Industrial chain electricity consumption prediction method, system and device based on TCN-LSTM model fusing causal relationship, and medium

Abstract

The invention belongs to the technical field of industrial power consumption prediction, and discloses an industrial chain power consumption prediction method, system and device based on a TCN-LSTM model fusing a causal relationship, and a medium, so as to solve the problems of poor interpretability, low generalization and low prediction precision. The method comprises the following steps: acquiring target industry chain information; acquiring a historical electricity consumption sequence and an external influence factor sequence of each industry; performing multi-band component decomposition on the historical electricity consumption time sequence and the external influence factor time sequence of each industry by adopting a wavelet decomposition algorithm; a causal analysis method is adopted to quantify causal intensity, directionality and transmission time lag among the industries; quantifying the correlation strength between each external influence factor and the industry power consumption component under each frequency band by adopting a grey correlation analysis method; a TCN-LSTM prediction model driven by the causal relationship is established; and the TCN-LSTM prediction model is used to predict the power consumption of each industry under each frequency band.

Description

Technical Field

[0001] This invention belongs to the field of industrial electricity consumption forecasting technology, specifically relating to a method, system, equipment, and medium for forecasting electricity consumption in the industrial chain based on a TCN-LSTM model that integrates causal relationships. Background Technology

[0002] The industrial chain is a core component of the modern economic system, and its operation is closely related to electricity demand, making electricity consumption forecasting a key link in optimizing energy dispatch and enhancing industrial resilience.

[0003] To gain a deeper understanding of the complex interactions within the industrial chain, causal analysis methods have been introduced into the field of industrial electricity consumption forecasting in recent years. For example, some studies combine industry correlation coefficients with Granger causality tests to construct industrial causal networks using electricity consumption data and employ algorithms such as minimum spanning trees to extract dominant paths, thereby revealing the transmission mechanism of the industrial chain. While these methods introduce the concept of causality, Granger causality tests are inherently applicable to linear systems and struggle to accurately capture the prevalent nonlinear dynamic relationships within the industrial chain, thus limiting their application.

[0004] To overcome this limitation, nonlinear causal inference methods such as convergent cross-mapping and partial cross-mapping have been proposed and gradually applied. These methods can identify strongly coupled, nonlinear direct causal relationships between variables and effectively eliminate interference from indirect causal paths, thus making the identification of causal structures more accurate and robust. In terms of feature processing, grey relational analysis, as an effective tool for measuring the correlation between external influencing factors and electricity consumption, is widely used in feature screening and variable selection for industrial systems, improving the scientific rigor of model input features.

[0005] However, most existing methods remain at the stage of causal relationship identification and network construction, exhibiting a significant technical deficiency: the difficulty in effectively embedding information such as causal strength, transmission lag, and correlation of external variables obtained from quantitative analysis into the final industrial electricity consumption forecasting model. This limits the practical utilization of valuable causal information in data-driven models, hindering the automatic screening of key influencing factors and the systematic optimization of model input features. Especially in real-world scenarios where various links within the industrial chain are highly coupled and external environmental variables are complex and ever-changing, how to organically integrate advanced causal analysis, correlation analysis, and modern data-driven models such as deep learning to construct an industrial electricity consumption forecasting method that combines interpretability, high generalization, and high predictive accuracy has become a critical technical challenge that urgently needs to be overcome in this field. Summary of the Invention

[0006] Based on the aforementioned shortcomings and deficiencies in the existing technology, one of the objectives of this invention is to at least solve one or more of the aforementioned problems in the existing technology. In other words, one of the objectives of this invention is to provide a method, system, device, and medium for predicting industrial electricity consumption based on a TCN-LSTM model that integrates causal relationships, which meets one or more of the aforementioned requirements, so as to improve the scientificity, accuracy, and adaptability of industrial electricity consumption prediction.

[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model that integrates causal relationships, comprising the following steps: S1. Obtain information on the industries and corresponding companies involved in the target industrial chain to determine the nodes and hierarchical relationships within the target industrial chain, thereby building the target industrial chain system; S2. Obtain the historical electricity consumption sequence of each industry within the target industrial chain system and the sequence of external influencing factors related to each industry, and perform preprocessing operations. S3. Using wavelet decomposition algorithm, multi-frequency component decomposition is performed on the preprocessed historical electricity consumption time series of each industry and the time series of external influencing factors to obtain the industry electricity consumption components and influencing factor components in different frequency bands. S4. Using causal analysis, conduct causal relationship analysis on the electricity consumption components of each industry under each frequency band, and quantify the causal strength, directionality and transmission time delay between industries. S5. Use grey relational analysis to quantify the correlation strength between external influencing factors and industry electricity consumption components in each frequency band. S6. Based on the causal relationship obtained in step S4 and the correlation strength obtained in step S5, construct input features and input them into the TCN-LSTM prediction model for training, so as to establish a causal relationship-driven TCN-LSTM prediction model. S7. Using the TCN-LSTM prediction model constructed in step S6, predict the electricity consumption of each industry in the target industrial chain system under each frequency band, and reconstruct and integrate the prediction results of different frequency bands to obtain the electricity consumption prediction results of the target industrial chain.

[0008] As a preferred embodiment, step S4 involves using a causal analysis method to analyze causal relationships, including the following steps: S41. Using a convergent cross-mapping algorithm, pairwise causal analysis is performed on each industry within the target industrial chain system to identify preliminary causal relationships. S42. The preliminary causal relationship is corrected using a partial cross-mapping algorithm to identify and eliminate indirect causal effects transmitted by intermediate variables in the causal network, thereby preserving the direct causal path.

[0009] As a preferred embodiment, step S41 specifically involves: For any two industries within the target industrial chain system and Each of them constructs a shadow manifold in phase space based on its electricity consumption time series components; Based on the industry The shadow manifold of the electricity consumption sequence for the industry The electricity consumption sequence is estimated to obtain the estimated value; Calculate the industry The Pearson correlation coefficient between the actual sequence and the estimated value is used as a reference from the industry. To the industry The initial causal strength; Introducing a time delay parameter, the industry The sequence is advanced in time, the causal strength under different time delays is calculated, and the optimal transmission time delay and the corresponding strongest causal strength are identified. If the strongest causal strength is greater than a preset first threshold, then it is determined that the industry... To the industry There is a significant causal relationship. The first threshold The value of is between 0 and 1.

[0010] As a preferred embodiment, step S42 specifically involves: After identifying the preliminary causal relationship network involving three or more industries in step S41, for any one of them, from industry Through industry To the industry Potential indirect causal pathways, perform the following operations: industry As a control variable, the calculation in the removal industry After the impact on electricity consumption, the industry With the industry The partial correlation coefficient between them is used as the correlation coefficient from the industry. To the industry The strength of direct causality; If the direct causal strength is less than a preset second threshold, then it is determined that the industry... To the industry The causal relationship is considered indirect and is eliminated; if the direct causal strength is greater than or equal to the second threshold, the direct causal relationship is retained. The second threshold... The value of is between 0 and 1.

[0011] As a preferred option, step S5 specifically involves: The industry electricity consumption sequence is set as the reference sequence, and the external influencing factor sequence is set as the comparison sequence; The initial value normalization method is used to perform dimensionless processing on each sequence; Calculate the grey relational coefficient between the sequence of external influencing factors and the sequence of industry electricity consumption at each time point; The average value of the grey relational coefficients at all times is used to obtain the grey relational degree between the external influencing factor and the industry's electricity consumption, which is then used as the correlation strength.

[0012] As a preferred approach, the construction of input features in step S6 specifically involves: For the target industry to be predicted, the electricity consumption components of the upstream and downstream industries corresponding to the direct causal relationship that was determined to exist and be retained in step S4 in the corresponding frequency band are weighted with the causal strength obtained in step S4 to form the first input part. The components of external influencing factors related to the target industry in the corresponding frequency band are weighted using the correlation strength obtained in step S5 as the weight to form the second input part; The first input portion and the second input portion are combined to form the input feature.

[0013] As a preferred approach, the TCN-LSTM prediction model described in step S6 is constructed in the following manner: The temporal convolutional network module receives the input features and performs feature extraction through a multi-layer structure containing dilated convolutions to capture long sequence dependencies; The Long Short-Term Memory (LSTM) network module receives the output of the temporal convolutional network module and learns a gating mechanism to memorize the long-term temporal dependencies of the time series. The fully connected layer processes the output of the long short-term memory network module to generate the final power consumption prediction value.

[0014] Secondly, the present invention provides an industrial chain electricity consumption prediction system based on a TCN-LSTM model that integrates causal relationships, for implementing the industrial chain electricity consumption prediction method as described in the first aspect.

[0015] Thirdly, the present invention provides an electronic device, the computer device including a memory, a processor and a computer program, wherein when the computer program is executed by the processor, it implements the industrial chain electricity consumption prediction method as described in the first aspect.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the industrial chain electricity consumption prediction method as described in the first aspect.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. Achieved accurate mining and purification of causal relationships within the industrial chain: Compared with technical solutions that only use linear methods such as Granger causality tests or solely use convergent cross-mapping (CCM), this invention innovatively combines convergent cross-mapping (CCM) with partial cross-mapping (PCM) algorithms. This combination not only identifies nonlinear causal relationships but also effectively identifies and eliminates indirect causal effects transmitted by intermediate variables, thereby accurately revealing the real and direct causal paths between upstream and downstream industries in the industrial chain, significantly improving the accuracy and reliability of causal network construction.

[0018] 2. A prediction model input based on deep fusion of multi-source and multi-scale information was constructed: Compared to traditional methods that only use raw electricity consumption sequences or simply splice together external factors, this invention introduces grey relational analysis, which is performed simultaneously with causal analysis. This enables simultaneous quantitative measurement of the causal relationships within the industrial chain and the correlation of external influencing factors. In this way, key internal and external driving variables can be scientifically selected, and the quantified causal strength and correlation degree can be used as weights to construct model input features with higher information density and clearer physical meaning, thereby improving the scientific rigor and comprehensiveness of the prediction model from the source.

[0019] 3. Enhanced overall performance and decision support value of the prediction model: This invention organically embeds the features selected and weighted through causal and correlation analysis into the TCN-LSTM deep learning model. The TCN module excels at capturing local dependency patterns in long sequences, while the LSTM module is adept at learning long-term temporal dynamics. The combination of the two forms a powerful prediction engine. Compared to "black box" purely data-driven models, the causal relationship-driven model constructed in this invention not only has higher prediction accuracy and generalization ability but also better interpretability. Decision-makers can trace the key causal chains and major influencing factors behind the prediction results, thus providing strong data support and decision-making basis for industrial chain risk management and energy scheduling planning.

[0020] Further or more detailed beneficial effects will be described in conjunction with specific embodiments in the detailed implementation. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the electricity consumption forecasting method for the industrial chain described in Embodiment 1 of the present invention.

[0023] Figure 2 This is a structural diagram of the electronic device described in Embodiment 3 of the present invention.

[0024] Icon labels: 200. Electronic devices; 201. Processor; 202. Communication bus; 203. User interface; 204. Network interface; 205. Memory. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0026] In the following description, several embodiments of the present invention are provided. Different embodiments can be substituted or combined. Therefore, the present invention can also be considered to include all possible combinations of the same and / or different embodiments described. Thus, if one embodiment includes features A, B, and C, and another embodiment includes features B and D, then the present invention should also be considered to include embodiments containing one or more other possible combinations of A, B, C, and D, even if such embodiments are not explicitly described in the following text.

[0027] The following description provides examples and does not limit the scope, applicability, or examples set forth in the claims. Changes may be made to the function and arrangement of the described elements without departing from the scope of the invention. Various processes or components may be appropriately omitted, substituted, or added to the various examples. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined into other examples.

[0028] To facilitate a better understanding of the embodiments of the present invention, its application scenarios will be explained before providing a detailed explanation of the specific implementation methods.

[0029] The electricity consumption forecasting method for the industrial chain described in the embodiments of this specification is applied to various scenarios that require accurate forecasting of the future electricity load of a specific industrial chain as a whole or its key links. In these scenarios, the application of the electricity consumption forecasting method for the industrial chain aims to: 1. Improve the accuracy of power system dispatching and planning: Provide power grid companies with forward-looking electricity consumption forecast data based on the operation logic of the industrial chain for power generation planning, power trading and transmission and distribution dispatch, thereby optimizing the allocation of power resources and ensuring the safe and stable operation of the power grid.

[0030] 2. Enhance the risk management and decision-making capabilities of the industrial chain itself: Help government macroeconomic management departments, industry associations or core enterprises to understand the dynamic transmission mechanism within the industrial chain, assess the potential impact of changes in the external environment on the overall energy consumption of the industrial chain, and provide data-driven decision support for formulating industrial policies, supply chain adjustments and energy efficiency management strategies.

[0031] 3. Optimize enterprise production, operation and energy management: Individual enterprises in the industrial chain, especially large electricity consumers, can use this method to more accurately predict their own electricity demand trends and understand their position in the causal network of the industrial chain, thereby rationally arranging production plans, participating in demand-side response, and achieving cost reduction and efficiency improvement.

[0032] The following is a brief explanation of the industrial chain, historical electricity consumption sequence, external influencing factor sequence, wavelet decomposition algorithm, causal analysis method, and TCN-LSTM prediction model involved in several embodiments of this specification: An industry chain refers to a collection of industries with inherent technological and economic connections, encompassing all stages from raw material procurement and intermediate product manufacturing to finished product production and distribution, all centered around a specific final product or service. In this invention, the industry chain is modeled as a complex network composed of multiple node industries linked by causal relationships.

[0033] Historical electricity consumption series refers to the sequence formed by arranging the electricity consumption data of various nodes in the industrial chain in chronological order over a period of time (such as daily, weekly, or monthly). This is the basic data for time series analysis and model training.

[0034] The external influencing factor sequence refers to the data sequence of external variables that may affect the electricity consumption of the target industry, excluding the electricity consumption of industries within the industrial chain. These factors may include, but are not limited to, product market price indices, macroeconomic indicators, temperature and weather data, etc., and they exert influence from outside the system.

[0035] Wavelet decomposition is a signal processing technique used to decompose non-stationary time series (such as electricity consumption series) into subsequences (i.e., components) of different frequencies. This invention employs this algorithm to separate the long-term trend, medium-term cycle, and short-term fluctuations in the original sequence, enabling more refined analysis of causal relationships and correlations across different time scales.

[0036] Causal analysis method: In this invention, it specifically refers to a nonlinear causal inference method that combines Convergent Cross Mapping (CCM) and Partial Cross Mapping (PCM). This method is used to accurately quantify the driving strength, direction of influence, and transmission time delay of changes in electricity consumption in one industry on changes in electricity consumption in another industry within a supply chain. It can also distinguish and eliminate indirect causal relationships, thereby revealing the true direct causal path.

[0037] The TCN-LSTM prediction model is a deep learning model composed of a Temporal Convolutional Network (TCN) and a Long Short-Term Memory (LSTM) network. The TCN module uses dilated convolutions to capture long-term dependency patterns in the input features, while the LSTM module further learns and memorizes complex dynamic characteristics over time. In this invention, this model is specifically built to fuse input features that have been filtered and weighted through causal and correlation analysis, and it is the core engine for achieving high-precision prediction.

[0038] Example 1: like Figure 1 As shown, this embodiment provides a method for predicting electricity consumption in the industrial chain based on a TCN-LSTM model that integrates causal relationships, including the following steps: Step S1: Obtain information on the industries and corresponding companies involved in the target industrial chain to determine the nodes and hierarchical relationships within the target industrial chain, thereby building the target industrial chain system.

[0039] Step S2: Obtain the historical electricity consumption sequence of each industry within the target industrial chain system. , External influencing factors related to each industry and The first in the industry A sequence of external influencing factors ,in This indicates the industry index number; there are a total of [number] in the target industry chain. o Each industry t=1,2,…,T represents time points, T is the length of the time series, and preprocessing operations (such as data cleaning, missing value imputation, and normalization) are performed.

[0040] Step S3: Using wavelet decomposition algorithm, multi-frequency component decomposition is performed on the preprocessed historical electricity consumption time series of each industry and the time series of external influencing factors to obtain the industry electricity consumption components and influencing factor components in different frequency bands.

[0041] Step S4: Using causal analysis, conduct causal relationship analysis on the electricity consumption components of each industry under each frequency band, and quantify the causal strength, directionality and transmission time delay between industries.

[0042] Specifically, step S4 includes the following steps: Step S41: Using a convergent cross-mapping algorithm, perform pairwise causal analysis to identify each industry within the target industrial chain system, obtaining preliminary causal relationships. The more specific calculation process is as follows: Assuming there is in the industrial chain o Each industry , and It is an index for the industry. , and Represent , and The time series of industry electricity consumption. When the industry electricity consumption time series The phase space contains the time series of industry electricity consumption. When the phase space information is obtained, that is... The estimated electricity consumption can be obtained through If it can be obtained, then the industry industry There is a causal relationship.

[0043] Constructing industry electricity consumption series and The shadow manifold in phase space is expressed as: (1), (2), In equations (1) and (2), and Represent and The shadow manifold, and Represents the time series of electricity consumption in the industry and exist t The value of the moment. NRepresents the length of the time series. E Represents the embedding dimension. Represents the time lag period.

[0044] The formula for calculating the shadow manifold estimate is: (3), (4), In equations (3) and (4), express based on Estimated electricity consumption in the industry express In shadow manifold The first l Neighboring points, express The weight, L Indicates the number of neighboring points. express and The Manhattan distance between them.

[0045] From the industry To the industry The causal strength between them is quantified by formula (5): (5), In equation (5), Indicates from the industry To the industry The causal strength, It is the Pearson correlation coefficient. and They represent and The average value.

[0046] Furthermore, considering the impact of industry inventory, there is a significant time lag in causal transmission between industries. This embodiment performs causal analysis considering the transmission time difference, showing that the difference compared to the above calculation lies only in the input, and its expression is: (6), In equation (6), Indicates the transmission delay. Numerically not less than 0, in a specific From the industry To the industry The strongest causal strength will be used as the basis for industry To the industry The most reasonable causal strength corresponds to the following transmission delay, causal strength, and estimated value: , and , (7), In equation (7), This is an empirical threshold, with a value between 0 and 1.

[0047] Step S42: Correct the preliminary causal relationship using a partial cross-mapping algorithm, identify and eliminate indirect causal effects caused by intermediate variables in the causal network, thereby preserving the direct causal path. More specifically, the operation is as follows: Assume that the causal relationship transmission direction among the three industries identified after analysis using the convergent cross-mapping algorithm is as follows: (8), Calculate using the convergent cross-mapping algorithm and , , and The partial correlation coefficient between them can be calculated using formula (9). As a third variable to eliminate the transitivity of causality: (9), In equation (9), This indicates a node to be removed, taking into account transmission latency. The impact of electricity consumption arrive The strength of direct causality.

[0048] if So from the industry To the industry There is a direct causal relationship between them. This is an empirical threshold, with a value between 0 and 1.

[0049] Step S5: Employ grey relational analysis to quantify the correlation strength between external influencing factors and industry electricity consumption components across different frequency bands. The specific operation is as follows: By industry For example, assuming it is affected Influenced by external factors Indicates the industry's historical electricity consumption The A series of external influencing factors, including the industry's electricity consumption series. The sequence is set as the reference sequence, while its external influencing factor sequence is set as the comparison sequence. To eliminate dimensional differences between different physical quantities, an initial value normalization method is used, calculated using the following formula: (10) In equation (10), This represents the normalized value of the initial value. express exist t The first moment The numerical values ​​of each external influencing factor. express The initial value is a normalized value.

[0050] The formulas for calculating the grey correlation coefficient and correlation degree between influencing factors and electricity consumption are as follows: (11), (12) In equations (11) and (12), Represents the industry's electricity consumption series and external influencing factors exist t The grey relational coefficient at time t, Represents the industry's electricity consumption series External influencing factors The degree of gray correlation between them This represents the distinguishing factor, which is usually taken as 0.5.

[0051] Step S6: Based on the causal relationship obtained in Step S4 and the correlation strength obtained in Step S5, construct input features and input them into the TCN-LSTM prediction model for training to establish a causal relationship-driven TCN-LSTM prediction model. The specific operation is as follows: Industry in the low frequency band Taking electricity consumption forecasting as an example, assuming the industry have Industries with strong causal relationships between upstream and downstream sectors External influencing factors. Representation and Industry The first with a strong relationship If there are multiple industries, then the input of the model can be constructed using formula (13): (13) In equation (13), yes t The input weight at a given time indicates the correlation between the input and the industry at low frequencies. There is a strong causal relationship, and This indicates the optimal conduction time delay in the low-frequency band, representing the industry... To the industry The causal strength value, Indicating in the low frequency band t At any time based The shadow manifold constructed from the electricity consumption sequence, yes t The weighted value of the model input at time t represents the industry External influencing factors This indicates the industry in the low-frequency band. External influencing factors The gray correlation value between them Indicates low frequency band t Time based The constructed shadow manifold values, Indicates in t The input time series of the proposed model is used to predict industry trends in the low-frequency range. Electricity consumption This represents the number of historical data samples used to predict the electricity consumption features of the training set.

[0052] Building a TCN framework, assuming the existence of layer: (14) In equation (14), Indicates in t At time 1, the TCN model is at time 2. nt Layer input, Indicates the first nt The output of the dilation convolution operation. Indicates the first nt The dilation factor of the second dilated convolution operation. Indicates the first i One filter, Indicates the size of the filter. This represents the activation function. The convolutional layer used to adjust the shape is shown in Equation (14). .

[0053] The expression for constructing an LSTM framework is: (15) In equation (15), express t The activation value of the forget gate at any given time. express t The activation value of the output gate at any given time. express t The activation value of the input gate at any given time. expresst The hidden state at any given moment express t The candidate cell state at time t. express t The cell state at time t, , , and The weight matrices represent the forget gate, input gate, output gate, and candidate cell states, respectively. , , and These represent the bias vectors for the forget gate, input gate, output gate, and candidate cell states, respectively. and This represents the activation function.

[0054] The expression for adjusting the shape of the final output using a fully connected layer is: (16) In equation (16), Indicates in The electricity consumption of the target industry as predicted at any given time. and This represents the final output weight matrix and bias values. Step S7: Using the TCN-LSTM prediction model constructed in step S6, predict the electricity consumption of each industry in the target industrial chain system under each frequency band, and reconstruct and integrate the prediction results of different frequency bands to obtain the electricity consumption prediction results of the target industrial chain.

[0055] Example 2: This embodiment provides a power consumption prediction system for the industrial chain based on the TCN-LSTM model that integrates causal relationships, which is used to implement the power consumption prediction method for the industrial chain as described in Embodiment 1.

[0056] Example 3: like Figure 2 As shown, this embodiment provides an electronic device, which may include: at least one processor, at least one network interface, a user interface, a memory, and at least one communication bus.

[0057] The communication bus can be used to enable communication between the various components mentioned above.

[0058] The user interface may include buttons, and optional user interfaces may also include standard wired interfaces and wireless interfaces.

[0059] The network interface may include, but is not limited to, Bluetooth modules, NFC modules, Wi-Fi modules, etc.

[0060] The processor may include one or more processing cores. It connects various parts of the electronic device via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in memory, and accessing data stored in memory to perform various functions and process data. Optionally, the processor can be implemented using at least one hardware form of DSP, FPGA, or PLA. The processor may integrate one or more of the following: CPU, GPU, and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor.

[0061] The memory may include RAM or ROM. Optionally, the memory may include a non-transitory computer-readable medium. The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor. The memory, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a prediction application. The processor can be used to call the prediction application stored in the memory and execute the steps of the industrial chain electricity consumption prediction method mentioned in the foregoing embodiments.

[0062] Example 4: This embodiment provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the above-described instructions. Figure 1 One or more steps in the illustrated embodiment. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0063] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0064] Those skilled in the art will understand that all or part of the processes in the method of Embodiment 1 described above can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and the implementation scheme can be combined arbitrarily.

[0065] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0066] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0067] The above description is merely an exemplary embodiment of the present invention and should not be construed as limiting the scope of the invention. Any equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of embodiments of the invention upon considering the specification and practicing the disclosure herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of the invention are defined by the claims.

Claims

1. A method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model that integrates causal relationships, characterized in that, Including the following steps: S1. Obtain information on the industries and corresponding companies involved in the target industrial chain to determine the nodes and hierarchical relationships within the target industrial chain, thereby building the target industrial chain system; S2. Obtain the historical electricity consumption sequence of each industry within the target industrial chain system and the sequence of external influencing factors related to each industry, and perform preprocessing operations. S3. Using wavelet decomposition algorithm, multi-frequency component decomposition is performed on the preprocessed historical electricity consumption time series of each industry and the time series of external influencing factors to obtain the industry electricity consumption components and influencing factor components in different frequency bands. S4. Using causal analysis, conduct causal relationship analysis on the electricity consumption components of each industry under each frequency band, and quantify the causal strength, directionality and transmission time delay between industries. S5. Use grey relational analysis to quantify the correlation strength between external influencing factors and industry electricity consumption components in each frequency band. S6. Based on the causal relationship obtained in step S4 and the correlation strength obtained in step S5, construct input features and input them into the TCN-LSTM prediction model for training, so as to establish a causal relationship-driven TCN-LSTM prediction model. S7. Using the TCN-LSTM prediction model constructed in step S6, predict the electricity consumption of each industry in the target industrial chain system under each frequency band, and reconstruct and integrate the prediction results of different frequency bands to obtain the electricity consumption prediction results of the target industrial chain.

2. The method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model with fused causal relationships as described in claim 1, characterized in that, Step S4, which involves performing causal relationship analysis using a causal analysis method, includes the following steps: S41. Using a convergent cross-mapping algorithm, pairwise causal analysis is performed on each industry within the target industrial chain system to identify preliminary causal relationships. S42. The preliminary causal relationship is corrected using a partial cross-mapping algorithm to identify and eliminate indirect causal effects transmitted by intermediate variables in the causal network, thereby preserving the direct causal path.

3. The method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model with fused causal relationships as described in claim 2, characterized in that, Step S41 is as follows: For any two industries within the target industrial chain system and Each of them constructs a shadow manifold in phase space based on its electricity consumption time series components; Based on the industry The shadow manifold of the electricity consumption sequence for the industry The electricity consumption sequence is estimated to obtain the estimated value; Calculate the industry The Pearson correlation coefficient between the actual sequence and the estimated value is used as a reference from the industry. To the industry The initial causal strength; Introducing a time delay parameter, the industry The sequence is advanced in time, the causal strength under different time delays is calculated, and the optimal transmission time delay and the corresponding strongest causal strength are identified. If the strongest causal strength is greater than a preset first threshold, then it is determined that the industry... To the industry There is a significant causal relationship.

4. The method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model with fused causal relationships as described in claim 3, characterized in that, Step S42 is as follows: After identifying the preliminary causal relationship network involving three or more industries in step S41, for any one of them, from industry Through industry To the industry Potential indirect causal pathways, perform the following operations: industry As a control variable, the calculation in the removal industry After the impact on electricity consumption, the industry With the industry The partial correlation coefficient between them is used as the correlation coefficient from the industry. To the industry The strength of direct causality; If the direct causal strength is less than a preset second threshold, then it is determined that the industry... To the industry The causal relationship is an indirect causal relationship and is eliminated; if the direct causal strength is greater than or equal to the second threshold, the direct causal relationship is retained.

5. The method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model with fused causal relationships as described in claim 4, characterized in that, Step S5 is as follows: The industry electricity consumption sequence is set as the reference sequence, and the external influencing factor sequence is set as the comparison sequence; The initial value normalization method is used to perform dimensionless processing on each sequence; Calculate the grey relational coefficient between the sequence of external influencing factors and the sequence of industry electricity consumption at each time point; The average value of the grey relational coefficients at all times is used to obtain the grey relational degree between the external influencing factor and the industry's electricity consumption, which is then used as the correlation strength.

6. The method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model with fused causal relationships as described in claim 5, characterized in that, The construction of input features in step S6 specifically involves: For the target industry to be predicted, the electricity consumption components of the upstream and downstream industries corresponding to the direct causal relationship that was determined to exist and be retained in step S4 in the corresponding frequency band are weighted with the causal strength obtained in step S4 to form the first input part. The components of external influencing factors related to the target industry in the corresponding frequency band are weighted using the correlation strength obtained in step S5 as the weight to form the second input part; The first input portion and the second input portion are combined to form the input feature.

7. The method for predicting electricity consumption in an industrial chain based on a TCN-LSTM model with fused causal relationships as described in claim 6, characterized in that, The TCN-LSTM prediction model described in step S6 is constructed in the following way: The temporal convolutional network module receives the input features and performs feature extraction through a multi-layer structure containing dilated convolutions to capture long sequence dependencies; The Long Short-Term Memory (LSTM) network module receives the output of the temporal convolutional network module and learns a gating mechanism to memorize the long-term temporal dependencies of the time series. The fully connected layer processes the output of the long short-term memory network module to generate the final power consumption prediction value.

8. A power consumption forecasting system for an industrial chain based on a TCN-LSTM model that integrates causal relationships, characterized in that, Used to implement the industrial chain electricity consumption forecasting method as described in any one of claims 1 to 7.

9. A computer device, the computer device comprising a memory, a processor, and a computer program, characterized in that, When the computer program is executed by the processor, it implements the power consumption forecasting method for the industrial chain as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the power consumption forecasting method for the industrial chain as described in any one of claims 1 to 7.

Images