Power technology development trend prediction method and system based on data correction

Abstract

The invention provides an electric power technology development trend prediction method and system based on data correction. The method comprises the following steps: constructing a PP I index calculation model and a CP I index calculation model; obtaining power time sequence data in the target area; according to the PP I index calculation model and the CP I index calculation model, constructing a relevance model of the unit added value power consumption with the original PP I index and the original CP I index, performing linear decomposition on the PP I index and the CP I index, and taking a decomposition result as an output neuron of the improved generative adversarial network; carrying out standardization processing on the load data based on a Z-score method, and summarizing the load data into the same dimension; and constructing a regression model about the technical development index, the power data and the technical data by adopting the corrected development index and the obtained load data and utilizing a multiple regression method. According to the invention, the accuracy of power technology development prediction can be improved.

Description

Technical Field

[0001] This invention relates to the field of power development trend prediction technology, and in particular to a method and system for predicting power technology development trends based on data correction. Background Technology

[0002] Load data, as a key synchronous indicator of technological development, shows a high correlation between its changing trends and the level of technological advancement. Typically, a significant increase in load directly reflects accelerated development; therefore, monitoring electricity data can directly reveal the macro-level technological development trend. Currently, academia widely employs methods such as multiple regression analysis, computable general equilibrium models, and mixed grey relational analysis to deeply explore the quantitative relationship between load and various technological indices. However, actual technological systems are dynamically influenced by multiple factors such as industrial restructuring, energy efficiency technology advancements, and population spatial agglomeration, leading to trend deviations between load data and technological development data at certain stages. This necessitates intelligent correction of these inconsistent data points.

[0003] On the other hand, current research focuses primarily on short-term corrections to real-time electricity prices and fitting the spatial distribution of energy data. Existing research is significantly insufficient regarding the mismatch between historical load data and technological development patterns caused by long-term, structural changes (such as industry transformation and upgrading). Although methods exist for detecting and correcting obvious outliers, a systematic and universally applicable method for the rationality assessment and trend correction of massive amounts of historical data is still lacking. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for predicting the development trend of power technology based on data correction. It aims to solve the technical problems in the prior art where there is a systematic deviation between historical load data and technological development trends due to structural changes in technology, and where existing correction methods only target outliers and cannot effectively correct general historical data, resulting in insufficient prediction accuracy of power-technology correlation models.

[0005] In a first aspect, the present invention provides a method for predicting the development trend of power technology based on data correction, the method comprising:

[0006] The Producer Product Development Index (PPI) and the Consumer Product Load Index (CPI) were analyzed to construct calculation models for the PPI and CPI indices.

[0007] Acquire power time-series data within the target area, the power time-series data including unit added value power consumption within the target area;

[0008] Based on the PPI and CPI calculation models, a correlation model between unit added value power consumption and the original PPI and CPI indices is constructed. The PPI and CPI indices are linearly decomposed into power consumption-driven and non-power consumption-interference parts, and the decomposition results are used as the output neurons of the improved generative adversarial network.

[0009] An improved generative adversarial network is used to correct the load data and PPI and CPI indices, and the load data is standardized based on the Z-score method to summarize the load data into the same dimension.

[0010] Using the modified development index and the acquired load data, a regression model on the relationship between technological development indicators, electricity data, and technical data was constructed using a multiple regression method.

[0011] Furthermore, a PPI index calculation model is constructed based on the following formula:

[0012]

[0013] Where, p t p0 represents the technical performance of the power products in the reporting period, q0 represents the sales volume of the power products in the base period, and PPI represents the producer technology index.

[0014] Construct a CPI index calculation model based on the following formula:

[0015]

[0016] Among them, P current P represents the sum of the current technical performance of a group of goods and services. base It represents the sum of the technical performance of a group of goods and services in the base period; CPI is the Consumer Price Index.

[0017] In some embodiments, the improved generative adversarial network includes a generator network and a discriminator network, which jointly correct the PPI and CPI indices through adversarial training, wherein:

[0018] Generator network structure: The input layer receives the raw payload data X. load The original PPI index PPI raw Compared with the original CPI index raw And an influencing factor vector, the influencing factor vector including the industrial electricity consumption growth rate g elec and the load share of the six major energy-intensive industries r high The hidden layers employ a multi-layer fully connected structure, with each layer using the ReLU activation function for non-linear transformation; the output layer outputs the corrected PPI exponent. corr and CPI indexcorr .

[0019] Discriminator network structure: The input layer receives real development index data Y. true Or the corrected data Y output by the generator corr The hidden layer adopts a multi-layer fully connected structure, and the activation function is LeakyReLU. The output layer is a single neuron, and the Sigmoid function is used to output the probability that the data is the real data.

[0020] In some embodiments, the training process of the improved generative adversarial network includes:

[0021] Wasserstein distance is used as the loss function, and a gradient penalty term is introduced to improve training stability;

[0022] The generator's loss function L G for

[0023]

[0024] Where G is the generator network; z represents the input noise vector, which follows a prior distribution p. z (z), where D represents the discriminator network, and c is the condition vector, containing load data, the original PPI index, the original CPI index, the industrial electricity consumption growth rate, and the load share of the six major energy-intensive industries. This represents the expected value of all noise vectors;

[0025] Discriminator loss function L D for:

[0026]

[0027] Where λ1 represents the gradient penalty coefficient, and x represents the real development index data, which follows the real data distribution p. data , For the expected value of all development index data, GP represents the gradient penalty term, which is calculated as follows:

[0028]

[0029] in, Represents the gradient operator, This represents a random sampling point on the line connecting the real data and the generated data. Let represent the expected value of all random sampling points.

[0030] In some embodiments, the training process of the improved generative adversarial network further includes:

[0031] Introduce a reconstruction error term into the generator loss function:

[0032]

[0033] Therefore, the expression for the generator multi-objective loss function is:

[0034]

[0035] in, Let α represent the generator's total loss function, where α is the reconstruction error weighting coefficient.

[0036] In some embodiments, the step of constructing a regression model of technological development indicators, electricity data, and technical data using a multiple regression method with the modified development index and the acquired load data includes:

[0037] The objective function of the regression model is expressed as follows:

[0038]

[0039] Where Obj(θ) is the objective function, Let y be the loss function. i A true index of technological development. The predicted technology development index is represented by the loss function calculated using the mean squared error, Ω(f k ) represents the regularization term of the k-th tree, N is the number of historical data sets, and K is the number of decision trees. In this case, it is related to x. i The quantity is related to the following calculation method:

[0040]

[0041] Where T represents the number of leaf nodes in the tree, λ2 represents the regularization parameter, and w represents the weight of the leaf nodes;

[0042] The approximate objective function Obj in the t-th iteration (t) as follows:

[0043]

[0044] Among them, g i H represents the first-order gradient. i Let f represent the second gradient. t (x i ) represents the t-th tree for sample x i The predicted value, x i =[PPI corr CPI corr ,X load ,g elec ,r high ];

[0045] The first-order gradient is calculated as follows:

[0046]

[0047] The second-order gradient is calculated as follows:

[0048]

[0049] in, To represent the previous round of forecast values The first-order partial derivative, For the true value y i The loss function of the previous round of predictions, The value predicted in the previous round The second-order partial derivatives of .

[0050] In some embodiments, the Pearson correlation coefficient is used to assess the linear correlation between the technology development index predicted by the regression model and the actual technology development index. The specific formula is as follows:

[0051] The Pearson correlation coefficient is calculated as follows:

[0052]

[0053] Where R is the Pearson correlation coefficient, y i Let i be the true technological development index of the i-th sample. The average of all real technology development indices. Let i be the predicted technology development index for the i-th sample. This is the average of all predictive technology development indices.

[0054] Secondly, the present invention provides a power technology development trend prediction system based on data correction, the system comprising:

[0055] The index calculation model construction module is used to analyze the Producer Index and the Residents' Load Index to construct the PPI index calculation model and the CPI index calculation model.

[0056] The time-series data acquisition module is used to acquire power time-series data within a target area, wherein the power time-series data includes unit added value power consumption within the target area;

[0057] The decomposition module is used to construct a correlation model between unit added value power consumption and the original PPI index and the original CPI index based on the PPI index calculation model and the CPI index calculation model. It also performs linear decomposition of the PPI and CPI indices into power consumption-driven parts and non-power consumption interference parts, and uses the decomposition results as the output neurons of the improved generative adversarial network.

[0058] The data correction module is used to correct the load data and PPI and CPI indices using an improved generative adversarial network, and to standardize the load data based on the Z-score method, so that the load data are summarized into the same dimension.

[0059] The regression model building module is used to construct a regression model on technological development indicators, electricity data, and technical data using a multiple regression method, based on the modified development index and the acquired load data.

[0060] Thirdly, the present invention provides a storage medium that stores one or more programs, which, when executed by a processor, implement the above-described method for predicting the development trend of power technology based on data correction.

[0061] Fourthly, the present invention provides an electronic device, the electronic device comprising a memory and a processor, wherein:

[0062] The memory is used to store computer programs;

[0063] When the processor executes the computer program stored in the memory, it implements the above-mentioned method for predicting the development trend of power technology based on data correction.

[0064] Compared with the prior art, the present invention has the following advantages:

[0065] This invention first decomposes the PPI / CPI index into power consumption-driven and non-power consumption-interference components, accurately separating the impact of technical performance and external interference factors on the price index. An improved generative adversarial network (GAN) is employed to introduce gradient penalties and reconstruction error optimization to optimize the adversarial training process. The generator integrates multi-dimensional influencing factors such as industrial electricity consumption and loads of high-energy-consuming industries, while the discriminator enhances outlier sensitivity through LeakyReLU, effectively correcting nonlinear distortions in load data and the price index. After standardizing and unifying data dimensions using Z-scores, a gradient boosting regression model with regularization terms is constructed. Overfitting is suppressed through first- / second-order gradient optimization and tree-structure regularization. Finally, the Pearson correlation coefficient is used to quantify and verify the linear correlation strength between predicted values ​​and actual technological development. This scheme achieves highly robust predictions of power technology development trends through multi-level data correction and model optimization. Attached Figure Description

[0066] Figure 1 This is a flowchart of a power technology development trend prediction method based on data correction proposed in an embodiment of the present invention;

[0067] Figure 2 This is a schematic diagram of the structure of a power technology development trend prediction system based on data correction proposed in an embodiment of the present invention.

[0068] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. 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. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects.

[0070] like Figure 1 As shown, an embodiment of the present invention proposes a method for predicting the development trend of power technology based on data correction. The method includes steps S101 to S105, wherein:

[0071] Step S101: Analyze the Producer Index and the Household Load Index to construct the PPI and CPI calculation models;

[0072] It should be noted that in this step, the causes of inconsistencies in time-series data such as the Producer Price Index (PPI) and Consumer Price Index (CPI) are analyzed, and relevant index models are constructed accordingly.

[0073] Specifically, in some embodiments, the PPI index calculation model is constructed according to the following formula:

[0074]

[0075] Where, p t p0 represents the technical performance of the power products in the reporting period, q0 represents the sales volume of the power products in the base period, and PPI represents the producer technology index.

[0076] Construct a CPI index calculation model based on the following formula:

[0077]

[0078] Among them, P current P represents the sum of the current technical performance of a group of goods and services. base It represents the sum of the technical performance of a group of goods and services in the base period; CPI is the Consumer Price Index.

[0079] Step S102: Obtain power time-series data within the target area, wherein the power time-series data includes unit incremental power consumption within the target area;

[0080] It should be noted that the electricity time-series data for the target area is obtained through web crawling technology. This electricity time-series data represents the unit value-added electricity consumption within that area, and the accounting scope covers the entire life cycle of high-energy-consuming industries, ensuring the accuracy and real-time nature of the data. The web crawling technology includes, but is not limited to, crawler scripts written in Python, which retrieve web page content via HTTP requests and parse HTML data to extract the required meteorological data.

[0081] Step S103: Based on the PPI index calculation model and the CPI index calculation model, construct a correlation model between unit added value power consumption and the original PPI index and the original CPI index, and linearly decompose the PPI and CPI indices into power consumption driven parts and non-power consumption interference parts, and use the decomposition results as the output neurons of the improved generative adversarial network.

[0082] Step S104: The load data, PPI index, and CPI index are corrected using an improved generative adversarial network, and the load data is standardized based on the Z-score method to summarize the load data into the same dimension.

[0083] In this embodiment, the improved generative adversarial network includes a generator network and a discriminator network, which jointly correct the PPI and CPI indices through adversarial training, wherein:

[0084] Generator network structure: The input layer receives the raw payload data X. load The original PPI index PPI raw Compared with the original CPI index raw And an influencing factor vector, the influencing factor vector including the industrial electricity consumption growth rate g elec and the load share of the six major energy-intensive industries r high The hidden layers employ a multi-layer fully connected structure, with each layer using the ReLU activation function for non-linear transformation; the output layer outputs the corrected PPI exponent. corr and CPI index ocrr .

[0085] Discriminator network structure: The input layer receives real development index data Y. true Or the corrected data Y output by the generator corr The hidden layer adopts a multi-layer fully connected structure, and the activation function is LeakyReLU. The output layer is a single neuron, and the Sigmoid function is used to output the probability that the data is the real data.

[0086] In summary, the improved generative adversarial network (GAN) significantly enhances the stability of adversarial training and the accuracy of data correction through a two-stage structural optimization (multi-layer fully connected generator with ReLU activation, and LeakyReLU discriminator with Sigmoid output). The generator input layer integrates key influencing factors such as industrial electricity consumption and load data from high-energy-consuming industries, enabling the correction process to adapt to industry characteristics. The discriminator enhances its sensitivity to identifying anomalous data through the LeakyReLU function, avoiding the collapse problem of traditional GAN ​​models. Experiments show that this structure greatly improves the recognition accuracy in power load data correction tasks, providing a cleaner data foundation for technology development prediction.

[0087] In addition, in some embodiments, Wasserstein distance is used as the loss function, and a gradient penalty term is introduced to improve training stability;

[0088] The generator's loss function L G for

[0089]

[0090] Where G is the generator network; z represents the input noise vector, which follows a prior distribution p. z (z), where D represents the discriminator network, and c is the condition vector, containing load data, the original PPI index, the original CPI index, the industrial electricity consumption growth rate, and the load share of the six major energy-intensive industries. This represents the expected value of all noise vectors;

[0091] Discriminator loss function L D for:

[0092]

[0093] Where λ1 represents the gradient penalty coefficient, and x represents the real development index data, which follows the real data distribution p. data , For the expected value of all development index data, GP represents the gradient penalty term, which is calculated as follows:

[0094]

[0095] in, Represents the gradient operator, This represents a random sampling point on the line connecting the real data and the generated data. Let represent the expected value of all random sampling points.

[0096] By introducing a training strategy incorporating Wasserstein distance and a gradient penalty term, the inherent problems of unstable training and vanishing gradients in the original model are fundamentally addressed. The Wasserstein distance measures the distributional difference between the real and generated data, while the gradient penalty term constrains the discriminator's gradient norm, enabling the generator to continuously receive effective feedback signals.

[0097] Furthermore, in some embodiments, a reconstruction error term is introduced into the generator loss function:

[0098]

[0099] Therefore, the expression for the generator multi-objective loss function is:

[0100]

[0101] in, Let α represent the generator's total loss function, where α is the reconstruction error weighting coefficient.

[0102] The aforementioned multi-objective constraint mechanism not only enhances the interpretability of data correction and avoids overfitting or distortion, but also significantly improves the training stability of generative adversarial networks by balancing adversarial game theory with the need for data fidelity. Ultimately, it provides a more reliable and structurally clear data foundation for predicting the development trend of power technology and significantly enhances the model's adaptability to complex power scenarios.

[0103] In addition, the following optimization mechanisms are included in the process of improving the training of generative adversarial networks:

[0104] (I) Conditional Constraint Mechanism:

[0105] Introduce technological development influencing factors as condition vectors into the generator input to ensure that the corrected development index is consistent with actual conditions such as power data and industrial structure.

[0106] (II) Dynamic Learning Rate Adjustment:

[0107] The learning rate is dynamically adjusted based on the convergence of the discriminator and generator during training to avoid mode collapse and training instability.

[0108] (III) Early Termination Mechanism:

[0109] Training is terminated early when the Pearson correlation coefficient between the generator's output corrected data and the real data no longer increases over several consecutive training cycles to prevent overfitting.

[0110] Step S105: Using the modified development index and the acquired load data, construct a regression model on the technology development index, electricity data, and technology data using a multiple regression method.

[0111] In this step, the objective function of the regression model is expressed as:

[0112]

[0113] Where Obj(θ) is the objective function, Let y be the loss function. i A true index of technological development. The predicted technology development index is represented by the loss function calculated using the mean squared error, Ω(f k ) represents the regularization term of the k-th tree, N is the number of historical data sets, and K is the number of decision trees. In this case, it is related to x. i The quantity is related to the following calculation method:

[0114]

[0115] Where T represents the number of leaf nodes in the tree, λ2 represents the regularization parameter, and w represents the weight of the leaf nodes;

[0116] The approximate objective function Obj in the t-th iteration (t) as follows:

[0117]

[0118] Among them, g i H represents the first-order gradient. i Let f represent the second gradient. t (x i ) represents the t-th tree for sample x i The predicted value, x i =[PPI corr CPI corr X load g elec r high ];

[0119] The first-order gradient is calculated as follows:

[0120]

[0121] The second-order gradient is calculated as follows:

[0122]

[0123] in, To represent the previous round of forecast values The first-order partial derivative, For the true value y i The loss function of the previous round of predictions, The value predicted in the previous round The second-order partial derivatives of .

[0124] By combining the loss function with a structured regularization term, the model's generalization ability and computational efficiency are significantly improved. Specifically, using mean squared error as the loss function ensures accurate measurement of prediction bias, while the regularization term based on the number of leaf nodes and the L2 norm of the weights effectively suppresses model complexity and avoids overfitting. By approximating the objective function through Taylor expansion and introducing first- and second-order gradient information, the convergence speed of the optimization process is accelerated, and dynamic gradient calculation enhances the model's adaptability to data distribution. Ultimately, a balanced optimization of complexity and performance is achieved while maintaining high prediction accuracy.

[0125] Furthermore, in some embodiments, the Pearson correlation coefficient is used to assess the linear correlation between the technology development index predicted by the regression model and the actual technology development index, as shown in the following formula:

[0126] The Pearson correlation coefficient is calculated as follows:

[0127]

[0128] Where R is the Pearson correlation coefficient, y i Let i be the true technological development index of the i-th sample. The average of all real technology development indices. Let i be the predicted technology development index for the i-th sample. This is the average of all predictive technology development indices.

[0129] In summary, based on the aforementioned data-corrected method for predicting power technology development trends, Python is used to crawl load data, improving the accuracy of data analysis. An adversarial neural network is employed to intelligently correct abnormal development indices. By combining the consistency of load data, technology data, and related influencing factor data over time, a multiple regression prediction model is constructed. Relevance parameters suitable for regression analysis are determined, enabling reasonable predictions of future technological development. This approach addresses the technical problem of insufficient prediction accuracy in the "power-technology" correlation model, stemming from systematic deviations between historical load data and technological development trends caused by structural technological changes, and the inability of existing correction methods to effectively correct outliers and general historical data.

[0130] like Figure 2 As shown, one embodiment of the present invention proposes a power technology development trend prediction system based on data correction, the system comprising:

[0131] The index calculation model construction module 10 is used to analyze the industrial producer output development index and the residential load index to construct the PPI index calculation model and the CPI index calculation model.

[0132] The time-series data acquisition module 20 is used to acquire power time-series data within a target area, the power time-series data including unit added value power consumption within the target area;

[0133] The decomposition module 30 is used to construct a correlation model between unit added value power consumption and the original PPI index and the original CPI index based on the PPI index calculation model and the CPI index calculation model, and to linearly decompose the PPI and CPI indices into power consumption driven parts and non-power consumption interference parts, and use the decomposition results as the output neurons of the improved generative adversarial network.

[0134] The data correction module 40 is used to correct the load data and PPI and CPI indices using an improved generative adversarial network, and to standardize the load data based on the Z-score method, so as to summarize the load data into the same dimension.

[0135] The regression model construction module 50 is used to construct a regression model on the technology development index, electricity data, and technology data using a multiple regression method with the modified development index and the acquired load data.

[0136] In another aspect, the present invention also proposes a storage medium on which one or more programs are stored, which, when executed by a processor, implement the above-described method for predicting the development trend of power technology based on data correction.

[0137] In another aspect, the present invention also proposes an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to realize the above-mentioned method for predicting the development trend of power technology based on data correction.

[0138] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain stored, communicated, propagated, or transmitted programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0139] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0140] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0141] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A method for predicting the development trend of power technology based on data correction, characterized in that, The method includes: The Producer Product Development Index (PPI) and the Consumer Load Index (CPI) were analyzed to construct calculation models for the PPI and CPI indices. Acquire power time-series data within the target area, the power time-series data including unit added value power consumption within the target area; Based on the PPI and CPI calculation models, a correlation model between unit added value power consumption and the original PPI and CPI indices is constructed. The PPI and CPI indices are linearly decomposed into power consumption-driven and non-power consumption-interference parts, and the decomposition results are used as the output neurons of the improved generative adversarial network. An improved generative adversarial network is used to correct the load data and PPI and CPI indices, and the load data is standardized based on the Z-score method to summarize the load data into the same dimension. Using the modified development index and the acquired load data, a regression model on the relationship between technological development indicators, electricity data, and technical data was constructed using a multiple regression method.

2. The method for predicting the development trend of power technology based on data correction according to claim 1, characterized in that, Construct a PPI index calculation model based on the following formula: Where, p t p0 represents the technical performance of the power products in the reporting period, q0 represents the sales volume of the power products in the base period, and PPI represents the producer technology index. Construct a CPI index calculation model based on the following formula: Among them, P current P represents the sum of the current technical performance of a group of goods and services. base It represents the sum of the technical performance of a group of goods and services in the base period; CPI is the Consumer Price Index.

3. The method for predicting the development trend of power technology based on data correction according to claim 2, characterized in that, The improved generative adversarial network comprises a generator network and a discriminator network, which jointly corrects the PPI and CPI indices through adversarial training, wherein: Generator network structure: The input layer receives the raw payload data X. load The original PPI index PPI raw Compared with the original CPI index raw And an influencing factor vector, the influencing factor vector including the industrial electricity consumption growth rate g elec and the load share of the six major energy-intensive industries r high The hidden layers employ a multi-layer fully connected structure, with each layer using the ReLU activation function for non-linear transformation; the output layer outputs the corrected PPI exponent. corr and CPI index corr . Discriminator network structure: The input layer receives real development index data Y. true Or the corrected data Y output by the generator corr The hidden layer adopts a multi-layer fully connected structure, and the activation function is LeakyReLU. The output layer is a single neuron, and the Sigmoid function is used to output the probability that the data is the real data.

4. The method for predicting the development trend of power technology based on data correction according to claim 3, characterized in that, The training process of the improved generative adversarial network includes: Wasserstein distance is used as the loss function, and a gradient penalty term is introduced to improve training stability; The generator's loss function L G for Where G is the generator network; z represents the input noise vector, which follows a prior distribution p. z (z), where D represents the discriminator network, and c is the condition vector, containing load data, the original PPI index, the original CPI index, the industrial electricity consumption growth rate, and the load share of the six major energy-intensive industries. This represents the expected value of all noise vectors; Discriminator loss function L D for: Where λ1 represents the gradient penalty coefficient, and x represents the real development index data, which follows the real data distribution p. data , For the expected value of all development index data, GP represents the gradient penalty term, which is calculated as follows: in, Represents the gradient operator, This represents a random sampling point on the line connecting the real data and the generated data. Let represent the expected value of all random sampling points.

5. The method for predicting the development trend of power technology based on data correction according to claim 4, characterized in that, The training process of the improved generative adversarial network also includes: Introduce a reconstruction error term into the generator loss function: Therefore, the expression for the generator multi-objective loss function is: in, Let α represent the generator's total loss function, where α is the reconstruction error weighting coefficient.

6. The method for predicting the development trend of power technology based on data correction according to claim 5, characterized in that, The steps of constructing a regression model on technological development indicators, electricity data, and technical data using a multiple regression method, based on the modified development index and the acquired load data, include: The objective function of the regression model is expressed as follows: Where Obj(θ) is the objective function. Let y be the loss function. i A true index of technological development. The predicted technology development index is represented by the loss function calculated using the mean squared error, Ω(f k ) represents the regularization term of the k-th tree, N is the number of historical data sets, K is the number of decision trees, and x i The quantity is related to the calculation method as follows: Where T represents the number of leaf nodes in the tree, λ2 represents the regularization parameter, and w represents the weight of the leaf nodes; The approximate objective function Obj in the t-th iteration (t) as follows: Among them, g i H represents the first-order gradient. i f represents the second-order gradient. t (x i ) represents the t-th tree for sample x i The predicted value, x i =[PPI corr CPI corr X load g elec r high ]; The first-order gradient is calculated as follows: The second-order gradient is calculated as follows: in, To represent the previous round of forecast values The first-order partial derivative, For the true value y i The loss function of the previous round of predictions, The value predicted in the previous round The second-order partial derivatives of .

7. The method for predicting the development trend of power technology based on data correction according to claim 6, characterized in that, The Pearson correlation coefficient is used to assess the linear correlation between the technology development index predicted by the regression model and the actual technology development index. The specific formula is as follows: The Pearson correlation coefficient is calculated as follows: Where R is the Pearson correlation coefficient, y i Let i be the true technological development index of the i-th sample. The average of all real technology development indices. Let i be the predicted technology development index for the i-th sample. This is the average of all predicted technology development indices.

8. A power technology development trend prediction system based on data correction, characterized in that, The system includes: The index calculation model construction module is used to analyze the Producer Index and the Residents' Load Index to construct the PPI index calculation model and the CPI index calculation model. The time-series data acquisition module is used to acquire power time-series data within a target area, wherein the power time-series data includes unit added value power consumption within the target area; The decomposition module is used to construct a correlation model between unit added value power consumption and the original PPI index and the original CPI index based on the PPI index calculation model and the CPI index calculation model. It also performs linear decomposition of the PPI and CPI index into power consumption driven part and non-power consumption interference part, and uses the decomposition result as the output neuron of the improved generative adversarial network. The data correction module is used to correct the load data and PPI and CPI indices using an improved generative adversarial network, and to standardize the load data based on the Z-score method, so that the load data are summarized into the same dimension. The regression model building module is used to construct a regression model on technological development indicators, electricity data, and technical data using a multiple regression method, based on the modified development index and the acquired load data.

9. A storage medium, characterized in that, The storage medium stores one or more programs that, when executed by a processor, implement the data-corrected power technology development trend prediction method as described in any one of claims 1-7.

10. An electronic device comprising a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the power technology development trend prediction method based on data correction as described in any one of claims 1-7.

Images