption prediction model to obtain the predicted electricity consumption of the first user; a set of electricity trading varieties is obtained, wherein the set of electricity trading varieties includes M electricity trading varieties, wherein M is a positive integer greater than 1; based on the predicted electricity consumption of the first user, an electricity auxiliary decision-making analysis is performed according to the set of electricity trading varieties to obtain a first recommended electricity trading decision, and the first recommended electricity trading decision is sent to the first user, which solves the problems in the prior art that it is impossible to adaptively recommend electricity trading varieties to users and the low accuracy of electricity trading auxiliary decision-making, and achieves the technical effect of improving the accuracy of predicted electricity consumption and thus improving the accuracy of electricity trading auxiliary decision-making.

[0010] It should be understood that the contents described in this section are not intended to indicate the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present disclosure 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.

[0012] Figure 1 A flowchart of a high-precision learning power auxiliary decision-making method provided by an embodiment of the present disclosure;

[0013] Figure 2 A schematic diagram of a process of building a first power consumption prediction model in a high-precision learning power auxiliary decision-making method according to an embodiment of the present disclosure;

[0014] Figure 3 A schematic diagram of a flow chart of obtaining a second power consumption prediction model in a power auxiliary decision-making method of high-precision learning according to an embodiment of the present disclosure;

[0015] Figure 4 A schematic diagram of the structure of a high-precision learning power auxiliary decision-making system provided by an embodiment of the present disclosure;

[0016] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure.

[0017] Explanation of the accompanying drawings: first user feature acquisition module 11, same group user electricity consumption record library acquisition module 12, first electricity consumption prediction model acquisition module 13, second electricity consumption prediction model acquisition module 14, first user predicted electricity consumption acquisition module 15, electricity trading variety set acquisition module 16, first recommended electricity trading decision acquisition module 17, electronic device 800, processor 801, memory 802, bus 803. DETAILED DESCRIPTION

[0018] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0019] Embodiment 1

[0020] Figure 1 A high-precision learning power auxiliary decision-making method provided in an embodiment of the present application includes:

[0021] Step S100: Collect basic information of a first user to obtain first user information, and perform feature recognition on the first user information to obtain first user features;

[0022] Specifically, the first user is a user for whom electricity consumption is to be predicted. The first user is retrieved based on big data with the first user as the index condition to obtain the historical electricity consumption data of the first user. The basic information of the first user is collected based on the historical electricity consumption data of the first user to obtain the first user information. Furthermore, the first user information includes information such as the first user's electricity consumption, electricity consumption time, and weather information. Furthermore, the first user information is feature identified, the first user information is selected for identification and extraction, and the first user feature is obtained. The first user feature includes electricity load, electricity consumption time information, electricity consumption weather information, etc. Further, the first user feature is obtained to analyze the first user's electricity consumption behavior based on the first user feature.

[0023] Step S200: obtaining users in the same group based on the first user feature, and collecting electricity consumption records based on the users in the same group to obtain an electricity consumption record library of users in the same group;

[0024] Specifically, based on the first user feature, a user group having the same feature as the first user feature is obtained as the same group of users. Further, historical electricity consumption data is obtained based on big data, electricity consumption records of the same group of users are collected according to the historical electricity consumption data, and the electricity consumption records of the same group of users are extracted to generate an electricity consumption record library of the same group of users. Among them, the electricity consumption record library includes the electricity consumption information of the same group of users, the electricity consumption time information of the same group of users, etc. Further, the electricity consumption record library of the same group of users is obtained to select indicators for the electricity consumption behavior of the same group of users, and then analyze the electricity consumption behavior of the same group of users.

[0025] Step S300: Perform high-precision learning based on the electricity consumption record library of users in the same group to build a first electricity consumption prediction model;

[0026] Specifically, the electricity consumption record library of the same group of users includes multiple electricity consumption records of users, wherein the multiple electricity consumption records of users have multiple correspondingly associated electricity consumption time zone identifiers. Further, the multiple electricity consumption records of users are screened based on the multiple electricity consumption time zone identifiers to obtain multiple associated user electricity consumption records. Further, multiple operators are configured according to the multiple associated user electricity consumption records, and the multiple associated user electricity consumption records are supervised and trained by the multiple operators to generate a first electricity consumption prediction model to improve the accuracy of electricity consumption prediction.

[0027] Step S400: collecting the electricity consumption record of the first user to obtain a first electricity consumption record library, and performing electricity consumption prediction sensitivity learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model;

[0028] Specifically, based on the historical electricity consumption data of the first user, the electricity consumption record of the first user is collected to obtain a first electricity consumption record library, wherein the first electricity consumption record library includes the electricity consumption information of the first user, the electricity consumption time information of the first user, etc. Further, based on the first electricity consumption record library, the first electricity consumption prediction model is subjected to electricity consumption prediction sensitivity learning, that is, the first electricity consumption record library is used as input data to perform incremental learning on the first electricity consumption prediction model, and the first electricity consumption prediction model is subjected to supervised training to obtain a second electricity consumption prediction model, thereby improving the accuracy of electricity consumption prediction.

[0029] Step S500: Based on the first prediction time zone, power consumption prediction is performed according to the second power consumption prediction model to obtain the predicted power consumption of the first user;

[0030] Specifically, the first prediction time zone is a time zone for predicting the power consumption of the first user. In the first prediction time zone, the first power consumption record library is input into the second power consumption prediction model to predict the power consumption, and the predicted power consumption of the first user is obtained. The predicted power consumption of the first user is the power consumption data for predicting the power consumption of the first user. Furthermore, the accuracy of recommending power trading plans to users can be improved according to the predicted power consumption of the first user.

[0031] Step S600: obtaining a power transaction type set, wherein the power transaction type set includes M power transaction types, wherein M is a positive integer greater than 1;

[0032] Specifically, a power transaction variety set is obtained based on historical power consumption data, wherein the power transaction variety set includes multiple power transaction varieties. For example, the power transaction variety set includes power generation rights trading, external power transmission trading, large-user direct power purchase trading, short-term trading, etc. Further, the power transaction variety set includes M power transaction varieties, and the power transaction variety set includes at least 1 power transaction variety, so M is a positive integer greater than 1. Further, the power transaction variety set is obtained for subsequent auxiliary decision-making.

[0033] Step S700: Based on the predicted power consumption of the first user, an auxiliary power decision analysis is performed according to the power trading product set to obtain a first recommended power trading decision, and the first recommended power trading decision is sent to the first user.

[0034] Specifically, the predicted power consumption of the first user is evaluated for matching with M power trading varieties to obtain the variety matching degree. The variety abundance and variety trading circulation are obtained according to the M power trading varieties. Further, the variety matching degree, variety abundance, and variety trading circulation are weighted according to the preset weight coefficient to obtain the M variety fitness. Further, the M variety fitness is compared with the preset variety fitness to obtain multiple power trading varieties that meet the preset variety fitness. Further, according to the M power trading varieties and the M variety fitness, multiple power trading varieties are matched to obtain multiple power trading varieties and their corresponding multiple variety fitness. Further, the fitness ratio of multiple variety fitness is calculated with the fitness of M varieties to obtain the coefficient of the fitness ratio of multiple variety fitness to the fitness of M varieties, that is, the fitness ratio coefficient of multiple varieties. Further, based on the fitness ratio coefficient of multiple varieties, multiple power trading varieties are serialized according to the numerical values ​​of the fitness ratio coefficients of multiple varieties to obtain the first recommended power trading decision.

[0035] Among them, this embodiment can solve the problems in the prior art of being unable to adaptively recommend electricity trading products to users and the low accuracy of electricity trading auxiliary decision-making, thereby achieving the effect of improving the accuracy of predicted electricity consumption and thus improving the accuracy of electricity trading auxiliary decision-making.

[0036] like Figure 2 As shown, step S300 in the method provided in the embodiment of the present application includes:

[0037] S310: The electricity consumption record library of the same group of users includes multiple electricity consumption records of users, and the multiple electricity consumption records of users have multiple electricity consumption time zone identifiers corresponding to each other;

[0038] S320: Filtering the multiple user electricity usage records based on the multiple user electricity usage time zone identifiers to obtain multiple associated user electricity usage records;

[0039] S330: configuring a plurality of high-precision learning operators based on the plurality of associated user electricity consumption records;

[0040] S340: Based on a fully connected neural network, cross-supervised training is performed on the electricity consumption records of the multiple associated users according to the multiple high-precision learning operators to generate the first electricity consumption prediction model.

[0041] Specifically, the electricity consumption record library of the same group of users includes multiple electricity consumption records of users, wherein the multiple electricity consumption records of users include electricity consumption records and electricity consumption time records. Further, the multiple electricity consumption records of users have multiple electricity consumption time zone identifiers corresponding to each other. Among them, the multiple electricity consumption time zone identifiers of users are electricity consumption date identifiers and electricity consumption time identifiers. Further, each user electricity consumption record is correspondingly associated with each user electricity consumption time zone identifier.

[0042] Furthermore, based on multiple user power usage time zone identifiers, a first user power usage time zone identifier and its corresponding first user power usage record are obtained. Furthermore, a correlation analysis is performed on the first user power usage time zone identifier and the first predicted time zone. If the first user power usage time zone identifier and the first predicted time zone have a high correlation, the first user power usage record is added to multiple associated user power usage records. The associated user power usage record is obtained by filtering multiple user power usage records through multiple user power usage time zone identifiers.

[0043] Further, based on the electricity consumption records of multiple associated users, the time zone correlations corresponding to the electricity consumption records of multiple associated users are obtained. Among them, the time zone correlations corresponding to the electricity consumption records of multiple associated users are calculated with the multiple time zone correlations to obtain the proportion coefficient. Further, according to the proportion coefficient, multiple high-precision learning operators are configured, wherein the operator is a calculation unit that constitutes the deep learning algorithm. Further, in the network model, the operator is the calculation logic in the network model layer. For example, multiple high-precision learning operators include statistical calculation operators, format conversion operators, and result presentation operators.

[0044] Furthermore, in the fully connected neural network, each neuron is connected to all neurons in the previous layer and the next layer to form a dense connection structure. Furthermore, multiple associated user electricity consumption records are input as input data into the fully connected neural network. The multiple associated user electricity consumption records are cross-supervised trained according to multiple high-precision learning operators, that is, multiple high-precision learning operators perform supervised training on each input data to generate a first electricity consumption prediction model. Among them, supervised training is a process of adjusting the parameters of the classifier using a set of samples of known categories to achieve the required performance.

[0045] Among them, the first electricity consumption prediction model is built to predict electricity consumption behavior and improve the accuracy of transaction-assisted decision-making.

[0046] Step S320 in the method provided in the embodiment of the present application includes:

[0047] S321: traverse the multiple user power usage time zone identifiers to obtain a first user power usage time zone identifier, and match a first user power usage record corresponding to the first user power usage time zone identifier based on the multiple user power usage records;

[0048] S322: Based on the twin neural network, performing a correlation analysis on the first user's electricity consumption time zone identifier and the first predicted time zone to obtain a first time zone correlation degree;

[0049] S323: Determine whether the first time zone correlation degree meets a preset time zone correlation degree;

[0050] S324: If the first time zone association degree meets the preset time zone association degree, add the first user power usage record to the multiple associated user power usage records.

[0051] Specifically, multiple user power usage time zone identifiers are accessed in sequence, and a group of user power usage time zone identifiers are randomly extracted as the first user power usage time zone identifier. Further, multiple user power usage records are accessed in sequence, and the first user power usage time zone identifier is extracted for matching to obtain the first user power usage record corresponding to the first user power usage time zone identifier.

[0052] Furthermore, the twin neural network takes two samples as input, and its two sub-networks each receive an input, output their representation embedded in a high-dimensional space, and compare the similarity of the two samples by calculating the distance between the two representations, such as the Euclidean distance. Furthermore, based on the twin neural network, the first user's electricity consumption time zone identifier and the first predicted time zone are analyzed for correlation through correlation analysis, regression analysis, cross-tabulation chi-square analysis and other methods to obtain the first time zone correlation. Among them, the first predicted time zone is the time zone for predicting the electricity consumption of the first user.

[0053] Furthermore, a time zone correlation is preset based on historical power consumption data, wherein the preset time zone correlation is that there is a strong correlation between multiple user power consumption time zone identifiers and corresponding multiple predicted time zones. Further, the first time zone correlation is compared with the preset time zone correlation to determine whether the first time zone correlation satisfies the preset time zone correlation, and a time zone determination result is obtained.

[0054] Furthermore, if the time zone determination result is that the first time zone correlation degree meets the preset time zone correlation degree, the first user power consumption record is added to multiple associated user power consumption records. The associated user power consumption record is obtained by screening multiple user power consumption records through multiple user power consumption time zone identifiers.

[0055] Among them, the associated user electricity consumption records are used to improve the accuracy of electricity consumption behavior prediction.

[0056] Step S330 in the method provided in the embodiment of the present application includes:

[0057] S331: Based on the multiple associated user electricity consumption records, obtain multiple record time zone association degrees;

[0058] S332: Calculating the correlation ratio based on the multiple recorded time zone correlations to obtain multiple record correlation ratio coefficients;

[0059] S333: Allocate training computing power to the multiple associated user electricity consumption records based on the multiple record association ratio coefficients to obtain the multiple high-precision learning operators.

[0060] Specifically, based on multiple associated user power consumption records, multiple record time zone associations are obtained, wherein the multiple record time zone associations are multiple time zone associations corresponding to the multiple associated user power consumption records. Further, the method for obtaining the multiple time zone associations is the same as the method for obtaining the first time zone association.

[0061] Furthermore, the correlation ratio is calculated based on the correlation of multiple time zones, that is, the coefficients of the correlation of multiple time zones are the correlation of multiple time zones, and multiple record correlation ratio coefficients are obtained. Furthermore, the training computing power is allocated to multiple associated user electricity consumption records based on the multiple record correlation ratio coefficients to obtain multiple high-precision learning operators. Among them, computing power is the computing ability to achieve the target result output by processing information data.

[0062] Among them, a plurality of high-precision learning operators are obtained to perform supervised training on the electricity consumption records of a plurality of associated users through the operators, so as to obtain a first electricity consumption prediction model and improve the accuracy of electricity consumption prediction.

[0063] like Figure 3 As shown, step S400 in the method provided in the embodiment of the present application includes:

[0064] S410: performing incremental learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a first incremental electricity consumption prediction model and a first electricity loss data set;

[0065] S420: Obtaining a first loss data amount of the first power loss data set;

[0066] S430: Determine whether the first loss data amount is greater than a preset loss data amount;

[0067] S440: If the first loss data volume is greater than the preset loss data volume, supervised training is performed on the first incremental electricity consumption prediction model based on the first electricity loss data set to generate the second electricity consumption prediction model.

[0068] Specifically, incremental learning is performed on the first electricity consumption prediction model based on the first electricity consumption record library, wherein incremental learning is to enhance the trained model by adding new data and labels without retraining the entire model. Further, the parameters of the first electricity consumption prediction model are initialized, the first electricity consumption record library of new samples is received, the membership of the first electricity consumption record library of new samples is calculated, the cluster center is updated, the membership of all samples is updated, and the cluster center and the membership of all samples are repeatedly updated until certain conditions are met to complete incremental learning. Further, by performing incremental learning on the first electricity consumption prediction model, a first incremental electricity consumption prediction model and a first electricity loss data set are obtained. The first electricity loss data set includes multiple first electricity loss data. Further, the first loss data amount of the first electricity loss data set is obtained. The first loss data amount is the number of first electricity loss data.

[0069] Furthermore, based on the historical incremental learning preset loss data amount, it is determined whether the first loss data amount is greater than the preset loss data amount to obtain a loss judgment result. Furthermore, if the first loss data amount is greater than the preset loss data amount, the first electricity loss data set is extracted as input data and input into the first incremental electricity prediction model, and the first incremental electricity prediction model is supervised trained to generate the second electricity prediction model. Among them, the first incremental electricity prediction model is supervised trained, and the supervised training process is to first train with known inputs and corresponding outputs to obtain an optimal first incremental electricity prediction model, and then apply the first incremental electricity prediction model to the new data, that is, extract the first electricity loss data set and input it into the first incremental electricity prediction model to map the output results. After going through this process, the model has predictive ability.

[0070] The first electricity consumption prediction model is trained and learned to obtain a second electricity consumption prediction model, and the second electricity consumption prediction model can obtain electricity consumption prediction output data with higher accuracy than the first electricity consumption prediction model.

[0071] Step S700 in the method provided in the embodiment of the present application includes:

[0072] S710: Based on the predicted power consumption of the first user, respectively evaluate the fitness of the M power trading varieties to obtain the fitness of the M varieties;

[0073] S720: Based on the preset variety fitness, the M power trading varieties are screened according to the M variety fitness to obtain Q power trading varieties that meet the preset variety fitness, where Q is a positive integer and 1<Q≤M;

[0074] S730: Identifying the fitness of the M varieties based on the Q electricity trading varieties to obtain the fitness of the Q varieties;

[0075] S740: Calculate the fitness ratio based on the fitness of the Q varieties to obtain the fitness ratio coefficients of the Q varieties, and make incentive adjustments to the Q power trading varieties based on the fitness ratio coefficients of the Q varieties to generate the first recommended power trading decision.

[0076] Specifically, the predicted power consumption of the first user is evaluated for matching with the M power trading varieties to obtain the variety matching degree. The variety sufficiency of the M power trading varieties is identified to obtain the variety sufficiency. The variety trading circulation of the M power trading varieties is identified to obtain the variety trading circulation. Furthermore, the variety matching degree, variety sufficiency, and variety trading circulation are weighted according to the preset weight coefficient to obtain the M variety adaptability.

[0077] Furthermore, the fitness of the varieties is preset according to the historical electricity consumption data. The fitness of the M varieties is compared with the preset variety fitness to obtain Q electricity trading varieties that meet the preset variety fitness, where Q is a positive integer and 1<Q≤M.

[0078] Furthermore, the Q power trading varieties are matched and identified with the M power trading varieties, and the Q power trading varieties and their corresponding Q variety fitness are obtained by matching. Furthermore, the fitness ratio of the Q variety fitness and the M variety fitness is calculated to obtain the coefficient of the fitness ratio of the Q variety fitness to the fitness ratio of the M variety, that is, the fitness ratio coefficient of the Q variety. Furthermore, the Q power trading varieties are incentive-regulated based on the fitness ratio coefficient of the Q varieties, and the first recommended power trading decision is generated. Among them, the incentive regulation method is to serialize the Q power trading varieties according to the numerical value of the fitness ratio coefficient of the Q varieties, and obtain the serialization processing result, that is, the first recommended power trading decision.

[0079] The first recommended electricity trading decision is obtained to recommend a decision to the first user.

[0080] Step S710 in the method provided in the embodiment of the present application includes:

[0081] S711: Based on the M power trading varieties, obtain the i-th power trading variety, and 1≤i≤M;

[0082] S712: performing a matching evaluation based on the predicted power consumption of the first user and the i-th power trading product to obtain a matching degree of the i-th product;

[0083] S713: Identify the abundance of the i-th power trading product based on the i-th power trading product to obtain the abundance of the i-th power trading product;

[0084] S714: Identifying the transaction circulation degree of the i-th power transaction product based on the i-th power transaction product to obtain the transaction circulation degree of the i-th power transaction product;

[0085] S715: Based on the preset fitness evaluation constraints, weighted calculation is performed on the i-th variety matching degree, the i-th variety abundance degree and the i-th variety transaction circulation degree to generate the i-th variety fitness, and the i-th variety fitness is added to the M variety fitnesses.

[0086] Specifically, based on M power trading varieties, the i-th power trading variety is randomly extracted to obtain. Further, the i-th power trading variety is evaluated for matching with the predicted power consumption of the first user to obtain the i-th variety matching degree. Among them, the i-th variety matching degree is used to evaluate the matching degree between the i-th power trading variety and the first user's power trading variety. Further, the i-th power trading variety is identified with the power trading variety set for variety abundance to obtain the i-th variety abundance. Among them, the i-th variety abundance is used to evaluate the abundance of the i-th power trading variety. Further, the i-th power trading variety is identified with the power trading variety set for variety trading circulation to obtain the i-th variety trading circulation. Among them, the i-th variety trading circulation is used to evaluate the circulation breadth of the i-th power trading variety.

[0087] Furthermore, a fitness evaluation constraint is preset based on historical electricity consumption data. The preset fitness evaluation constraint is a weight coefficient of multiple evaluation categories. The multiple evaluation categories include the i-th variety matching degree, the i-th variety abundance, and the i-th variety transaction circulation degree. Further, based on the weight coefficients of multiple evaluation categories in the preset fitness evaluation constraint, the i-th variety matching degree, the i-th variety abundance, and the i-th variety transaction circulation degree are weightedly calculated to generate the i-th variety fitness, so as to distinguish the differences between the effects produced by different i-th variety fitnesses. Further, the i-th variety fitness is added to the M variety fitnesses.

[0088] Among them, the fitness of M power trading varieties is evaluated respectively, and the fitness of M varieties is obtained to determine the difference between the effects produced by the fitness of different i-th varieties, and then conduct subsequent screening.

[0089] Embodiment 2

[0090] Based on the same inventive concept as the power auxiliary decision method with high precision learning in the above-mentioned embodiment, Figure 4 As shown, the present application also provides a high-precision learning power auxiliary decision-making system, the system comprising:

[0091] A first user feature acquisition module 11, the first user feature acquisition module is used to collect basic information of the first user, obtain first user information, and perform feature recognition on the first user information to obtain first user features;

[0092] A module 12 for obtaining a power consumption record library of users in the same group, wherein the module is used to obtain users in the same group based on the first user characteristics, and to collect power consumption records based on the users in the same group to obtain a power consumption record library of users in the same group;

[0093] A first electricity consumption prediction model acquisition module 13, the first electricity consumption prediction model acquisition module is used to perform high-precision learning based on the electricity consumption record library of the same group of users to build a first electricity consumption prediction model;

[0094] a second electricity consumption prediction model obtaining module 14, the second electricity consumption prediction model obtaining module being used to collect the electricity consumption record of the first user, obtain a first electricity consumption record library, and perform electricity consumption prediction sensitivity learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model;

[0095] A first user predicted power consumption obtaining module 15, the first user predicted power consumption obtaining module is used to perform power consumption prediction based on the first prediction time zone and the second power consumption prediction model to obtain the first user predicted power consumption;

[0096] An electric power transaction variety set obtaining module 16, wherein the electric power transaction variety set obtaining module is used to obtain an electric power transaction variety set, wherein the electric power transaction variety set includes M electric power transaction varieties, wherein M is a positive integer greater than 1;

[0097] The first recommended power trading decision obtaining module 17 is used to perform power auxiliary decision analysis based on the predicted power consumption of the first user and the power trading product set to obtain the first recommended power trading decision, and send the first recommended power trading decision to the first user.

[0098] Furthermore, the system further comprises:

[0099] A user electricity consumption record acquisition module, wherein the user electricity consumption record acquisition module is used for the electricity consumption record library of the same group of users to include multiple user electricity consumption records, and the multiple user electricity consumption records have multiple user electricity consumption time zone identifiers corresponding to each other;

[0100] an associated user electricity consumption record obtaining module, the associated user electricity consumption record obtaining module being used to filter the multiple user electricity consumption records based on the multiple user electricity consumption time zone identifiers to obtain multiple associated user electricity consumption records;

[0101] A high-precision learning operator processing module, the high-precision learning operator processing module is used to configure multiple high-precision learning operators based on the multiple associated user electricity consumption records;

[0102] The first electricity consumption prediction model acquisition module is used to generate the first electricity consumption prediction model by performing cross-supervised training on the electricity consumption records of the multiple associated users based on a fully connected neural network and according to the multiple high-precision learning operators.

[0103] Furthermore, the system further comprises:

[0104] a first user electricity usage time zone identification obtaining module, the first user electricity usage time zone identification obtaining module being used to traverse the multiple user electricity usage time zone identifications, obtain the first user electricity usage time zone identification, and match the first user electricity usage record corresponding to the first user electricity usage time zone identification based on the multiple user electricity usage records;

[0105] A first time zone correlation degree obtaining module, wherein the first time zone correlation degree obtaining module is used to perform a correlation analysis on the first user's electricity consumption time zone identifier and the first predicted time zone based on a twin neural network to obtain a first time zone correlation degree;

[0106] A first time zone correlation processing module, the first time zone correlation processing module is used to determine whether the first time zone correlation satisfies a preset time zone correlation;

[0107] The associated user electricity consumption record obtaining module is used for adding the first user electricity consumption record to the multiple associated user electricity consumption records if the first time zone association degree meets the preset time zone association degree.

[0108] Furthermore, the system further comprises:

[0109] A recording time zone association degree obtaining module, the recording time zone association degree obtaining module is used to obtain multiple recording time zone association degrees based on the multiple associated user power consumption records;

[0110] A record correlation ratio coefficient obtaining module, the record correlation ratio coefficient obtaining module is used to calculate the correlation ratio based on the multiple record time zone correlations to obtain multiple record correlation ratio coefficients;

[0111] A plurality of high-precision learning operator acquisition modules are used to allocate training computing power to the plurality of associated user electricity consumption records based on the plurality of record association ratio coefficients to obtain the plurality of high-precision learning operators.

[0112] Furthermore, the system further comprises:

[0113] a first power loss data set acquisition module, the first power loss data set acquisition module being used to perform incremental learning on the first power consumption prediction model based on the first power consumption record library to obtain a first incremental power consumption prediction model and a first power loss data set;

[0114] A first loss data amount obtaining module, the first loss data amount obtaining module is used to obtain a first loss data amount of the first power loss data set;

[0115] A first loss data amount processing module, the first loss data amount processing module is used to determine whether the first loss data amount is greater than a preset loss data amount;

[0116] The second electricity consumption prediction model acquisition module is used to perform supervised training on the first incremental electricity consumption prediction model based on the first electricity loss data set to generate the second electricity consumption prediction model if the first loss data amount is greater than the preset loss data amount.

[0117] Furthermore, the system further comprises:

[0118] M variety fitness obtaining modules, the M variety fitness obtaining modules being used to respectively evaluate the fitness of the M power trading varieties based on the predicted power consumption of the first user, and obtain the fitness of the M varieties;

[0119] Q power trading variety acquisition modules, the Q power trading variety acquisition modules are used to screen the M power trading varieties according to the M variety fitnesses based on preset variety fitnesses, and obtain Q power trading varieties that meet the preset variety fitnesses, wherein Q is a positive integer, and 1<Q≤M;

[0120] Q variety fitness acquisition modules, the Q variety fitness acquisition modules are used to identify the fitness of the M varieties based on the Q power trading varieties to obtain the fitness of the Q varieties;

[0121] An incentive adjustment processing module, wherein the incentive adjustment processing module is used to calculate the fitness ratio based on the fitness ratios of the Q varieties, obtain the fitness ratio coefficients of the Q varieties, and perform incentive adjustment on the Q power trading varieties based on the fitness ratio coefficients of the Q varieties to generate the first recommended power trading decision.

[0122] Furthermore, the system further comprises:

[0123] The i-th power transaction type acquisition module, the Q power transaction type acquisition modules are used to obtain the i-th power transaction type based on the M power transaction types, and 1≤i≤M;

[0124] An i-th variety matching degree obtaining module, the i-th variety matching degree obtaining module is used to evaluate the matching degree based on the first user's predicted power consumption and the i-th power trading variety, and obtain the i-th variety matching degree;

[0125] an i-th variety sufficiency obtaining module, the i-th variety sufficiency obtaining module being used to identify variety sufficiency based on the i-th power trading variety and obtain the i-th variety sufficiency;

[0126] The module for obtaining the transaction circulation degree of the i-th variety is used to identify the transaction circulation degree of the variety based on the i-th power trading variety and obtain the transaction circulation degree of the i-th variety;

[0127] M variety fitness acquisition modules, the M variety fitness acquisition modules are used to perform weighted calculations on the i-th variety matching degree, the i-th variety abundance degree and the i-th variety transaction circulation degree based on preset fitness evaluation constraints, generate the i-th variety fitness, and add the i-th variety fitness to the M variety fitnesses.

[0128] The specific example of a high-precision learning power auxiliary decision-making method in the aforementioned embodiment 1 is also applicable to a high-precision learning power auxiliary decision-making system in this embodiment. Through the aforementioned detailed description of a high-precision learning power auxiliary decision-making method, those skilled in the art can clearly know a high-precision learning power auxiliary decision-making system in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0129] Embodiment 3

[0130] Figure 5 is a schematic diagram according to a third embodiment of the present disclosure, Figure 5 As shown, the computer device 800 in the present disclosure may include: a processor 801 and a memory 802 .

[0131] The memory 802 is used to store programs; the memory 802 may include volatile memory (English: volatile memory), such as random-access memory (English: random-access memory, abbreviated: RAM), such as static random-access memory (English: static random-access memory, abbreviated: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviated: DDR SDRAM), etc.; the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory). The memory 802 is used to store computer programs (such as applications, functional modules, etc. that implement the above method), computer instructions, etc., and the above computer programs, computer instructions, etc. can be partitioned and stored in one or more memories 802. And the above computer programs, computer instructions, data, etc. can be called by the processor 801.

[0132] The above-mentioned computer programs, computer instructions, etc. may be stored in partitions in one or more memories 802 . And the above-mentioned computer programs, computer instructions, etc. may be called by the processor 801 .

[0133] The processor 801 is used to execute the computer program stored in the memory 802 to implement each step of the method involved in the above embodiment.

[0134] For details, please refer to the relevant description in the previous method embodiment.

[0135] The processor 801 and the memory 802 may be independent structures or integrated structures. When the processor 801 and the memory 802 are independent structures, the memory 802 and the processor 801 may be coupled and connected via a bus 803 .

[0136] The electronic device of this embodiment can execute the technical solution in the above method, and its specific implementation process and technical principle are the same, which will not be repeated here.

[0137] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0138] According to an embodiment of the present disclosure, the present disclosure also provides a computer program product, which includes: a computer program, the computer program is stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device executes the solution provided by any of the above embodiments.

[0139] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of this disclosure can be achieved, and this document is not limited here.

[0140] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A high-precision learning power auxiliary decision-making method, characterized in that: The method comprises: Collecting basic information of the first user to obtain first user information, and performing feature recognition on the first user information to obtain first user features; Based on the first user feature, obtaining users in the same group, and collecting electricity usage records based on the users in the same group to obtain an electricity usage record library of users in the same group; Based on the electricity consumption record library of the same group of users, high-precision learning is performed to build a first electricity consumption prediction model; Collecting the electricity consumption record of the first user to obtain a first electricity consumption record library, and performing electricity consumption prediction sensitivity learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model; Based on the first prediction time zone, predicting the power consumption according to the second power consumption prediction model to obtain the predicted power consumption of the first user; Obtaining a power transaction variety set, wherein the power transaction variety set includes M power transaction varieties, wherein M is a positive integer greater than 1; Based on the predicted power consumption of the first user, performing power auxiliary decision analysis according to the power trading variety set, obtaining a first recommended power trading decision, and sending the first recommended power trading decision to the first user; Based on the electricity consumption record library of the same group of users, high-precision learning is performed to build a first electricity consumption prediction model, including: The electricity consumption record library of the same group of users includes multiple electricity consumption records of users, and the multiple electricity consumption records of users have multiple electricity consumption time zone identifiers corresponding to each other; Filtering the multiple user electricity usage records based on the multiple user electricity usage time zone identifiers to obtain multiple associated user electricity usage records; configuring a plurality of high-precision learning operators based on the plurality of associated user electricity consumption records; Based on a fully connected neural network, cross-supervised training is performed on the electricity consumption records of the multiple associated users according to the multiple high-precision learning operators to generate the first electricity consumption prediction model; The method of performing power consumption prediction sensitivity learning on the first power consumption prediction model based on the first power consumption record library to obtain a second power consumption prediction model includes: Performing incremental learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a first incremental electricity consumption prediction model and a first electricity loss data set; Obtaining a first loss data amount of the first power loss data set; Determining whether the first loss data amount is greater than a preset loss data amount; If the first loss data amount is greater than the preset loss data amount, performing supervised training on the first incremental power consumption prediction model based on the first power consumption loss data set to generate the second power consumption prediction model; Configuring multiple high-precision learning operators based on the multiple associated user electricity consumption records includes: Based on the multiple associated user electricity consumption records, obtaining multiple record time zone association degrees; Calculate the correlation ratio based on the multiple recorded time zone correlations to obtain multiple record correlation ratio coefficients; Based on the multiple record correlation ratio coefficients, training computing power is allocated to the multiple associated user electricity consumption records to obtain the multiple high-precision learning operators.

2. The method according to claim 1, characterized in that The plurality of user electricity usage records are filtered based on the plurality of user electricity usage time zone identifiers, including: Traversing the multiple user electricity usage time zone identifiers, obtaining a first user electricity usage time zone identifier, and matching a first user electricity usage record corresponding to the first user electricity usage time zone identifier based on the multiple user electricity usage records; Based on the twin neural network, a correlation analysis is performed on the first user's electricity consumption time zone identifier and the first predicted time zone to obtain a first time zone correlation degree; Determining whether the first time zone correlation degree meets a preset time zone correlation degree; If the first time zone association degree meets the preset time zone association degree, the first user power usage record is added to the multiple associated user power usage records.

3. The method according to claim 1, characterized in that Based on the predicted power consumption of the first user, performing power auxiliary decision analysis according to the power transaction variety set to obtain a first recommended power transaction decision, including: Based on the predicted power consumption of the first user, respectively evaluate the fitness of the M power trading varieties to obtain the fitness of the M varieties; Based on the preset variety fitness, the M power trading varieties are screened according to the M variety fitness to obtain Q power trading varieties that meet the preset variety fitness, where Q is a positive integer and 1<Q≤M; Identify the fitness of the M varieties based on the Q electricity trading varieties to obtain the fitness of the Q varieties; The fitness ratio is calculated based on the fitness ratios of the Q varieties to obtain the fitness ratio coefficients of the Q varieties, and incentives are adjusted for the Q power trading varieties based on the fitness ratio coefficients of the Q varieties to generate the first recommended power trading decision.

4. The method according to claim 3, characterized in that Based on the predicted power consumption of the first user, respectively evaluating the fitness of the M power trading varieties to obtain the fitness of the M varieties includes: Based on the M power trading varieties, obtain the i-th power trading variety, and 1≤i≤M; Performing a matching evaluation based on the predicted power consumption of the first user and the i-th power trading variety to obtain a matching degree of the i-th variety; Identify the abundance of the i-th power trading product based on the i-th power trading product to obtain the abundance of the i-th power product; Based on the i-th power trading variety, the variety trading circulation degree is identified to obtain the i-th variety trading circulation degree; Based on the preset fitness evaluation constraints, the matching degree of the i-th variety, the abundance of the i-th variety and the transaction circulation degree of the i-th variety are weightedly calculated to generate the fitness of the i-th variety, and the i-th variety fitness is added to the fitness of the M varieties.

5. A high-precision learning power auxiliary decision-making system, characterized in that: A method for implementing a high-precision learning power-assisted decision-making method as described in any one of claims 1 to 4, the system comprising: A first user feature acquisition module, the first user feature acquisition module is used to collect basic information of the first user, obtain first user information, and perform feature recognition on the first user information to obtain first user features; A module for obtaining a power consumption record library of users in the same group, wherein the module is used to obtain users in the same group based on the first user characteristics, and to collect power consumption records based on the users in the same group to obtain a power consumption record library of users in the same group; A first electricity consumption prediction model acquisition module, the first electricity consumption prediction model acquisition module is used to perform high-precision learning based on the electricity consumption record library of the same group of users to build a first electricity consumption prediction model; a second electricity consumption prediction model acquisition module, the second electricity consumption prediction model acquisition module is used to collect the electricity consumption record of the first user, obtain a first electricity consumption record library, and perform electricity consumption prediction sensitivity learning on the first electricity consumption prediction model based on the first electricity consumption record library to obtain a second electricity consumption prediction model; A first user predicted power consumption obtaining module, the first user predicted power consumption obtaining module is used to perform power consumption prediction based on the first prediction time zone and the second power consumption prediction model to obtain the first user predicted power consumption; An electric power transaction variety set acquisition module, wherein the electric power transaction variety set acquisition module is used to obtain an electric power transaction variety set, wherein the electric power transaction variety set includes M electric power transaction varieties, wherein M is a positive integer greater than 1; a first recommended power trading decision obtaining module, the first recommended power trading decision obtaining module being used to perform power auxiliary decision analysis based on the predicted power consumption of the first user and the power trading variety set, obtain a first recommended power trading decision, and send the first recommended power trading decision to the first user; The system further comprises: A user electricity consumption record acquisition module, wherein the user electricity consumption record acquisition module is used for the electricity consumption record library of the same group of users to include multiple user electricity consumption records, and the multiple user electricity consumption records have multiple user electricity consumption time zone identifiers corresponding to each other; an associated user electricity consumption record obtaining module, the associated user electricity consumption record obtaining module being used to filter the multiple user electricity consumption records based on the multiple user electricity consumption time zone identifiers to obtain multiple associated user electricity consumption records; A high-precision learning operator processing module, the high-precision learning operator processing module is used to configure multiple high-precision learning operators based on the multiple associated user electricity consumption records; a first electricity consumption prediction model acquisition module, the first electricity consumption prediction model acquisition module being used to perform cross-supervisory training on the electricity consumption records of the multiple associated users according to the multiple high-precision learning operators based on a fully connected neural network to generate the first electricity consumption prediction model; The system further comprises: a first power loss data set acquisition module, the first power loss data set acquisition module being used to perform incremental learning on the first power consumption prediction model based on the first power consumption record library to obtain a first incremental power consumption prediction model and a first power loss data set; A first loss data amount obtaining module, the first loss data amount obtaining module is used to obtain a first loss data amount of the first power loss data set; A first loss data amount processing module, the first loss data amount processing module is used to determine whether the first loss data amount is greater than a preset loss data amount; a second electricity consumption prediction model acquisition module, the second electricity consumption prediction model acquisition module being used to perform supervised training on the first incremental electricity consumption prediction model based on the first electricity loss data set to generate the second electricity consumption prediction model if the first loss data amount is greater than the preset loss data amount; The system further comprises: A recording time zone association degree obtaining module, the recording time zone association degree obtaining module is used to obtain multiple recording time zone association degrees based on the multiple associated user power consumption records; A record correlation ratio coefficient obtaining module, the record correlation ratio coefficient obtaining module is used to calculate the correlation ratio based on the multiple record time zone correlations to obtain multiple record correlation ratio coefficients; A plurality of high-precision learning operator acquisition modules are used to allocate training computing power to the plurality of associated user electricity consumption records based on the plurality of record association ratio coefficients to obtain the plurality of high-precision learning operators.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images