A multi-station fusion model construction method

Abstract

The application discloses a kind of multi-station fusion model construction methods, specifically related to multi-station fusion technical field, including influencing factor data integration, fusion feature based on Transformer multi-station fusion model and the establishment of energy storage cost fluctuation prediction model;The application is extracted by convolutional neural network model to data feature, the sequence form is constructed to the feature after extraction, then sequence is input into Transformer multi-station fusion model and is trained, the features of different modalities are fused to obtain the feature after fusion by attention mechanism, the energy storage cost fluctuation index is obtained by establishing energy storage cost fluctuation prediction model to the feature after fusion, according to energy storage cost fluctuation index numerical value judges the feasibility of energy storage equipment operation, energy storage cost fluctuation prediction model helps real-time monitoring the operating condition of energy storage system, timely discovers potential cost fluctuation risk, helps to guarantee the stability of multi-station fusion system, avoids system failure caused by cost fluctuation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-station fusion, and more particularly to a multi-station fusion model building method. BACKGROUND

[0002] With the passage of time, under the impetus of "new infrastructure", traditional substations are also quietly changing. So-called "multi-station fusion" is to integrate traditional substations, cold and heat supply stations and other "new infrastructure" elements together to form a new type of infrastructure integrating power, cold and heat, and to provide cloud, 5G and charging services for users. The "multi-station fusion" technology is gradually entering people's lives, and it brings people more efficient and more convenient comprehensive energy services. Multi-station fusion can optimally integrate urban resources, and a smart comprehensive energy station integrating substations, photovoltaic stations, energy storage stations, charging stations and data centers can realize comprehensive energy management, solve power supply and demand balance problems, provide comprehensive energy services and realize energy comprehensive and efficient utilization, which has important theoretical and practical significance for promoting China's power grid "source, network, load and storage" integration, developing comprehensive energy internet and building new power grid. However, in the actual application of the multi-station fusion model, due to the great change of external factors, the contradiction between the feasibility and economy of energy storage is prone to occur. In order to solve the above problems, the present application provides a technical solution. SUMMARY

[0003] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a multi-station fusion model building method. The data is feature-extracted by a convolutional neural network model, the extracted features are constructed into a sequence form, and then the sequence is input into a Transformer multi-station fusion model for training. The features of different modalities are fused by an attention mechanism to obtain fused features. A storage cost fluctuation prediction model is established based on the fused features to obtain a storage cost fluctuation index. The feasibility of the operation of the energy storage equipment is determined according to the value of the storage cost fluctuation index. The storage cost fluctuation prediction model is helpful for real-time monitoring of the operation status of the energy storage system, timely discovery of potential cost fluctuation risks, and protection of the stability of the multi-station fusion system to avoid system failure caused by cost fluctuation, thereby solving the problems raised in the above background.

[0004] To achieve the above object, the present application provides the following technical scheme:

[0005] A multi-station fusion model building method comprises the following steps:

[0006] Step one, influence factor data integration: collect the charging efficiency, discharging efficiency, cycle service life, purchase cost, operation cost, maintenance cost, market demand, monthly average solar energy, maturity of energy storage technology, battery cost, raw material cost and energy cost of the energy storage device, and generate an influence factor database;

[0007] Step two, fusion features based on the Transformer multi-station fusion model: features are extracted from the data through a convolutional neural network model, the extracted features are constructed into a sequence form, and then the sequence is input into the Transformer multi-station fusion model for training, different modal features are fused through an attention mechanism to obtain fused features, the fused features are input into a regression task model for task processing, and the model results are evaluated through cross-validation;

[0008] Step three, establish an energy storage cost fluctuation prediction model: an energy storage cost fluctuation prediction model is established based on the fused features, and an energy storage cost fluctuation index is obtained, and the feasibility of the operation of the energy storage device is determined according to the energy storage cost fluctuation index value.

[0009] As a further scheme of the present application, the specific steps of step two, fusion features based on the Transformer multi-station fusion model, are as follows:

[0010] Step S1, data preprocessing: the influence factor database is subjected to data cleaning, abnormal value processing and missing value processing;

[0011] Step S2, feature extraction: different modal data are aligned to have the same dimension through maximum pooling, and then the processed influence factor database is subjected to feature extraction through a convolutional neural network model;

[0012] Step S3, training based on the Transformer multi-station fusion model: the extracted features are constructed into a sequence form, and then the sequence is input into the Transformer multi-station fusion model for training;

[0013] Step S4, feature fusion: different modal features are aligned and the similarity between different modal features is calculated, different modal features are fused through an attention mechanism to obtain fused features, and the fused features are input into a regression task model for task processing;

[0014] Step S5, model evaluation: the Transformer multi-station fusion model is evaluated through cross-validation.

[0015] As a further aspect of the present invention, step S4, feature fusion, aligns features from different modalities and calculates the similarity between features from different modalities. The similarity between features from different modalities is calculated using Manhattan distance, and the smaller the Manhattan distance value, the higher the similarity between features from different modalities. The formula for calculating Manhattan distance is:

[0016] z = |x1-x2|+|y1-y2|;

[0017] In the formula: z is the Manhattan distance, x1 is the x-coordinate of a feature in the standard coordinate system, y1 is the y-coordinate of a feature in the standard coordinate system, x2 is the x-coordinate of another feature in the standard coordinate system, and y2 is the y-coordinate of another feature in the standard coordinate system.

[0018] As a further aspect of the present invention, step S4 feature fusion uses an attention mechanism to fuse features from different modalities to obtain fused features. The inputs of the attention mechanism are the charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, maturity of energy storage technology, battery cost, raw material cost, and energy cost of the energy storage device. The outputs are the operating efficiency, operating cost, financing difficulty, and market demand of the energy storage device.

[0019] As a further aspect of the present invention, the integrated features include the operating efficiency, operating cost, financing difficulty, and market demand of the energy storage device.

[0020] As a further aspect of the present invention, step S4, feature fusion, inputs the fused features into the regression task model for task processing, wherein the formula of the regression task model is:

[0021]

[0022] In the formula: MA is the absolute error, N is the number of features after fusion, and y i The true value of the fused features. These are the predicted values ​​of the fused features.

[0023] As a further aspect of this invention, step three establishes an energy storage cost fluctuation prediction model. This model is built based on the fused features to obtain energy storage cost fluctuation indicators. The fused features include the operating efficiency, operating cost, financing difficulty, and market demand of the energy storage equipment. The specific calculation formula for the energy storage cost fluctuation prediction model is as follows:

[0024]

[0025] In the formula: P MO R is an indicator of energy storage cost fluctuation. MIFor the operating efficiency of energy storage devices, R ME For the operating costs of energy storage devices, R MV Due to the difficulty in financing energy storage equipment, R MQ This reflects the market demand for energy storage equipment.

[0026] As a further aspect of the present invention, step three establishes an energy storage cost fluctuation prediction model to determine the feasibility of operating the energy storage equipment based on the energy storage cost fluctuation index value, specifically as follows:

[0027] When 0≤P MO When the energy storage capacity is less than 30%, the energy storage equipment operates normally and at a low cost.

[0028] When 30≤P MO When the energy storage capacity is less than 50%, the energy storage equipment operates normally but at a high cost.

[0029] When 50≤P MO When the efficiency is less than 70%, the energy storage equipment experiences high operating losses and high costs.

[0030] When 70≤P MO When the energy storage capacity is less than 100%, the energy storage equipment suffers high operating losses and costs, and must be shut down.

[0031] The technical effects and advantages of the multi-site fusion model construction method of the present invention are as follows:

[0032] 1. By predicting energy storage cost fluctuation indicators, this invention can better schedule and allocate energy storage resources in multi-site integrated systems, improve energy utilization efficiency, help reduce overall operating costs, and maximize economic benefits.

[0033] 2. This invention, through an energy storage cost fluctuation prediction model, helps to monitor the operation of energy storage systems in real time, promptly identify potential cost fluctuation risks, help ensure the stability of multi-site integrated systems, and avoid system failures caused by cost fluctuations.

[0034] 3. By establishing an energy storage cost fluctuation prediction model, this invention can help investors more accurately assess the investment value of energy storage projects, thereby making more informed investment decisions, which helps to improve investors' returns and reduce investment risks. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the structure of a multi-station fusion model construction method according to the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0037] A method for building a multi-site fusion model includes the following steps:

[0038] Step 1, Data Integration of Influencing Factors: Collect data on the charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, maturity of energy storage technology, battery cost, raw material cost, and energy cost of energy storage devices to generate a database of influencing factors;

[0039] Step 2, feature fusion based on Transformer multi-station fusion model: feature extraction is performed on the data through convolutional neural network model, the extracted features are constructed into sequence form, and the sequence is then input into Transformer multi-station fusion model for training. The features of different modalities are fused through attention mechanism to obtain fused features. The fused features are then input into regression task model for task processing, and the model results are evaluated through cross-validation.

[0040] Step 3: Establish an energy storage cost fluctuation prediction model: By establishing an energy storage cost fluctuation prediction model based on the fused features, an energy storage cost fluctuation index is obtained, and the feasibility of operating the energy storage equipment is judged based on the value of the energy storage cost fluctuation index.

[0041] Features are extracted from the data using a convolutional neural network model, and the extracted features are constructed into a sequence. This sequence is then input into a Transformer multi-site fusion model for training. The features from different modalities are fused using an attention mechanism to obtain fused features. An energy storage cost fluctuation prediction model is then established based on the fused features to obtain an energy storage cost fluctuation index. The feasibility of energy storage equipment operation is judged based on the value of the energy storage cost fluctuation index. The energy storage cost fluctuation prediction model helps to monitor the operation of the energy storage system in real time, promptly detect potential cost fluctuation risks, help ensure the stability of the multi-site fusion system, avoid system failures caused by cost fluctuations, and help improve the return on investment for investors while reducing investment risks.

[0042] Step two in this embodiment of the invention, based on the fusion features of the Transformer multi-site fusion model, specifically involves the following steps:

[0043] Step S1, Data Preprocessing: This involves cleaning, handling outliers, and removing missing values ​​from the database of influencing factors.

[0044] Step S2, Feature Extraction: Data from different modalities are aligned to have the same dimension by max pooling, and then features are extracted from the processed influencing factor database by a convolutional neural network model.

[0045] Step S3, Training based on the Transformer multi-station fusion model: Construct the extracted features into a sequence, and then input the sequence into the Transformer multi-station fusion model for training;

[0046] Step S4, Feature Fusion: Align the features of different modalities and calculate the similarity between the features of different modalities. Fusion of the features of different modalities is achieved through an attention mechanism to obtain fused features. The fused features are then input into the regression task model for task processing.

[0047] Step S5, Model Evaluation: Evaluate the Transformer multi-site fusion model through cross-validation.

[0048] This study cleans and processes data on energy storage devices, including charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, energy storage technology maturity, battery cost, raw material cost, and energy cost. Outlier and missing value handling are performed. Max pooling is then used to align data from different modalities to ensure they have the same dimensionality. A convolutional neural network model is then used to extract features from the processed database of influencing factors. These extracted features are constructed into sequences and input into a Transformer multi-site fusion model for training. The features from different modalities are aligned, and the similarity between them is calculated. An attention mechanism is then used to fuse these features, resulting in a fused feature set. This fused feature set is input into a regression task model for processing. Finally, cross-validation is used to evaluate the Transformer multi-site fusion model.

[0049] In this embodiment of the invention, step S4, feature fusion, aligns features from different modalities and calculates the similarity between features from different modalities. The similarity between features from different modalities is calculated using Manhattan distance, and the smaller the Manhattan distance value, the higher the similarity between features from different modalities. The formula for calculating Manhattan distance is:

[0050] z = |x1-x2|+|y1-y2|;

[0051] In the formula: z is the Manhattan distance, x1 is the x-coordinate of a feature in the standard coordinate system, y1 is the y-coordinate of a feature in the standard coordinate system, x2 is the x-coordinate of another feature in the standard coordinate system, and y2 is the y-coordinate of another feature in the standard coordinate system.

[0052] The Manhattan distance method calculates the similarity between different modal features, directly measuring the difference between two feature vectors. This difference can be used to assess the similarity between different modal features. Furthermore, it exhibits good robustness to outliers and noise. In practical applications, data may contain noise or outliers; the Manhattan distance method can resist these interfering factors, improving the accuracy of similarity calculations.

[0053] In this embodiment of the invention, step S4 feature fusion uses an attention mechanism to fuse features from different modalities to obtain fused features. The inputs of the attention mechanism are the charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, maturity of energy storage technology, battery cost, raw material cost, and energy cost of the energy storage device. The outputs are the operating efficiency, operating cost, financing difficulty, and market demand of the energy storage device.

[0054] The integrated features in this embodiment of the invention are the operating efficiency, operating cost, financing difficulty, and market demand of the energy storage device.

[0055] In this embodiment of the invention, step S4, feature fusion, inputs the fused features into the regression task model for task processing. The formula for the regression task model is:

[0056]

[0057] In the formula: MA is the absolute error, N is the number of features after fusion, and y i The true value of the fused features. These are the predicted values ​​of the fused features.

[0058] Calculating the absolute error allows for a direct measurement of the difference between the measured result and the true value, without needing to consider the positive or negative relationship between the measured and true values. This helps to more intuitively understand the magnitude of the error and the accuracy of the measurement result. Furthermore, by identifying the difference between the measured result and the true value, it provides guidance for improving measurement methods and enhancing measurement accuracy, which is of great significance for scientific research and engineering practice.

[0059] When calculating energy storage cost fluctuation indicators, the following rules apply to the numerical changes of each indicator:

[0060] When the absolute value of the difference between the operating efficiency of energy storage equipment and that of standard energy storage equipment is within a set threshold range, the absolute value of the operating cost of energy storage equipment and that of standard energy storage equipment is within a set threshold range, the absolute value of the financing difficulty of energy storage equipment and that of standard energy storage equipment is within a set threshold range, and the absolute value of the difference between the market demand of energy storage equipment and that of standard energy storage equipment is large or small, the impact on the energy storage cost fluctuation index is significant.

[0061] When the absolute value of the difference between the market demand for energy storage equipment and the market demand for standard energy storage equipment is within a set threshold range, the absolute value of the operating cost of energy storage equipment and the operating cost of standard energy storage equipment is within a set threshold range, the absolute value of the financing difficulty of energy storage equipment and the financing difficulty of standard energy storage equipment is within a set threshold range, and the absolute value of the difference between the operating efficiency of energy storage equipment and the operating efficiency of standard energy storage equipment is large or small, the impact on the energy storage cost fluctuation index is small.

[0062] When the absolute value of the difference between the market demand for energy storage equipment and the market demand for standard energy storage equipment is within a set threshold range, the absolute value of the difference between the operating efficiency of energy storage equipment and the operating efficiency of standard energy storage equipment is within a set threshold range, the absolute value of the financing difficulty of energy storage equipment and the financing difficulty of standard energy storage equipment is within a set threshold range, and the absolute value of the operating cost of energy storage equipment and the operating cost of standard energy storage equipment is large or small, the impact on the energy storage cost fluctuation index is small.

[0063] When the absolute value of the difference between the market demand for energy storage equipment and the market demand for standard energy storage equipment is within a set threshold range, the absolute value of the difference between the operating efficiency of energy storage equipment and the operating efficiency of standard energy storage equipment is within a set threshold range, the absolute value of the operating cost of energy storage equipment and the operating cost of standard energy storage equipment is within a set threshold range, and the absolute value of the financing difficulty of energy storage equipment and the financing difficulty of standard energy storage equipment is large or small, the impact on the energy storage cost fluctuation index is small.

[0064] Step three in this embodiment of the invention establishes an energy storage cost fluctuation prediction model. This model is built based on the fused features to obtain energy storage cost fluctuation indicators. The fused features include the operating efficiency, operating cost, financing difficulty, and market demand of energy storage equipment. The specific calculation formula for the energy storage cost fluctuation prediction model is as follows:

[0065]

[0066] In the formula: P MO R is an indicator of energy storage cost fluctuation. MI For the operating efficiency of energy storage devices, R ME For the operating costs of energy storage devices, RMV Due to the difficulty in financing energy storage equipment, R MQ This reflects the market demand for energy storage equipment.

[0067] Among them, the energy storage cost fluctuation index of energy storage equipment is related to the square root function of the market demand for energy storage equipment (e is a natural logarithm function), and is related to the fifth root function of the cube of the 23 times operating cost of energy storage equipment, the cube of the financing difficulty of energy storage equipment, and the cube of the logarithm of the operating efficiency of energy storage equipment. The functional relationship reflects that the energy storage cost fluctuation index of energy storage equipment is greatly affected by the market demand for energy storage equipment, and is less affected by the operating efficiency, operating cost, and financing difficulty of energy storage equipment.

[0068] Step three in this embodiment of the invention establishes an energy storage cost fluctuation prediction model to determine the feasibility of operating the energy storage device based on the energy storage cost fluctuation index value. Specifically, it involves:

[0069] When 0≤P MO When the energy storage capacity is less than 30%, the energy storage equipment operates normally and at a low cost.

[0070] When 30≤P MO When the energy storage capacity is less than 50%, the energy storage equipment operates normally but at a high cost.

[0071] When 50≤P MO When the efficiency is less than 70%, the energy storage equipment experiences high operating losses and high costs.

[0072] When 70≤P MO When the energy storage capacity is less than 100%, the energy storage equipment suffers high operating losses and costs, and must be shut down.

[0073] When the multi-site fusion model is used in practice, data from the influencing factor database is first collected. The operating efficiency, operating cost, financing difficulty, and market demand of the energy storage equipment are obtained through the fusion features of the Transformer multi-site fusion model. These four fusion features are then input into the energy storage cost fluctuation prediction model to calculate the energy storage cost fluctuation index. When the energy storage cost fluctuation index is within 30%, the energy storage equipment can operate normally with low operating costs. When the energy storage cost fluctuation index is between 30% and 50%, the energy storage equipment can still operate normally, but the operating cost is higher. When the energy storage cost fluctuation index is between 50% and 70%, the energy storage equipment can still operate normally, but the operating losses are high and the operating cost is high. When the energy storage cost fluctuation index is between 70% and 100%, the energy storage equipment has high operating losses and high costs, and can no longer operate normally, requiring shutdown.

[0074] This invention employs a convolutional neural network (CNN) model to extract features from data. It then performs data cleaning, outlier handling, and missing value processing on data related to energy storage devices, including charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, energy storage technology maturity, battery cost, raw material cost, and energy cost. Max pooling is used to align data from different modalities to ensure they have the same dimensionality. The CNN model then extracts features from the processed database of influencing factors, constructing a sequence of these features. This sequence is then input into a Transformer multi-site fusion model for training. The features from different modalities are aligned, and the similarity between them is calculated. An attention mechanism is used to fuse these features, resulting in a fused feature set. This fused feature set is then input into a regression task model for processing. Finally, cross-validation is used to evaluate the Transformer multi-site fusion model. Subsequently, by establishing a prediction model for energy storage cost fluctuations based on the integrated features, an energy storage cost fluctuation index is obtained. The feasibility of operating energy storage equipment is judged based on the value of the energy storage cost fluctuation index. The energy storage cost fluctuation prediction model helps to monitor the operation status of the energy storage system in real time, promptly identify potential cost fluctuation risks, help ensure the stability of the multi-site integrated system, avoid system failures caused by cost fluctuations, and help improve the return on investment and reduce investment risks.

[0075] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0076] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for building a multi-site fusion model, characterized in that, Includes the following steps: Step 1, Data Integration of Influencing Factors: Collect data on the charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, maturity of energy storage technology, battery cost, raw material cost, and energy cost of energy storage devices to generate a database of influencing factors; Step 2, feature fusion based on Transformer multi-station fusion model: feature extraction is performed on the data through convolutional neural network model, the extracted features are constructed into sequence form, and the sequence is then input into Transformer multi-station fusion model for training. The features of different modalities are fused through attention mechanism to obtain fused features. The fused features are then input into regression task model for task processing, and the model results are evaluated through cross-validation. Step 3: Establish an energy storage cost fluctuation prediction model: An energy storage cost fluctuation prediction model is established based on the integrated features to obtain energy storage cost fluctuation indicators. The integrated features include the operating efficiency, operating cost, financing difficulty, and market demand of energy storage equipment. The specific calculation formula for the energy storage cost fluctuation prediction model is as follows: In the formula: P MO R is an indicator of energy storage cost fluctuation. MI For the operating efficiency of energy storage devices, R ME For the operating costs of energy storage devices, R MV Due to the difficulty in financing energy storage equipment, R MQ The market demand for energy storage equipment; By establishing an energy storage cost fluctuation prediction model based on the fused features, an energy storage cost fluctuation index is obtained. The energy storage cost fluctuation prediction model judges the feasibility of energy storage equipment operation based on the value of the energy storage cost fluctuation index, specifically: When 0≤P MO When the energy storage capacity is less than 30%, the energy storage equipment operates normally and at a low cost. When 30≤P MO When the energy storage capacity is less than 50%, the energy storage equipment operates normally but at a high cost. When 50≤P MO When the efficiency is less than 70%, the energy storage equipment experiences high operating losses and high costs. When 70≤P MO When the energy storage capacity is less than 100%, the energy storage equipment suffers high operating losses and costs, and must be shut down.

2. The method for building a multi-site fusion model according to claim 1, characterized in that, Step two, based on the Transformer multi-site fusion model, involves the following specific steps for feature fusion: Step S1, Data Preprocessing: This involves cleaning, handling outliers, and removing missing values ​​from the database of influencing factors. Step S2, Feature Extraction: Data from different modalities are aligned to have the same dimension by max pooling, and then features are extracted from the processed influencing factor database by a convolutional neural network model. Step S3, Training based on the Transformer multi-station fusion model: Construct the extracted features into a sequence, and then input the sequence into the Transformer multi-station fusion model for training; Step S4, Feature Fusion: Align the features of different modalities and calculate the similarity between the features of different modalities. Fusion of the features of different modalities is achieved through an attention mechanism to obtain fused features. The fused features are then input into the regression task model for task processing. Step S5, Model Evaluation: Evaluate the Transformer multi-site fusion model through cross-validation.

3. The method for building a multi-site fusion model according to claim 2, characterized in that, Step S4, feature fusion, aligns features from different modalities and calculates the similarity between these features. The similarity is calculated using Manhattan distance; a smaller Manhattan distance indicates higher similarity. The formula for calculating Manhattan distance is: z = |x1-x2|+|y1-y2|; In the formula: z is the Manhattan distance, x1 is the x-coordinate of a feature in the standard coordinate system, y1 is the y-coordinate of a feature in the standard coordinate system, x2 is the x-coordinate of another feature in the standard coordinate system, and y2 is the y-coordinate of another feature in the standard coordinate system.

4. The method for building a multi-site fusion model according to claim 2, characterized in that, Step S4 feature fusion uses an attention mechanism to fuse features from different modalities to obtain fused features. The inputs of the attention mechanism are the charging efficiency, discharging efficiency, cycle life, purchase cost, operating cost, maintenance cost, market demand, monthly average solar energy, maturity of energy storage technology, battery cost, raw material cost, and energy cost of the energy storage device. The outputs are the operating efficiency, operating cost, financing difficulty, and market demand of the energy storage device.

5. The method for building a multi-site fusion model according to claim 2, characterized in that, The characteristics of the integration are the operating efficiency, operating cost, financing difficulty, and market demand of energy storage equipment.

6. The method for building a multi-site fusion model according to claim 2, characterized in that, Step S4, feature fusion, inputs the fused features into the regression task model for task processing. The formula for the regression task model is: In the formula: MA is the absolute error, N is the number of features after fusion, and y i The true value of the fused features. These are the predicted values ​​of the fused features.

Images