A power transaction market dispatching method

Abstract

The application provides a power transaction market scheduling method, and belongs to the technical field of power transaction market scheduling. The method comprises the following steps: step S1, establishing a DCGAN network structure and initializing network parameters; step S2, simulating and establishing a power transaction-analysis hybrid database by using an HTAP database; step S3, simulating and designing a power transaction database on the transaction side of the HTAP database; step S4, establishing a monitoring sample statistical matrix set of the DCGAN network; step S5, training a discriminator model; step S6, setting a monitoring sample statistical matrix set update frequency parameter. Step S7, statistical matrix; the method has good natural logical fitting and interpretability with the real relationship between the production side and the consumption side in the power transaction market. Both theory and experiment prove that the method is a better power transaction market scheduling model method.

Description

Technical Field

[0001] This invention relates to the field of power trading market dispatching technology, and more specifically, to a power trading market dispatching method. Background Technology

[0002] Currently, due to the inability to effectively store large quantities of electricity over long periods, the most significant challenge in this field is achieving a real-time dispatch model that precisely matches electricity production and demand. This model must ensure uninterrupted power supply for consumers while simultaneously guaranteeing that all power generation output is consumed in real time. Supply-demand mismatch can lead to either insufficient or redundant power supply. Traditional methods, such as using empirical models to pre-set production and consumption parameters, predicting electricity consumption based on historical consumption data through statistical analysis, and providing pre-sale quotas to secondary suppliers, have failed to adequately address the real-time supply-demand matching problem and lack interpretability.

[0003] How to invent a power trading market dispatch method to improve these problems has become an urgent issue for those skilled in the art. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a power trading market dispatching method, which aims to improve the traditional power trading market dispatching method's inability to effectively solve the problem of real-time supply and demand matching and its unsatisfactory interpretability.

[0005] This invention is implemented as follows: a method for dispatching electricity trading markets, comprising the following steps:

[0006] Step S1: Establish the DCGAN network structure and initialize the network parameters;

[0007] Step S2: Using the HTAP database, simulate and establish a hybrid database for power trading and analysis.

[0008] Step S3: Simulate and design an electricity trading database on the trading side of the HTAP database. The main database fields consist of trading time, trading region, trading volume, and trading price.

[0009] Step S4: On the analysis side of the HTAP database, based on the real-time synchronization table of the power trading data table on the trading side, and using scripts to simulate the power trading market's sales and consumption transactions, establish a set of monitoring sample statistical matrices for the DCGAN network;

[0010] Step S5: Using the statistical matrix set of monitoring samples generated above as the "real samples" (standard is 1), mix the "fake samples" (labeled as 0) generated by the initial parameters of the Generator, and train the Discriminator model;

[0011] Step S6: Set the update frequency parameter u of the monitoring sample statistical matrix set, which represents the frequency of updating and iterating the sample matrix using real-time consistent synchronization of power trading data on the HTAP database trading side;

[0012] Step S7: After the iteration training of steps S5-S6 is completed, the DCGAN model is obtained. The statistical matrix of the simulated monitoring samples generated by the Generator, that is, the electricity available for sale in a certain region for a certain period of time, can be used as the standard for electricity consumption forecast data to drive the scheduling strategy of the generation side and the consumption side of electricity trading, thereby forming a new type of electricity trading market scheduling model.

[0013] In a preferred embodiment of the present invention, in step S1, the DCGAN network structure is formed by two convolutional neural networks acting as a generator and a discriminator respectively. After establishing the DCGAN network structure, the neural network parameters are initialized, and the two convolutional neural networks are each selected with 3-5 layers to accelerate the training speed.

[0014] In a preferred embodiment of the present invention, in step S2, the transaction database side is used to support real-time power trading, and the analysis database side is used to support the generation of monitoring sample matrix datasets for DCGAN neural network learning.

[0015] In a preferred embodiment of the present invention, in step S3, the transaction area is defined by district or county.

[0016] In a preferred embodiment of the present invention, in step S4, the monitoring sample statistical matrix set of the DCGAN network is established as: electricity sales and consumption time interval - electricity sales and consumption area - electricity sales and consumption statistical value matrix.

[0017] In a preferred embodiment of the present invention, in step S6, the frequency parameter u is set according to the statistical side matrix generation efficiency of the HTAP database.

[0018] In a preferred embodiment of the present invention, the iteration termination condition for steps S5-S6 is a set number of iteration cycles, which is then terminated.

[0019] In a preferred embodiment of the present invention, the iteration termination condition for steps S5-S6 is to set a minimum error tolerance, and to use the logical expected state of the Generator and Discriminator as the termination condition for updating the DCGAN model with real-time updated power trading data.

[0020] The beneficial effects of this invention are as follows: This invention provides a power trading market scheduling method based on the above design. This method uses a Generative Adversarial Network (GAN) framework, specifically a Deep Convolutional Generative Adversarial Network (DCGAN) model structure, and combines the consistency characteristics of real-time data between the HTAP database trading and analysis ends to realize a power trading market scheduling model design. The scheduling model designed in this patent can utilize power trading data continuously synchronized to the analysis end in real time. After rapid statistical analysis, the training data matrix of this design model is continuously obtained, which then drives the Generator sub-model used to simulate the generation side, connected to the Discriminator sub-model used to simulate the consumption side. This continuously iteratively updates the DCGAN model parameters, ultimately obtaining a model that accurately meets the scheduling needs of the power trading market. Firstly, the design logic of Generative Adversarial Models (GANs) shares a natural similarity with the logical relationship between electricity production and consumption. It dynamically adjusts to achieve a balance through adversarial interaction, thus possessing good applicability and innovation in theory, while also exhibiting superior interpretability compared to existing methods. Secondly, using the HTAP database as the infrastructure for model learning and iteration ensures consistent synchronization of real-time electricity trading data, and then rapidly updates the training data matrix of the GAN through a high-performance analysis engine, guaranteeing the accuracy of the trained model from another perspective. In summary, the proposed method offers significant advantages for solving such real-world problems. The generative adversarial process of this design model has a strong natural logical fit and interpretability with the real-world relationship between the production and consumption sides in the electricity trading market. Both theoretical and experimental results demonstrate that it is a superior method for electricity trading market scheduling. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart provided by an embodiment of the present invention;

[0023] Figure 2 This is a schematic diagram of the DCGAN network structure and the statistical matrix structure of the monitoring samples provided in the embodiments of the present invention;

[0024] Figure 3 A schematic diagram illustrating the process of generating a statistical matrix for HTAP database power transaction synchronization and monitoring samples, provided for an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0026] Please see Figure 1 This invention provides a technical solution: a method for dispatching electricity trading markets, comprising the following steps:

[0027] Step S1: Establish a DCGAN network structure consisting of two convolutional neural networks that act as the generator and discriminator, respectively. You can choose 3-5 layers of convolutional neural networks to speed up the training process and initialize the neural network parameters.

[0028] Step S2: Using the HTAP database, simulate the establishment of a hybrid power trading-analysis database. The trading database side is used to support real-time power trading, and the analysis database side is used to support the generation of monitoring sample matrix datasets for DCGAN neural network learning.

[0029] Step S3: Simulate and design the power trading database on the trading side of the HTAP database. The main database fields consist of trading time, trading region, trading volume, and trading price. The trading region is in units of districts and counties.

[0030] Step S4: On the analysis side of the HTAP database, based on the real-time synchronization table of the power transaction data table on the trading side, and using a script to simulate the power transaction market, establish a set of monitoring sample statistical matrices for the DCGAN network: namely, a matrix of power transaction time intervals, power transaction regions, and power transaction statistics. Specifically, the "power transaction time interval" can be statistically analyzed in hours, and the "power transaction region" can be statistically analyzed in districts and counties of a province / city. One matrix can be created per day, and the end point of the day is "Now()".

[0031] Step S5: Using the statistical matrix set of monitoring samples generated above as the "real samples" (standard is 1), mix the "fake samples" (labeled as 0) generated by the initial parameters of the Generator, and train the Discriminator model;

[0032] Step S6: Set the update frequency parameter u of the monitoring sample statistical matrix set, which represents the frequency of updating and iterating the sample matrix using real-time consistent power trading data synchronized with the HTAP database trading side. If set to 1, it means that the parameters of the Discriminator model need to be generated using the latest trading data statistics in each iteration; if set to 10, it means that it is updated once every 10 iterations. Specifically, the setting of the frequency parameter u can be adjusted according to the matrix generation efficiency of the HTAP database statistical side. Generally speaking, the higher the update frequency, the higher the model accuracy.

[0033] In some specific implementation schemes, the iteration termination conditions for steps 5-7 can be adopted in the following two ways, depending on the efficiency of the computing environment: (1) Set the number of iteration cycles (e.g., 500 times), and terminate when the time is up; (2) Or set the minimum error tolerance (e.g., 0.01), and use the logical state that the Generator and Discriminator should reach as the termination condition for updating the DCGAN model with real-time updated power trading data.

[0034] Step S7: After the iterative training of Steps 5-6 above is completed, the DCGAN model is obtained. The statistical matrix of the simulated monitoring samples generated by the Generator, that is, the electricity available for sale in a certain region for a certain period of time, can be used as the standard for electricity consumption forecast data to drive the scheduling strategy of the power generation side and the power consumption side of the power trading, thereby forming a new type of power trading market scheduling model.

[0035] This invention first utilizes the idea of ​​obtaining a high-quality realistic model simulation through continuous generation-adversarial iteration of the generator and discriminator in a Generative Adversarial Network (GAN). The generation side and the consumption side are respectively mapped to the generator and discriminator in the GAN, resulting in a logical model design that simulates the generation-consumption relationship with extremely high similarity. Furthermore, combining the consistency between real-time data from the HTAP database transaction and analysis terminals, real-time electricity consumption transaction data is continuously synchronized to the analysis terminal for rapid statistical analysis, continuously generating a training data matrix for the generator / discriminator model on the generation side. The generation-adversarial iterative process of the generation-consumption model is based as much as possible on the statistical update results of historical data from real-time transactions, ultimately obtaining a more accurate matching power production and consumption scheduling model and method.

[0036] Experiments have shown that the design method and model of this patent can effectively match the scheduling needs of power production and consumption. At the same time, because the design concept is highly consistent with the natural state of power production and consumption, it has good interpretability and significant theoretical value for continued research and practical application.

[0037] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for dispatching electricity in a trading market, characterized in that, Includes the following steps: Step S1: Establish the DCGAN network structure and initialize the network parameters; Step S2: Using an HTAP database, simulate and establish a hybrid database for electricity trading and analysis; Step S3: Simulate and design an electricity trading database on the trading side of the HTAP database. The main database fields consist of trading time, trading region, trading volume, and trading price. Step S4: On the analysis side of the HTAP database, based on the real-time synchronization table of the power trading data table on the trading side, and using scripts to simulate the power trading market's sales and consumption transactions, establish a set of monitoring sample statistical matrices for the DCGAN network; Step S5: Using the statistical matrix set of monitoring samples generated above as the "real samples", mix the "fake samples" generated by the initial parameters of the Generator to train the Discriminator model; Step S6: Set the update frequency parameter u of the monitoring sample statistical matrix set, which represents the frequency of updating and iterating the sample matrix using real-time consistent synchronization of power trading data on the HTAP database trading side; Step S7: After the iteration training of steps S5-S6 is completed, the DCGAN model is obtained. The statistical matrix of the simulated monitoring samples generated by the Generator, that is, the electricity available for sale in a certain region for a certain period of time, can be used as the standard for electricity consumption forecast data to drive the scheduling strategy of the generation side and the consumption side of electricity trading, thereby forming an electricity trading market scheduling model.

2. The power trading market dispatching method according to claim 1, characterized in that, In step S1, the DCGAN network structure is formed by two convolutional neural networks acting as a generator and a discriminator, respectively. After establishing the DCGAN network structure, the neural network parameters are initialized. The two convolutional neural networks each have 3-5 layers to accelerate the training speed.

3. The power trading market dispatching method according to claim 1, characterized in that, In step S2, the transaction database side is used to support real-time power trading, and the analysis database side is used to support the generation of the monitoring sample matrix dataset for DCGAN neural network learning.

4. The power trading market dispatching method according to claim 1, characterized in that, In step S3, the transaction area is divided into districts and counties.

5. The power trading market dispatching method according to claim 1, characterized in that, In step S4, the monitoring sample statistical matrix set of the DCGAN network is established as: electricity sales and consumption time interval - electricity sales and consumption area - electricity sales and consumption statistical value matrix.

6. The power trading market dispatching method according to claim 1, characterized in that, In step S6, the frequency parameter u is set according to the statistical side matrix generation efficiency of the HTAP database.

7. The power trading market dispatching method according to claim 1, characterized in that, The iteration termination condition for steps S5-S6 is the set number of iterations, at which point the iteration will terminate.

8. The power trading market dispatching method according to claim 1, characterized in that, The iteration termination condition for steps S5-S6 is to set a minimum error tolerance, and use the logical expected state of the Generator and Discriminator as the termination condition for updating the DCGAN model with real-time updated power trading data.

Images