Data driving control method and system for multi-input multi-output information fusion of energy storage system

Abstract

The invention discloses a data driving control method and system for multi-input and multi-output information fusion of an energy storage system, and relates to the technical field of stable control of the energy storage system, and the method comprises the steps: S1, carrying out the preprocessing of multi-input and multi-output information data of an energy storage converter in the energy storage system, obtaining a data set, and carrying out the calculation of the data set; wherein the multiple-input-multiple-output information data comprises direct-current side voltage and current, alternating-current side voltage and current and alternating-current side power of the energy storage converter; s2, constructing a data driving model of the energy storage converter based on a GAN network, training the data driving model by using the data set, and judging whether the model meets requirements or not; and S3, based on the trained data driving model, completing control of the energy storage system. According to the invention, the nonlinear mapping capability of the controller can be enhanced, and the dynamic behavior of the energy storage system can be accurately captured.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage system stability control, and more particularly to a data-driven control method and system for energy storage system multi-input and multi-output information fusion. Background Art

[0002] With the rapid development of energy storage technology, energy storage systems (ESS) are increasingly being used in energy systems and have become an indispensable part of the energy transition process. The main function of ESS is to store and release energy to address the imbalance between energy supply and demand. ESS includes various forms such as batteries, pumped hydro, and flywheel energy storage. The use of ESS not only improves energy efficiency but also ensures the stability of power supply. For example, a photovoltaic charging station optimizes the charge and discharge power sequence of the energy storage system by combining photovoltaic output and charging load data, thereby implementing a multi-mode integrated operation strategy.

[0003] Data-driven control (DDC) is an emerging control method that does not rely on a mathematical model of the controlled object. Instead, it designs the controller by analyzing the system's input and output data. This method is particularly suitable for complex systems where precise mathematical models are difficult to establish. Data-driven control can be divided into two categories: data-based control and model-free control. The former first uses data to model the system and then performs control based on this model; the latter directly uses input and output data for control without the need for direct modeling. In recent years, with the development of artificial intelligence and machine learning technologies, data-driven control has become increasingly common in fields such as power systems and industrial automation.

[0004] At present, the control methods of energy storage systems mainly include model-based control and data-driven control. Model-based control methods rely on precise mathematical descriptions of system behavior, but this method often faces modeling difficulties and inaccurate models in practical applications. In contrast, data-driven control methods achieve control by analyzing historical data and have better adaptability and robustness. However, existing patents on data-driven control methods, such as patent CN202310109273.9 and patent CN202210968138.5, mostly focus on simple systems with a single input-output relationship. For complex multi-input and multi-output energy storage systems, how to effectively integrate multi-source heterogeneous data and design an efficient and stable controller remains an urgent problem that those skilled in the art need to solve. Summary of the Invention

[0005] In view of this, the present invention provides a data-driven control method and system for multi-input and multi-output information fusion of an energy storage system, which can not only enhance the nonlinear mapping capability of the controller, but also accurately capture the dynamic behavior of the energy storage system.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A data-driven control method for multi-input and multi-output information fusion of an energy storage system, comprising:

[0008] S1: Preprocessing the multi-input multi-output information data of the energy storage converter in the energy storage system to obtain a data set, wherein the multi-input multi-output information data includes the DC side voltage and current, AC side voltage and current, and AC side power of the energy storage converter;

[0009] S2: constructing a data-driven model of the energy storage converter based on the GAN network, training the data-driven model using the data set, and determining whether the model meets the requirements;

[0010] S3: Complete the control of the energy storage system based on the trained data-driven model.

[0011] Specifically, the S1 includes:

[0012] S11: Collect the DC side input voltage u of the energy storage converter d and input current i d , AC side output voltage u and current i, and calculate the AC side voltage amplitude U based on the AC side output voltage u and current i m , voltage phase angle and the current amplitude I m , current phase angle

[0013] S12: Determine the current iac and current amplitude I on the AC side m , phase angle The relationship between frequencies f is expressed as follows:

[0014]

[0015] S13: Calculate the active power P and reactive power Q of the energy storage converter, and the relationship is as follows:

[0016]

[0017] S14: Determine the relationship between the AC side current iac of the energy storage converter and the active power P and reactive power Q. The expression is as follows:

[0018]

[0019] Specifically, the S1 further includes:

[0020] S15: Generate a complete set of data from the data set of AC side current, AC side voltage, active power and reactive power containing missing values;

[0021] S16: Generate a random sample of missing values ​​and then perform multiple difference filling;

[0022] S17: Merge the data of AC side current, AC side voltage, active power and reactive power into a set of data storage;

[0023] S18: Use the Z-score normalization method to normalize the AC side current, AC side voltage, active power, and reactive power. The formula is as follows:

[0024]

[0025] Among them, X is the original data, μ is the mean of the feature, σ is the standard deviation of the feature, X scaled is the normalized data;

[0026] S19: The normalized processing results obtained in S18 are formed into a data set, and the data set is divided into a training data set and a verification data set.

[0027] Specifically, the S2 includes:

[0028] S21: Build a data-driven model, where the data-driven model includes a generation network G(x) and a discrimination network D(x), using noise z for random sampling, inputting the G network, generating new data G(z) and probability distribution P g (G(z));

[0029] The new data G(z) is used to determine the authenticity of the missing value in S16, and the probability distribution P g (G(z)) creates a random data set to train the random samples in s16; when the random sample is identified as true by the identification network D(x), the current random sample is output; when it is identified as false, a new random sample is generated until the data in the data set is identical to the real data, and the current random sample is output;

[0030] S22: Determine the value function V(D,G) of the data-driven model, which is expressed as follows:

[0031]

[0032] Where E is the expectation, The expectation of sampling current, voltage and power data realistically; is the expectation of the data generated by the generative network G(x); Pdata (x) is the probability distribution of the data set; P g (x) is the probability distribution of the data set generated by the generator network G(x); x is the real sample vector; z is the random noise data; D(x) is the discriminator network;

[0033] S23: Train the identification network D(x), the expression is as follows:

[0034]

[0035] S24. Train the generative network G(x), which is expressed as follows:

[0036]

[0037] S25: Output the judgment category through the Sigmoid function;

[0038] S26: Optimize the loss function of the identification network D(x) and the generation network G(x), and adjust the generation network G(x) according to the identification network D(x) until the identification network D(x) and the generation network G(x) meet any one of convergence, optimal value, or a preset number of iterations.

[0039] Specifically, the control process of the energy storage system includes:

[0040] Collect AC bus voltage signal u d , determine the grid phase through a phase-locked loop;

[0041] Based on the calculated active power P and reactive power Q of the energy storage converter, the outer loop DC voltage control and active power control are realized;

[0042] The AC current output by the energy storage converter is obtained through a data-driven model, and the coordinate transformation is performed and output to the current inner loop controller;

[0043] Through the pulse width modulation method, the switching signal is output to control the operation of the energy storage system.

[0044] Specifically, a system for a data-driven control method for multi-input and multi-output information fusion of an energy storage system includes:

[0045] A preprocessing module: preprocesses the multi-input multi-output information data of the energy storage converter in the energy storage system to obtain a data set, wherein the multi-input multi-output information data includes the DC side voltage and current, AC side voltage and current, and AC side power of the energy storage converter;

[0046] Model construction module: constructing a data-driven model of the energy storage converter based on the GAN network, training the data-driven model using the data set, and determining whether the model meets the requirements;

[0047] System control module: Based on the trained data-driven model, complete the control of the energy storage system.

[0048] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a data-driven control method and system for multi-input and multi-output information fusion of an energy storage system, which has the following beneficial effects:

[0049] (1) Using a data-driven approach to model the energy storage system can not only enhance the nonlinear mapping capability of the controller, but also accurately capture the dynamic behavior of the energy storage system.

[0050] (2) Data-driven modeling can be directly based on actual operating data (such as current, voltage, and power) without the need for complex mathematical expressions, making the energy storage system closer to the actual operating state.

[0051] (3) Generative Adversarial Networks (GANs), which consist of a generator and a discriminator, can generate high-quality synthetic data. The generator is responsible for generating data, while the discriminator is responsible for distinguishing true from false data. This is particularly useful when real data is difficult to obtain or expensive. Synthetic data can be used to enhance the training set and improve the generalization ability of the model.

[0052] (4) Using a GAN network, adversarial training between the generator and the discriminator during training can improve the model's robustness to abnormal data and noise; in energy storage systems, this means that the model can better handle sensor noise and uncertainty in operation.

[0053] (5) The use of GAN network can achieve self-supervision of the training process without the need for additional label information; it can reduce the dependence on large amounts of labeled data and reduce the cost and workload of data preparation; in the modeling of energy storage systems, unlabeled operating data can be used to train the model to obtain actual economic benefits.

[0054] (6) The GAN network can not only generate synthetic data, but also be used for feature extraction; in energy storage systems, GAN can be used to identify and extract key features that have an important impact on system performance, thereby improving the effectiveness of control strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0056] Figure 1Flowchart of the data-driven control method for multi-input and multi-output information fusion of the energy storage system provided by the present invention;

[0057] Figure 2 This is a diagram of the structure of the multi-input and multi-output information fusion data generation adversarial network for the energy storage converter provided by the present invention;

[0058] Figure 3 A data-driven control block diagram of multi-input and multi-output information fusion for the energy storage system provided by the present invention;

[0059] Figure 4 This is a flow chart of the multi-input and multi-output information fusion data-driven training of the energy storage converter provided by the present invention. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] The embodiment of the present invention discloses a data-driven control method for multi-input and multi-output information fusion of an energy storage system, comprising:

[0062] S1: Preprocessing the multi-input multi-output information data of the energy storage converter in the energy storage system to obtain a data set, wherein the multi-input multi-output information data includes the DC side voltage and current, AC side voltage and current, and AC side power of the energy storage converter;

[0063] S2: constructing a data-driven model of the energy storage converter based on the GAN network, training the data-driven model using the data set, and determining whether the model meets the requirements;

[0064] S3: Based on the trained data-driven model, complete the control of the energy storage system. Figure 1 .

[0065] Said S1 comprises:

[0066] S11: Collect the DC side input voltage u of the energy storage converter d and input current i d , AC side output voltage u and current i, and calculate the AC side voltage amplitude U based on the AC side output voltage u and current i m , voltage phase angle and the current amplitude I m , current phase angle

[0067] S12: Determine the current iac and current amplitude I on the AC side m , phase angle The relationship between frequencies f is expressed as follows:

[0068]

[0069] S13: Calculate the active power P and reactive power Q of the energy storage converter, and the relationship is as follows:

[0070]

[0071] S14: Determine the relationship between the AC side current iac of the energy storage converter and the active power P and reactive power Q. The expression is as follows:

[0072]

[0073] Furthermore, the S1 further includes:

[0074] S15: Generate a complete set of data from the data set of AC side current, AC side voltage, active power and reactive power containing missing values;

[0075] S16: Perform multiple operations to generate a random sample of missing values, and then perform multiple interpolation;

[0076] S17: Merge the data of AC side current, AC side voltage, active power and reactive power into a set of data storage;

[0077] S18: Use the Z-score normalization method to normalize the AC side current, AC side voltage, active power, and reactive power. The formula is as follows:

[0078]

[0079] Among them, X is the original data, μ is the mean of the feature, σ is the standard deviation of the feature, X scaled is the normalized data;

[0080] S19: The normalized processing results obtained in S18 are formed into a data set, and the data set is divided into a training data set and a verification data set.

[0081] Preferably, the S2 includes:

[0082] S21: Build a data-driven model, where the data-driven model includes a generation network G(x) and a discrimination network D(x), using noise z for random sampling, inputting the G network, generating new data G(z) and probability distribution P g (G(z));

[0083] The new data G(z) is used to determine the authenticity of the missing value in S16, and the probability distribution P g (G(z)) creates a random data set to train the random samples in s16; when the random sample is identified as true by the identification network D(x), the current random sample is output; when it is identified as false, a new random sample is generated until the data in the data set is identical to the real data, and the current random sample is output;

[0084] S22: Determine the value function V(G,D) of the data-driven model, which is expressed as follows:

[0085]

[0086] Where E is the expectation, The expectation of sampling current, voltage and power data realistically; is the expectation of the data generated by the generative network G(x); P data (x) is the probability distribution of the data set; P g (x) is the probability distribution of the data set generated by the generator network G(x); x is the real sample vector; z is the random noise data; D(x) is the discriminator network;

[0087] S23: Train the discriminator network D(x) to classify the labels of the training data. The expression is as follows:

[0088]

[0089] S24. Train the generative network G(x), which is expressed as follows:

[0090]

[0091] S25: Output the judgment category through the Sigmoid function;

[0092] S26: Optimize the loss function of the identification network D(x) and the generation network G(x), and adjust the generation network G(x) according to the identification network D(x) until the identification network D(x) and the generation network G(x) meet any one of convergence, optimal value, or a preset number of iterations.

[0093] Preferably, Figure 2 : is a GAN network structure diagram. In an embodiment of the present invention, the energy storage converter data-driven model is constructed based on the GAN network.

[0094] For further information, see Figure 3-4 , the control process of the energy storage system includes:

[0095] Collect AC bus voltage signal u d , determine the grid phase through a phase-locked loop;

[0096] Outer loop DC voltage control and active power control;

[0097] The output current of the energy storage converter is obtained through data driving, and the coordinate transformation is performed and output to the current inner loop controller;

[0098] Through the pulse width modulation (PWM) method, the switching signal is output to control the operation of the energy storage inverter.

[0099] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0100] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data-driven control method for multi-input and multi-output information fusion of an energy storage system, characterized in that: include: S1: preprocessing the multi-input multi-output information data of the energy storage converter in the energy storage system to obtain a data set, wherein the multi-input multi-output information data includes the DC side voltage and current, AC side voltage and current, and AC side power of the energy storage converter; S2: constructing a data-driven model of the energy storage converter based on the GAN network, training the data-driven model using the data set, and determining whether the model meets the requirements; S3: Complete the control of the energy storage system based on the trained data-driven model.

2. A data driven control method suitable for multi-input and multi-output information fusion of energy storage system according to claim 1, characterized in that: The S1 includes: S11: Collect the DC side input voltage u of the energy storage converter d and input current i d , the AC side output voltage u and current i, and the AC side voltage amplitude U is calculated based on the AC side output voltage u and current i m , voltage phase angle and the current amplitude I m , current phase angle S12: Determine the current i on the AC side ac With the current amplitude I m , Phase Angle The relationship between the frequencies f is expressed as follows: S13: Calculate the active power P and reactive power Q of the energy storage converter, and the relationship is as follows: S14: Determine the AC side current i of the energy storage converter ac The relationship between active power P and reactive power Q is as follows:

3. A data driven control method suitable for multi-input and multi-output information fusion of energy storage system according to claim 2, characterized in that: The S1 further comprises: S15: Generate a complete set of data from the data sets of AC side current, AC side voltage, active power and reactive power containing missing values; S16: Generate a random sample of missing values ​​and then perform multiple difference filling; S17: combining the data of the AC side current, the AC side voltage, the active power and the reactive power into a set of data storage; S18: The AC side current, AC side voltage, active power and reactive power are normalized using the Z-score standardization method, and the formula is as follows: Among them, X is the original data, μ is the mean of the feature, σ is the standard deviation of the feature, X scaled is the normalized data; S19: The normalized processing result obtained in S18 is used to form a data set, and the data set is divided into a training data set and a verification data set.

4. A data driven control method suitable for multi-input and multi-output information fusion of energy storage system according to claim 1, characterized in that: The S2 includes: S21: Build a data-driven model, where the data-driven model includes a generation network G(x) and a discrimination network D(x), use noise z to perform random sample expansion, input G network, generate new data G(z) and probability distribution P g (G(z)); The new data G(z) is used to determine the authenticity of the missing value in S16. The probability distribution P g (G(z)) establishes a random data set to train the random samples in s16; when the random sample is identified as true by the identification network D(x), the current random sample is output; when it is identified as false, a new random sample is generated until the data in the data set is the same as the real data, and the current random sample is output; S22: Determine the value function V(D,G) of the data-driven model, which is expressed as follows: In the formula, E is the expectation, The expectation of sampling current, voltage and power data realistically; is the expectation of the data generated by the generative network G(x); P data (x) is the probability distribution of the data set; P g (x) is the probability distribution of the data set generated by the generating network G(x); x is the real sample vector; z is the random noise data; D(x) is the discriminator network; S23: Train the discriminator network D(x) to enable it to classify the labels of the training data, as expressed as follows: S24. The expression of training generation network G(x) is as follows: S25: Output the determination category through the Sigma function, S26: Optimize the loss functions of the identification network D(x) and the generation network G(x), and adjust the generation network G(x) according to the identification network D(x) until the identification network D(x) and the generation network G(x) meet any one of reaching convergence, reaching the optimal value, or reaching a preset number of iterations.

5. A data driven control method suitable for multi-input and multi-output information fusion of energy storage system according to claim 1, characterized in that: The control process of the energy storage system includes: Collect AC bus voltage signal u d , determine the grid phase through a phase-locked loop; According to the calculated active power P and reactive power Q of the energy storage converter, the outer loop DC voltage control and active power control are realized; The AC current output by the energy storage converter is obtained through the data-driven model, and the coordinate transformation is performed and output to the current inner loop controller; Through the pulse width modulation method, the switching signal is output to control the operation of the energy storage system.

6. A system for a data-driven control method for multi-input and multi-output information fusion of an energy storage system as claimed in any one of claims 1 to 5, characterized in that: include: A preprocessing module: preprocessing the multi-input multi-output information data of the energy storage converter in the energy storage system to obtain a data set, wherein the multi-input multi-output information data includes the DC side voltage and current, AC side voltage and current, and AC side power of the energy storage converter; Model construction module: constructing a data-driven model of the energy storage converter based on the GAN network, training the data-driven model using the data set, and determining whether the model meets the requirements; System control module: Based on the trained data-driven model, complete the control of the energy storage system.

Images