Virtual power plant transaction risk assessment method and system based on assessment model
By using evaluation models of principal component analysis and neural network training in virtual power plant transaction risk assessment, combining sensitivity and non-sensitivity indicators, the problem of poor evaluation robustness caused by data loss is solved, and the full indicator evaluation of virtual power plant is achieved. risk assessment.
Patent Information
- Application Number
- CN202410827601.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-25
- Publication Date
- 2024-11-08
AI Technical Summary
The existing technology requires complete transaction data when evaluating virtual power plant transaction risk, which makes it difficult to achieve risk assessment in the scenario of missing data and poor evaluation robustness.
Using an evaluation model-based approach, the data is divided into sensitive data and non-sensitive data by monitoring transaction data between virtual power plants and multiple trading markets using principal component analysis, and the sensitivity indicator evaluation model and non-sensitive indicator evaluation are used to train sensitivity indicators. The model combines the two for combined evaluation to achieve full indicator evaluation of virtual power plants.
In the scenario of missing transaction data, the risk assessment can be effectively carried out, which improves the robustness of virtual power plant transaction risk assessment.
Smart Images

Figure CN118917940A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of virtual power plants, and in particular to a virtual power plant transaction risk assessment method and system based on an assessment model. Background Art
[0002] A virtual power plant is a power coordination management system that uses advanced information and communication technologies and software systems to achieve the aggregation and coordinated optimization of DERs such as DG, energy storage systems, controllable loads, and electric vehicles, so that it can participate in the power market and power grid operation as a special power plant. The core of the concept of a virtual power plant can be summarized as "communication" and "aggregation". The key technologies of a virtual power plant mainly include coordinated control technology, smart metering technology, and information and communication technology.
[0003] At present, in order to promote the enthusiasm of virtual power plant operators, increase the level of new energy consumption, and form a stable coordination and interaction capability between virtual power plants and power grids, virtual power plants are traded in a graded manner, which greatly optimizes the resource allocation between virtual power plants. Therefore, the transaction management of virtual power plants can ensure the stability of virtual power plant transactions.
[0004] When managing transactions of virtual power plants, existing technologies usually analyze transaction risks from transaction data and manage transactions by understanding transaction risks. However, transaction risk assessment must be based on complete transaction data, which makes it difficult to implement risk assessment in scenarios where transaction data is missing, and transaction risk assessment has poor robustness. Summary of the invention
[0005] The purpose of the present invention is to provide a virtual power plant transaction risk assessment method based on an assessment model to solve the technical problems in the prior art that transaction risk assessment must be based on complete transaction data, which makes it difficult to implement risk assessment in scenarios where transaction data is missing and the transaction risk assessment has poor robustness.
[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:
[0007] A virtual power plant transaction risk assessment method based on an assessment model comprises the following steps:
[0008] Monitor transaction data used for virtual power plants to settle with multiple trading markets as evaluation data for assessing virtual power plant transaction risks;
[0009] Using principal component analysis, the evaluation data is divided into evaluation sensitive data and non-evaluation sensitive data;
[0010] Using a neural network to train the evaluation sensitive data, a sensitivity index evaluation model is obtained to obtain a transaction risk value by evaluating the sensitive data;
[0011] Using non-evaluation sensitive data to perform transfer learning in the sensitivity indicator evaluation model, a non-sensitivity indicator evaluation model is obtained by evaluating the non-sensitive data to obtain the transaction risk value;
[0012] The transaction data is evaluated in combination through the sensitivity index evaluation model and the non-sensitivity index evaluation model, so as to achieve full index evaluation of the virtual power plant. (Significance: when the principal component is missing, non-sensitive data is used for alternative evaluation to improve the robustness of the evaluation)
[0013] As a preferred solution of the present invention, the trading market includes: an electric energy market, a natural gas market, a carbon trading market, and a peak-shaving market.
[0014] As a preferred solution of the present invention, the transaction data of the electric energy market includes the scale of aggregatable resources, the scale of total load, and the electricity clearing price of the electric energy market;
[0015] The transaction data of the peak-shaving market includes the peak-shaving declared quantity and electricity clearing price generated by the peak-shaving demand of the peak-shaving market;
[0016] The transaction data of the carbon trading market includes the carbon trading volume and carbon clearing price of the carbon trading market;
[0017] The transaction data of the natural gas market include the natural gas trading volume and natural gas clearing price in the natural gas market.
[0018] As a preferred solution of the present invention, the evaluation data includes the scale of aggregatable resources, total load scale, and electricity-clearing price in the electric energy market, the peak-shaving declaration volume and electricity-clearing price generated by the peak-shaving demand in the peak-shaving market, the carbon trading volume and carbon-clearing price in the carbon trading market, and the natural gas trading volume and natural gas-clearing price in the natural gas market.
[0019] As a preferred solution of the present invention, the principal component analysis method of the evaluation data includes:
[0020] The principal components of the evaluation data are obtained using the principal component analysis method;
[0021] The principal components of the evaluation data are taken as evaluation sensitive data, and the remaining data in the evaluation data are taken as non-evaluation sensitive data.
[0022] As a preferred embodiment of the present invention, the method for constructing the sensitivity index evaluation model includes:
[0023] Manually mark the virtual power plant transaction risk value corresponding to the sensitive data;
[0024] The assessed sensitive data is used as an input item of the neural network, and the virtual power plant transaction risk value corresponding to the assessed sensitive data is used as an output item of the neural network;
[0025] The neural network is used to train the input items and output items of the neural network, and a sensitivity index evaluation model for obtaining the risk value of the virtual power plant transaction based on the evaluation of sensitive data is obtained;
[0026] The sensitivity index evaluation model is:
[0027] Score A =NET(D A );
[0028] In the formula, Score A The transaction risk value output by the sensitivity index evaluation model, D A To evaluate sensitive data, NET is a neural network;
[0029] Loss function of the sensitivity index evaluation model: LA = || Score A -Score r || 1 , where LA is the loss function of the sensitivity assessment model, Score A Score is the transaction risk value output by the sensitivity index evaluation model. r is the true value of the transaction risk value, ||Score A -Score r || 1 is the L1 norm.
[0030] As a preferred embodiment of the present invention, the method for constructing the non-sensitive index evaluation model includes:
[0031] Manually mark the virtual power plant transaction risk value corresponding to non-assessed sensitive data;
[0032] The non-assessment sensitive data is used as the migration input item of the sensitivity indicator evaluation model, and the virtual power plant transaction risk value corresponding to the non-assessment sensitive data is used as the migration output item of the sensitivity indicator evaluation model;
[0033] The migration input items and migration output items are trained using the sensitivity indicator evaluation model to obtain a non-sensitive indicator evaluation model for obtaining the transaction risk value of the virtual power plant based on non-evaluation sensitive data;
[0034] The non-sensitive index evaluation model is:
[0035] Score B =NET(D B );
[0036] In the formula, Score B The transaction risk value output by the non-sensitive indicator evaluation model, D Bis non-evaluation sensitive data, NET is neural network;
[0037] Loss function of the non-sensitive indicator evaluation model: LB = -||D B -D A || 1 +||Score A -Score B || 1 , where LB is the loss function of the non-sensitive evaluation model, Score B The transaction risk value output by the non-sensitive indicator evaluation model, D B For non-evaluation sensitive data, Score A The transaction risk value output by the sensitivity index evaluation model, D A To evaluate sensitive data, ||D B -D A || 1 and||Score A -Score B || 1 All are L1 norm forms.
[0038] As a preferred solution of the present invention, the method for performing combined evaluation on transaction data by using the sensitivity index evaluation model and the non-sensitivity index evaluation model includes:
[0039] When the evaluation sensitive data in the transaction data has missing items, and the non-evaluation sensitive data is not missing, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the virtual power plant transaction risk value;
[0040] When the evaluation sensitive data in the transaction data is not missing, the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain the virtual power plant transaction risk value;
[0041] When both the evaluation sensitive data and the non-evaluation sensitive data in the transaction data have missing items, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity index evaluation model to obtain a first transaction risk value, and then the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain a second transaction risk value, and finally the first transaction risk value and the second transaction risk value are weighted averaged to obtain the virtual power plant transaction risk value, wherein,
[0042] The weighted weight of the first transaction risk value and the second transaction risk value is: W1 = N A-miss / N A , W1=N B-miss / N B, where W1 and W2 are the weighted weights of the first transaction risk value and the second transaction risk value, respectively, N A-miss N is the non-missing item in the transaction data for evaluating sensitive data. A is the total number of data items in transaction data that are evaluated for sensitive data, N B-miss is the non-missing item of non-evaluation sensitive data in the transaction data, N B It is the total data items of non-evaluation sensitive data in transaction data.
[0043] As a preferred solution of the present invention, the present invention provides a virtual power plant transaction risk assessment system based on an assessment model, which is applied to the virtual power plant transaction risk assessment method based on the assessment model, and the system includes:
[0044] A data collection unit for monitoring transaction data used for settlement between the virtual power plant and multiple trading markets;
[0045] The data processing unit is used to perform a combined evaluation of the transaction data through the sensitivity index evaluation model and the non-sensitivity index evaluation model to achieve a full index evaluation of the virtual power plant.
[0046] The data storage unit is used to store the sensitivity index evaluation model and the non-sensitivity index evaluation model.
[0047] As a preferred solution of the present invention, the method in which the data processing unit performs combined evaluation on the transaction data by using the sensitivity index evaluation model and the non-sensitivity index evaluation model comprises:
[0048] When the evaluation sensitive data in the transaction data has missing items, and the non-evaluation sensitive data is not missing, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the virtual power plant transaction risk value;
[0049] When the evaluation sensitive data in the transaction data is not missing, the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain the virtual power plant transaction risk value;
[0050] When both the evaluation sensitive data and the non-evaluation sensitive data in the transaction data have missing items, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity index evaluation model to obtain a first transaction risk value, and then the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain a second transaction risk value, and finally the first transaction risk value and the second transaction risk value are weighted averaged to obtain the virtual power plant transaction risk value, wherein,
[0051] The weighted weight of the first transaction risk value and the second transaction risk value is: W1 = N A-miss / N A, W1=N B-miss / N B , where W1 and W2 are the weighted weights of the first transaction risk value and the second transaction risk value, respectively, N A-miss N is the non-missing item in the transaction data for evaluating sensitive data. A is the total number of data items in transaction data that are evaluated for sensitive data, N B-miss is the non-missing item of non-evaluation sensitive data in the transaction data, N B It is the total data items of non-evaluation sensitive data in transaction data.
[0052] Compared with the prior art, the present invention has the following beneficial effects:
[0053] The present invention utilizes a neural network to train the evaluation sensitive data to obtain a sensitivity index evaluation model for obtaining a transaction risk value by evaluating the sensitive data, and utilizes non-evaluation sensitive data for transfer learning to obtain a non-sensitive index evaluation model for obtaining a transaction risk value by evaluating the non-sensitive data, thereby realizing a full index evaluation of the virtual power plant, and realizing risk evaluation even in a scenario where transaction data is missing, thereby enhancing the robustness of the transaction risk evaluation of the virtual power plant. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.
[0055] Figure 1 A virtual power plant transaction risk assessment flow chart provided by an embodiment of the present invention;
[0056] Figure 2 A block diagram of a virtual power plant transaction risk assessment system provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0058] like Figure 1 As shown, the present invention provides a virtual power plant transaction risk assessment method based on an assessment model, comprising the following steps:
[0059] Monitor transaction data used for virtual power plants to settle with multiple trading markets as evaluation data for assessing virtual power plant transaction risks;
[0060] The evaluation data is divided into evaluation sensitive data and non-evaluation sensitive data using principal component analysis;
[0061] A neural network is used to train sensitive data and obtain a sensitivity index evaluation model that obtains transaction risk value by evaluating sensitive data;
[0062] In the sensitivity index evaluation model, non-evaluation sensitive data is used for transfer learning to obtain a non-sensitive index evaluation model that obtains transaction risk value by evaluating non-sensitive data;
[0063] Through the sensitivity index evaluation model and the non-sensitivity index evaluation model, the transaction data is evaluated in combination to achieve full index evaluation of the virtual power plant. (Significance: when the main component is missing, non-sensitive data is used for alternative evaluation to improve the robustness of the evaluation)
[0064] In order to improve the robustness of the model used for virtual power plant transaction risk assessment, the present invention uses the principal components of the assessment data to train the assessment model. The principal components of the assessment data are also called characteristic data in the assessment data that are sensitive to the transaction risk value (i.e., non-assessment sensitive data), and a model that can complete transaction risk assessment based on the principal components of the assessment data is obtained (i.e., a sensitivity indicator assessment model).
[0065] Since the principal components of the evaluation data are characteristic data in the evaluation data that are sensitive to the transaction risk value, it is easy to obtain the transaction risk value through data analysis of the principal components of the evaluation data. Therefore, the sensitivity indicator evaluation model can easily obtain high accuracy, that is, a highly accurate transaction risk assessment result can be obtained by evaluating the principal components of the data.
[0066] The non-principal components of the evaluation data, or the characteristic data in the evaluation data that is not sensitive to the transaction risk value (i.e., non-evaluation sensitive data), refer to the difficulty in obtaining the transaction risk value through data analysis of the non-principal components of the evaluation data.
[0067] Therefore, in order to allow the characteristic data in the evaluation data that is insensitive to the transaction risk value (non-evaluation sensitive data), or the non-principal components of the evaluation data, to inherit the accuracy of the transaction risk assessment results obtained by the principal components of the evaluation data, the present invention uses a transfer learning method to train the non-evaluation sensitive data on the sensitivity index evaluation model to obtain a non-sensitivity index evaluation model, so that a highly accurate transaction risk assessment result can be obtained through the non-principal components of the evaluation data, that is, the highly accurate transaction risk value obtained by the principal components of the evaluation data can be obtained through the non-principal components of the evaluation data.
[0068] Therefore, after the principal component of the evaluation data is missing, the same accurate transaction risk value can be obtained by using characteristic data that is not sensitive to the transaction risk value (ie, non-evaluation sensitive data).
[0069] Moreover, further, when both the principal component and the non-principal component are missing in the evaluation data, the present invention will calculate the missing principal component and the non-principal component through the corresponding evaluation model, and perform weighted combination on the calculation results of the two evaluation models. The more missing items there are, the lower the credibility of the output result of the corresponding evaluation model, that is, the smaller the weight of the calculation result. Thus, in the independent calculation of the two evaluation models, a calculation constraint / evidence is formed to avoid the occurrence of random results, and a relatively accurate transaction risk value is obtained when both the principal component and the non-principal component are missing. Although it is not completely accurate, it can be used as reference data in risk assessment to roughly grasp the transaction risk.
[0070] In summary, the present invention can realize full-index evaluation of virtual power plants and improve the robustness of the model.
[0071] Trading markets include: electricity market, natural gas market, carbon trading market, and peak-shaving market.
[0072] The transaction data of the electric energy market includes the scale of aggregated resources, total load scale, and electricity clearing price in the electric energy market;
[0073] The transaction data of the peak-shaving market includes the peak-shaving declared quantity and electricity clearing price generated by the peak-shaving demand of the peak-shaving market;
[0074] The transaction data of the carbon trading market includes the carbon trading volume and carbon clearing price of the carbon trading market;
[0075] The transaction data of the natural gas market include the natural gas trading volume and natural gas clearing price in the natural gas market.
[0076] The evaluation data include the scale of aggregable resources, total load scale, and electricity-clearing price in the electric energy market, the peak-shaving declaration volume and electricity-clearing price generated by the peak-shaving demand in the peak-shaving market, the carbon trading volume and carbon-clearing price in the carbon trading market, and the natural gas trading volume and natural gas-clearing price in the natural gas market.
[0077] Principal component analysis methods for evaluating data include:
[0078] The principal component analysis method is used to obtain the principal components of the evaluation data. The principal component analysis method obtains the contribution of each data item and selects the principal components according to the contribution;
[0079] The principal components of the evaluation data are taken as evaluation sensitive data, and the remaining data in the evaluation data are taken as non-evaluation sensitive data.
[0080] In order to improve the robustness of the model used for virtual power plant transaction risk assessment, the present invention uses the principal components of the assessment data to train the assessment model. The principal components of the assessment data are also called feature data in the assessment data that are sensitive to the transaction risk value (i.e., non-assessment sensitive data). A model that can complete the transaction risk assessment based on the principal components of the assessment data (i.e., a sensitivity index assessment model) is obtained, which is as follows:
[0081] The construction method of the sensitivity index evaluation model includes:
[0082] Manually mark the virtual power plant transaction risk value corresponding to the sensitive data;
[0083] The assessed sensitive data is used as an input item of the neural network, and the virtual power plant transaction risk value corresponding to the assessed sensitive data is used as an output item of the neural network;
[0084] The neural network is used to train the input items and output items of the neural network, and a sensitivity index evaluation model for obtaining the risk value of the virtual power plant transaction based on the evaluation of sensitive data is obtained;
[0085] The sensitivity index evaluation model is:
[0086] Score A =NET(D A );
[0087] In the formula, Score A The transaction risk value output by the sensitivity index evaluation model, D A To evaluate sensitive data, NET is a neural network;
[0088] Loss function of the sensitivity index evaluation model: LA = || Score A -Score r || 1 , where LA is the loss function of the sensitivity assessment model, Score A Score is the transaction risk value output by the sensitivity index evaluation model. r is the true value of the transaction risk value, ||Score A -Score r || 1 is the L1 norm.
[0089] In order to make the feature data (non-assessment sensitive data) in the assessment data that is insensitive to the transaction risk value, or the non-principal component of the assessment data, inherit the accuracy of the transaction risk assessment result obtained by the principal component of the assessment data, the present invention uses a transfer learning method to train the non-assessment sensitive data on the sensitivity index evaluation model to obtain a non-sensitivity index evaluation model, so that a highly accurate transaction risk assessment result can be obtained through the non-principal component of the assessment data, that is, a highly accurate transaction risk value obtained by the principal component of the assessment data can be obtained through the non-principal component of the assessment data, as follows:
[0090] The construction method of the non-sensitivity indicator evaluation model includes:
[0091] Manually mark the virtual power plant transaction risk value corresponding to non-assessed sensitive data;
[0092] The non-assessment sensitive data is used as the migration input item of the sensitivity indicator evaluation model, and the virtual power plant transaction risk value corresponding to the non-assessment sensitive data is used as the migration output item of the sensitivity indicator evaluation model;
[0093] The migration input items and migration output items are trained using the sensitivity indicator evaluation model to obtain a non-sensitive indicator evaluation model for obtaining the transaction risk value of the virtual power plant based on non-evaluation sensitive data;
[0094] The non-sensitive indicator evaluation model is:
[0095] Score B =NET(D B );
[0096] In the formula, Score B The transaction risk value output by the non-sensitive indicator evaluation model, D B is non-evaluation sensitive data, NET is neural network;
[0097] Loss function of the non-sensitive indicator evaluation model: LB = -||D B -D A || 1 +||Score A -Score B || 1 , where LB is the loss function of the non-sensitive evaluation model, Score B The transaction risk value output by the non-sensitive indicator evaluation model, D B For non-evaluation sensitive data, Score A The transaction risk value output by the sensitivity index evaluation model, D A To evaluate sensitive data, ||D B -D A || 1and||Score A -Score B || 1 All are L1 norm forms.
[0098] Methods for combining and evaluating transaction data using sensitivity index evaluation models and non-sensitivity index evaluation models include:
[0099] When the evaluation sensitive data in the transaction data has missing items, but the non-evaluation sensitive data is not missing, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the virtual power plant transaction risk value;
[0100] When the assessed sensitive data in the transaction data is not missing, the assessed sensitive data in the transaction data is measured through the sensitivity index assessment model to obtain the virtual power plant transaction risk value;
[0101] When both the evaluation sensitive data and the non-evaluation sensitive data in the transaction data have missing items, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the first transaction risk value, and then the evaluation sensitive data in the transaction data is measured by the sensitivity indicator evaluation model to obtain the second transaction risk value. Finally, the first transaction risk value and the second transaction risk value are weighted averaged to obtain the virtual power plant transaction risk value, where:
[0102] The weighted weight of the first transaction risk value and the second transaction risk value is: W1 = N A-miss / N A , W1=N B-miss / N B , where W1 and W2 are the weighted weights of the first transaction risk value and the second transaction risk value, respectively, N A-miss N is the non-missing item in the transaction data for evaluating sensitive data. A is the total number of data items in transaction data that are evaluated for sensitive data, N B-miss is the non-missing item of non-evaluation sensitive data in the transaction data, N B It is the total data items of non-evaluation sensitive data in transaction data.
[0103] When the principal component of the evaluation data is missing, the same accurate transaction risk value can be obtained by using characteristic data that is not sensitive to the transaction risk value (ie, non-evaluation sensitive data).
[0104] When both the principal component and the non-principal component are missing in the evaluation data, the present invention calculates the missing principal component and the non-principal component through the corresponding evaluation model, and performs weighted combination on the calculation results of the two evaluation models. The more missing items there are, the lower the credibility of the output result of the corresponding evaluation model, that is, the smaller the weight of the calculation result. Thus, in the independent calculation of the two evaluation models, a calculation constraint / evidence is formed to avoid the occurrence of random results, and a relatively accurate transaction risk value is obtained when both the principal component and the non-principal component are missing. Although it is not completely accurate, it can be used as reference data in risk assessment to roughly grasp the transaction risk.
[0105] like Figure 2 As shown, the present invention provides a virtual power plant transaction risk assessment system based on an assessment model, which is applied to a virtual power plant transaction risk assessment method based on an assessment model, and the system includes:
[0106] A data collection unit for monitoring transaction data used for settlement between the virtual power plant and multiple trading markets;
[0107] The data processing unit is used to perform combined evaluation of transaction data through a sensitivity index evaluation model and a non-sensitivity index evaluation model to achieve full index evaluation of the virtual power plant.
[0108] The data storage unit is used to store the sensitivity index evaluation model and the non-sensitivity index evaluation model.
[0109] The method for the data processing unit to perform combined evaluation on transaction data by using the sensitivity index evaluation model and the non-sensitivity index evaluation model includes:
[0110] When the evaluation sensitive data in the transaction data has missing items, but the non-evaluation sensitive data is not missing, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the virtual power plant transaction risk value;
[0111] When the assessed sensitive data in the transaction data is not missing, the assessed sensitive data in the transaction data is measured through the sensitivity index assessment model to obtain the virtual power plant transaction risk value;
[0112] When both the evaluation sensitive data and the non-evaluation sensitive data in the transaction data have missing items, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the first transaction risk value, and then the evaluation sensitive data in the transaction data is measured by the sensitivity indicator evaluation model to obtain the second transaction risk value. Finally, the first transaction risk value and the second transaction risk value are weighted averaged to obtain the virtual power plant transaction risk value, where:
[0113] The weighted weight of the first transaction risk value and the second transaction risk value is: W1 = NA-miss / N A , W1=N B-miss / N B , where W1 and W2 are the weighted weights of the first transaction risk value and the second transaction risk value, respectively, N A-miss N is the non-missing item in the transaction data for evaluating sensitive data. A is the total number of data items in transaction data that are evaluated for sensitive data, N B-miss is the non-missing item of non-evaluation sensitive data in the transaction data, N B It is the total data items of non-evaluation sensitive data in transaction data.
[0114] The present invention utilizes a neural network to train evaluation sensitive data to obtain a sensitivity index evaluation model that obtains transaction risk values by evaluating sensitive data, and utilizes non-evaluation sensitive data for transfer learning to obtain a non-sensitivity index evaluation model that obtains transaction risk values by evaluating non-sensitive data, thereby achieving full index evaluation of virtual power plants, and realizing risk evaluation even in scenarios where transaction data is missing, thereby enhancing the robustness of virtual power plant transaction risk evaluation.
[0115] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.
Claims
1. A virtual power plant transaction risk assessment method based on an assessment model, characterized in that: The following steps are involved: Monitor transaction data used for virtual power plants to settle with multiple trading markets as evaluation data for assessing virtual power plant transaction risks; Using principal component analysis, the evaluation data is divided into evaluation sensitive data and non-evaluation sensitive data; Using a neural network to train the evaluation sensitive data, a sensitivity index evaluation model is obtained to obtain a transaction risk value by evaluating the sensitive data; Using non-evaluation sensitive data to perform transfer learning in the sensitivity indicator evaluation model, a non-sensitivity indicator evaluation model is obtained by evaluating the non-sensitive data to obtain the transaction risk value; The transaction data is evaluated in combination through the sensitivity index evaluation model and the non-sensitivity index evaluation model, so as to realize the full index evaluation of the virtual power plant.
2. According to claim 1, a virtual power plant transaction risk assessment method based on an assessment model is characterized by: The trading markets include: electricity market, natural gas market, carbon trading market, and peak-shaving market.
3. The virtual power plant transaction risk assessment method based on the assessment model according to claim 2 is characterized by: The transaction data of the electric energy market includes the scale of aggregated resources, total load scale, and electricity clearing price in the electric energy market; The transaction data of the peak-shaving market includes the peak-shaving declared quantity and electricity clearing price generated by the peak-shaving demand of the peak-shaving market; The transaction data of the carbon trading market includes the carbon trading volume and carbon clearing price of the carbon trading market; The transaction data of the natural gas market include the natural gas trading volume and natural gas clearing price in the natural gas market.
4. The virtual power plant transaction risk assessment method based on the assessment model according to claim 1 is characterized by: The evaluation data include the scale of aggregable resources, total load scale, and electricity-clearing price in the electric energy market, the peak-shaving declaration volume and electricity-clearing price generated by the peak-shaving demand in the peak-shaving market, the carbon trading volume and carbon-clearing price in the carbon trading market, and the natural gas trading volume and natural gas-clearing price in the natural gas market.
5. According to claim 4, a virtual power plant transaction risk assessment method based on an assessment model is characterized in that: The principal component analysis method of the evaluation data includes: The principal components of the evaluation data are obtained using the principal component analysis method; The principal components of the evaluation data are taken as evaluation sensitive data, and the remaining data in the evaluation data are taken as non-evaluation sensitive data.
6. A virtual power plant transaction risk assessment method based on an assessment model according to claim 5, characterized in that: The method for constructing the sensitivity index evaluation model includes: Manually mark the virtual power plant transaction risk value corresponding to the sensitive data; The assessed sensitive data is used as an input item of the neural network, and the virtual power plant transaction risk value corresponding to the assessed sensitive data is used as an output item of the neural network; The neural network is used to train the input items and output items of the neural network, and a sensitivity index evaluation model for obtaining the risk value of the virtual power plant transaction based on the evaluation of sensitive data is obtained; The sensitivity index evaluation model is: Score A =NET(D A ); In the formula, Score A The transaction risk value output by the sensitivity index evaluation model, D A To evaluate sensitive data, NET is a neural network; Loss function of the sensitivity index evaluation model: LA = || Score A -Score r ||1, where LA is the loss function of the sensitivity assessment model, Score A Score is the transaction risk value output by the sensitivity index evaluation model. r is the true value of the transaction risk value, ||Score A -Score r ||1 is the L1 norm.
7. The virtual power plant transaction risk assessment method based on the assessment model according to claim 6 is characterized by: The construction method of the non-sensitivity indicator evaluation model includes: Manually mark the virtual power plant transaction risk value corresponding to non-assessed sensitive data; The non-assessment sensitive data is used as the migration input item of the sensitivity indicator evaluation model, and the virtual power plant transaction risk value corresponding to the non-assessment sensitive data is used as the migration output item of the sensitivity indicator evaluation model; The migration input items and migration output items are trained using the sensitivity indicator evaluation model to obtain a non-sensitive indicator evaluation model for obtaining the transaction risk value of the virtual power plant based on non-evaluation sensitive data; The non-sensitive index evaluation model is: Score B =NET(D B ); In the formula, Score B The transaction risk value output by the non-sensitive indicator evaluation model, D B is non-evaluation sensitive data, NET is neural network; Loss function of the non-sensitive indicator evaluation model: LB = -||D B -D A ||1+||Score A -Score B ||1, where LB is the loss function of the non-sensitive evaluation model, Score B The transaction risk value output by the non-sensitive indicator evaluation model, D B For non-evaluation sensitive data, Score A The transaction risk value output by the sensitivity index evaluation model, D A To evaluate sensitive data, ||D B -D A ||1 and ||Score A -Score B ||1 are all L1 norm forms.
8. The virtual power plant transaction risk assessment method based on the assessment model according to claim 7 is characterized by: The method for performing combined evaluation on transaction data by using the sensitivity index evaluation model and the non-sensitivity index evaluation model includes: When the evaluation sensitive data in the transaction data has missing items, and the non-evaluation sensitive data is not missing, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the virtual power plant transaction risk value; When the evaluation sensitive data in the transaction data is not missing, the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain the virtual power plant transaction risk value; When both the evaluation sensitive data and the non-evaluation sensitive data in the transaction data have missing items, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity index evaluation model to obtain a first transaction risk value, and then the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain a second transaction risk value, and finally the first transaction risk value and the second transaction risk value are weighted averaged to obtain the virtual power plant transaction risk value, wherein, The weighted weight of the first transaction risk value and the second transaction risk value is: W1 = N A-miss / N A , W1=N B-miss / N B , where W1 and W2 are the weighted weights of the first transaction risk value and the second transaction risk value, respectively, N A-miss N is the non-missing item in the transaction data for evaluating sensitive data. A is the total number of data items in transaction data that are evaluated for sensitive data, N B-miss is the non-missing item of non-evaluation sensitive data in the transaction data, N B It is the total data items of non-evaluation sensitive data in transaction data.
9. A virtual power plant transaction risk assessment system based on an assessment model, characterized in that: A virtual power plant transaction risk assessment method based on an assessment model applied to any one of claims 1 to 8, the system comprising: A data collection unit for monitoring transaction data used for settlement between the virtual power plant and multiple trading markets; The data processing unit is used to perform a combined evaluation of the transaction data through the sensitivity index evaluation model and the non-sensitivity index evaluation model to achieve a full index evaluation of the virtual power plant. The data storage unit is used to store the sensitivity index evaluation model and the non-sensitivity index evaluation model.
10. A virtual power plant transaction risk assessment system based on an assessment model according to claim 9, characterized in that: The method for the data processing unit to perform combined evaluation on the transaction data by using the sensitivity index evaluation model and the non-sensitivity index evaluation model includes: When the evaluation sensitive data in the transaction data has missing items, and the non-evaluation sensitive data is not missing, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity indicator evaluation model to obtain the virtual power plant transaction risk value; When the evaluation sensitive data in the transaction data is not missing, the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain the virtual power plant transaction risk value; When both the evaluation sensitive data and the non-evaluation sensitive data in the transaction data have missing items, the non-evaluation sensitive data in the transaction data is measured by the non-sensitivity index evaluation model to obtain a first transaction risk value, and then the evaluation sensitive data in the transaction data is measured by the sensitivity index evaluation model to obtain a second transaction risk value, and finally the first transaction risk value and the second transaction risk value are weighted averaged to obtain the virtual power plant transaction risk value, wherein, The weighted weight of the first transaction risk value and the second transaction risk value is: W1 = N A-miss / N A , W1=N B-miss / N B , where W1 and W2 are the weighted weights of the first transaction risk value and the second transaction risk value, respectively, N A-miss N is the non-missing item in the transaction data for evaluating sensitive data. A is the total number of data items in transaction data that are evaluated for sensitive data, N B-miss is the non-missing item of non-evaluation sensitive data in the transaction data, N B It is the total data items of non-evaluation sensitive data in transaction data.
