Multi-energy-storage-inverter grid-connected system SOC balance control method based on neural network

Abstract

The invention relates to the technical field of energy storage inverters, and discloses a multi-energy storage inverter grid-connected system SOC balance control method based on a neural network, and the method comprises the steps: carrying out the real-time learning capability of a GRU-LSTM mixed neural network model for the current working condition and future trend of each energy storage inverter; a charging and discharging power instruction matched with the global state of the system is generated on line, so that on one hand, continuous differentiation of charge states among modules can be actively inhibited, the bus voltage loss risk caused by the fact that a low-SOC module exits in advance is remarkably reduced, and the continuity of system power supply in an off-grid mode is guaranteed; on the other hand, the traditional limitation of self-generation and self-use can be broken through, flexible allocation of photovoltaic energy among modules is achieved, a low-SOC unit can obtain support of a high-SOC unit or surplus photovoltaic power in time, energy idling and waste are avoided, and therefore the energy utilization efficiency and operation reliability of the whole household energy storage system are improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage inverter technology, and in particular to a SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks. Background Technology

[0002] In the actual operation of current home energy storage systems, when multiple energy storage inverters work together in parallel, they often face a series of key issues that urgently need to be addressed:

[0003] (1) System failure risk caused by SOC (State of Charge) imbalance: In a home energy storage system with multiple energy storage inverters operating in parallel, the differences in manufacturing processes, usage time, working environment, and charge / discharge management strategies among the batteries of each inverter module can easily lead to significant deviations in the SOC of different modules. This SOC imbalance problem becomes even more pronounced in off-grid operation mode. When the SOC of some modules in the system is at a low level, these low-SOC modules will be unable to continue working normally due to premature depletion of power. More seriously, due to the shutdown of these modules, the remaining modules may not be able to independently bear the responsibility of maintaining the stability of the bus voltage, which will lead to a power outage of the entire energy storage inverter system, making it impossible for household electrical appliances to be powered normally, seriously affecting the user's electricity experience and electricity safety.

[0004] (2) Obvious limitations of traditional power allocation strategies: In the existing power allocation mechanism of home energy storage systems, when the battery SOC of a certain module in the system is at a very low state, it usually relies solely on the PV (Photovoltaic) system connected to that module for charging. However, this single charging method has many drawbacks. On the one hand, due to the intermittent and unstable nature of photovoltaic power generation, its output power is affected by factors such as light intensity and weather conditions, and may not be able to provide enough power to the low SOC module in a short period of time, resulting in a slow charging speed and a prolonged power restoration time. On the other hand, under sufficient sunlight, if only the low SOC module is allowed to use its own photovoltaic system for charging, while other modules are idle or lightly loaded, it will waste the energy stored in other modules, reduce the energy utilization efficiency of the entire home energy storage system, and fail to fully utilize the advantages of multi-module parallel operation.

[0005] Given the numerous problems exposed by the parallel operation of the aforementioned energy storage inverters, it is imperative to improve and optimize the relevant technologies of existing home energy storage systems. Technological innovation to address issues such as unbalanced State of Charge (SOC) and limitations of traditional power distribution will help improve the stability, reliability, and energy utilization efficiency of home energy storage systems, thus promoting the further development of home energy storage technology.

[0006] The above information is provided as background information only to aid in understanding this disclosure and does not constitute an assertion or admission that any of the above content can be used as prior art relative to this disclosure. Summary of the Invention

[0007] This invention provides a SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks, in order to solve the problems existing in the prior art.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] In a first aspect, the present invention provides a SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network, the method comprising:

[0010] S1. Construct and train the GRU-LSTM hybrid neural network model;

[0011] S2. Apply the GRU-LSTM hybrid neural network model to the real-time operating conditions, and calculate the ideal charging and discharging power of each energy storage inverter online by collecting input parameters.

[0012] S3. Based on the ideal charging and discharging power of each energy storage inverter, achieve SOC balance control of the multi-energy storage inverter grid-connected system.

[0013] Furthermore, in the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method, step S3 includes:

[0014] S31. The ideal charging and discharging power of each energy storage inverter is used as the given value for the battery charging and discharging loop. By adjusting the duty cycle of the DC-DC conversion module, each energy storage inverter is made to operate in an ideal state in real time.

[0015] Furthermore, in the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method, S1 includes:

[0016] S11. Construct a GRU-LSTM hybrid neural network model based on the MATLAB Neural Network Toolbox. The GRU-LSTM hybrid neural network model includes an input layer, a Dropout layer, an LSTM layer, a GRU layer, a fully connected layer, a weight layer, and an output layer.

[0017] S12. Train the GRU-LSTM hybrid neural network model.

[0018] Furthermore, in the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method, step S12 includes:

[0019] S121. In a laboratory environment, several energy storage inverters are set up. The maximum PV power value and the backup load power are kept constant. Record the real-time charging and discharging power of the energy storage inverter's own battery, SOC, real-time PV power, backup load power, system average SOC, and total system PV power at a time of five minutes. These data are used as inputs for training and testing the GRU-LSTM hybrid neural network model.

[0020] S122. Mark the ideal charging and discharging power data of each energy storage inverter and use it as data output for training and testing of the GRU-LSTM hybrid neural network model;

[0021] S123. Randomly divide the acquired data into training data and test data;

[0022] S124. Import the training data into the GRU-LSTM hybrid neural network model and train the GRU-LSTM hybrid neural network model until the GRU-LSTM hybrid neural network model converges.

[0023] S125. Import the test data into the GRU-LSTM hybrid neural network model, verify the converged GRU-LSTM hybrid neural network model, and confirm that the evaluation result of the GRU-LSTM hybrid neural network model has reached the expected target when the maximum error MAX of the GRU-LSTM hybrid neural network model is less than or equal to 5% and the average error MAE is less than or equal to 3%.

[0024] Furthermore, in the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method, after step S121, the method further includes:

[0025] S121.5 Modify the maximum power value of different PVs and the backup load power, and return to perform the experiment described in S121 to obtain more data for training and testing the GRU-LSTM hybrid neural network model.

[0026] Furthermore, in the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method, after step S123, the method further includes:

[0027] S123.5. Map the training data uniformly to the interval [0, 1] to obtain the normalized training data.

[0028] Furthermore, in the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method, step S124 includes:

[0029] S1241. Select the Adam stochastic gradient descent algorithm to update the network weights;

[0030] S1242. Import the training data into the GRU-LSTM hybrid neural network model for a set number of training rounds until the GRU-LSTM hybrid neural network model converges, and set the initial learning rate to 0.0002.

[0031] Furthermore, in the SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks, in step S125, the formula for calculating the maximum error MAX is:

[0032] ;

[0033] The formula for calculating the average error (MAE) is as follows:

[0034] .

[0035] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network as provided in the first aspect above.

[0036] Thirdly, the present invention provides a computer-readable storage medium having computer-executable instructions stored thereon, the computer-executable instructions being executed by a computer processor to implement the neural network-based multi-energy storage inverter grid-connected system SOC equalization control method provided in the first aspect above.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention provides a neural network-based SOC equalization control method for multi-energy storage inverter grid-connected systems. By utilizing the real-time learning capability of the GRU-LSTM hybrid neural network model to study the current operating conditions and future trends of each energy storage inverter, it generates online charging and discharging power commands that match the global state of the system. This proactively suppresses the continuous differentiation of the state of charge among modules, significantly reducing the risk of bus voltage loss caused by the premature exit of low-SOC modules and ensuring the continuity of power supply in off-grid mode. Furthermore, it breaks through the traditional "self-generation and self-consumption" limitation, enabling flexible allocation of photovoltaic energy among modules. This allows low-SOC units to receive timely support from high-SOC units or surplus photovoltaic power, avoiding energy idleness and waste. Consequently, it improves the energy utilization efficiency and operational reliability of the entire home energy storage system, and further enhances the power supply continuity and user experience in home energy storage scenarios.

[0039] The present invention has other features and advantages, which will be apparent from or will be set forth in detail in the accompanying drawings and the following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description

[0040] 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.

[0041] Figure 1 This is a flowchart illustrating a method for SOC equalization control of a multi-energy storage inverter grid-connected system based on neural networks, provided in Embodiment 1 of the present invention.

[0042] Figure 2 This is a topology diagram of several energy storage inverters provided in Embodiment 1 of the present invention;

[0043] Figure 3 This is a data flow diagram of the neural network model provided in Embodiment 1 of the present invention;

[0044] Figure 4 This is a flowchart of the neural network model training process provided in Embodiment 1 of the present invention;

[0045] Figure 5 This is a schematic diagram of the battery charging and discharging loop provided in Embodiment 1 of the present invention;

[0046] Figure 6 This is a schematic diagram of the structure of a computer device provided in Embodiment 2 of the present invention. Detailed Implementation

[0047] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this application in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this application and are therefore intended to limit the scope of protection of this application.

[0048] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this application, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0049] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit this application.

[0050] In the description of this application, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" logical relationship.

[0051] In this application, terms such as “first” and “second” are used only to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy or order relationship between these entities or operations.

[0052] Unless otherwise specified, the use of terms such as “comprising,” “including,” “having,” or other similar expressions in this application is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0053] In this application, expressions such as "greater than", "less than", and "exceeding" are understood to exclude the stated number; expressions such as "above", "below", and "within" are understood to include the stated number. Furthermore, in the description of the embodiments of this application, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times", unless otherwise explicitly specified.

[0054] In the description of the embodiments of this application, the space-related expressions used, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "vertical," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or drawings. They are only for the purpose of describing the specific embodiments of this application or for the reader's understanding, and do not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0055] Unless otherwise expressly specified or limited, the terms "installation," "connection," "linking," "fixing," and "setting," as used in the description of the embodiments of this application, should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two components or the interaction between two components. For those skilled in the art to which this application pertains, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0056] Example 1

[0057] In view of the deficiencies in the existing technology, the applicant, based on years of extensive practical experience and professional knowledge in this field, and in conjunction with the application of theoretical principles, has actively conducted research and innovation in order to create a technology that can overcome the deficiencies in the existing technology. After continuous research, design, and repeated prototype production and improvement, this invention with practical value has finally been created.

[0058] Please refer to Figure 1 This invention provides a SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks. The topology diagram of several energy storage inverters is shown below. Figure 2 As shown, the method includes:

[0059] S1. Construct and train the GRU-LSTM hybrid neural network model.

[0060] It should be noted that GRU (Gated Recurrent Unit) and LSTM (Long Short-Term Memory) are both variants of recurrent neural networks (RNNs) and have the ability to process sequential data. GRU has a relatively simple structure and high computational efficiency, and can capture short-term dependencies in a sequence; LSTM, on the other hand, uses a complex gating mechanism to better handle long-term dependencies in a sequence.

[0061] This embodiment combines the two to form a GRU-LSTM hybrid neural network model, which can give full play to their respective advantages and more accurately learn the operating rules and future trends of energy storage inverters under different operating conditions.

[0062] S2. Apply the GRU-LSTM hybrid neural network model to the real-time operating conditions, and calculate the ideal charging and discharging power of each energy storage inverter online by collecting input parameters.

[0063] It should be noted that during the real-time operation of the grid-connected energy storage inverter system, the system's operating conditions are constantly changing. For example, changes in sunlight intensity will affect photovoltaic power generation, and changes in load will affect the system's power demand. Applying the trained GRU-LSTM hybrid neural network model to the actual system can dynamically adjust the charging and discharging power of the energy storage inverter according to the current real-time operating conditions.

[0064] Specifically, various parameters related to system operation, such as real-time battery charging and discharging power, SOC, and PV power, are collected in real time by sensors installed on the energy storage inverter and related equipment, and used as inputs for the model.

[0065] After the collected input parameters are fed into the trained GRU-LSTM hybrid neural network model, the model calculates the ideal charging and discharging power of each energy storage inverter online based on the rules and patterns it has learned internally. Figure 3 As shown in the figure. Because the model has real-time learning capabilities, it can predict the most suitable charging and discharging power for the system based on current operating conditions and historical data, enabling the system to maintain stable operation under different operating conditions.

[0066] S3. Based on the ideal charging and discharging power of each energy storage inverter, achieve SOC balance control of the multi-energy storage inverter grid-connected system.

[0067] It should be noted that in a multi-energy storage inverter grid-connected system, the State of Charge (SOC) of each module may differ due to various reasons, such as inconsistent battery performance or unreasonable charging and discharging strategies. An unbalanced SOC can lead to decreased system performance and even affect the safe operation of the system. The control objective of this step is to adjust the charging and discharging power of each energy storage inverter to gradually balance the SOC of each module, thereby improving the overall performance and reliability of the system.

[0068] Specifically, based on the ideal charging and discharging power of each energy storage inverter calculated in step S2, the charging and discharging process of the energy storage inverter is adjusted in real time by the control system. For example, for modules with a higher SOC, their charging power can be appropriately reduced or their discharging power increased; for modules with a lower SOC, their charging power can be increased or their discharging power decreased. Through this dynamic adjustment, the SOC of each module gradually reaches a balanced state during operation.

[0069] In one embodiment of this example, step S3 can be further refined to include the following steps:

[0070] S31. The ideal charging and discharging power of each energy storage inverter is used as the given value for the battery charging and discharging loop. By adjusting the duty cycle of the DC-DC conversion module, each energy storage inverter is made to operate in an ideal state in real time.

[0071] It should be noted that this step is the key execution link in step S3 to realize the SOC balance control of the multi-energy storage inverter grid-connected system. It converts the ideal charging and discharging power calculated by the neural network model into actual operable battery charging and discharging control commands. By adjusting the duty cycle of the DC-DC converter module, it allows each energy storage inverter to operate in the ideal state, thereby achieving the goal of SOC balance control.

[0072] Specifically, in an energy storage inverter system, the battery charge / discharge loop (such as...) Figure 5 (As shown) is the core control loop used for precise control of the battery charging and discharging process. The ideal charging and discharging power is calculated in step S2 based on the GRU-LSTM hybrid neural network model and combined with the real-time acquired input parameters. It represents the charging and discharging power value that each energy storage inverter should achieve in order to realize the overall system optimization (such as SOC balance, efficient operation, etc.) under the current operating conditions. Using this ideal charging and discharging power as the setpoint for the battery charging and discharging loop is equivalent to setting a clear target for the entire charging and discharging control process.

[0073] For example, suppose a home energy storage system has three energy storage inverter modules A, B, and C. Based on previous calculations, the ideal charging power of module A is 2kW, the ideal discharging power of module B is 1kW, and the ideal charging power of module C is 1.5kW. These power values ​​will then be used as the setpoints for the battery charging and discharging loops of modules A, B, and C, respectively, and subsequent control operations will revolve around achieving these power values.

[0074] The DC-DC converter module is a key component in an energy storage inverter system, enabling voltage conversion and power regulation. It converts the input DC voltage into DC voltage outputs of different amplitudes. By changing its duty cycle (i.e., the ratio of the on-time of the switching transistor to the switching cycle), the output voltage and current can be precisely controlled, thereby regulating the charging and discharging power of the battery.

[0075] Regulation principle: During battery charging and discharging, the DC-DC converter module compares the ideal charging and discharging power given by the battery charging and discharging loop with the actual detected battery charging and discharging power. If the actual power is less than the ideal power, it means that the battery's charging and discharging current or voltage needs to be increased. At this time, the DC-DC converter module will increase the duty cycle, allowing more energy to flow into or out of the battery. Conversely, if the actual power is greater than the ideal power, the duty cycle will be decreased to limit the battery's charging and discharging energy.

[0076] For example, taking module A as an example, its ideal charging power is 2kW. When the actual detected charging power of module A is 1.8kW, it is less than the ideal value. At this time, the control circuit of the DC-DC converter module will increase the duty cycle of the switching transistor according to this deviation signal, so as to increase the voltage or current input to the battery, thereby increasing the charging power and gradually bringing it closer to the ideal value of 2kW.

[0077] Because the operating conditions of a home energy storage system are constantly changing—for example, changes in sunlight intensity affect photovoltaic power generation, and increases or decreases in load alter the system's power demand—it is necessary to adjust the operating status of each energy storage inverter in real time. By continuously collecting real-time system parameters, recalculating the ideal charging and discharging power, and using this as a new setpoint, the duty cycle of the DC-DC converter module is adjusted in real time. This allows the energy storage inverter to respond promptly to changes in operating conditions and consistently maintain ideal operation.

[0078] Understandably, the ideal state refers to the gradual balancing of the State of Charge (SOC) of each energy storage inverter, while the overall system operates in a highly efficient and stable manner. When each energy storage inverter operates at its ideal charging and discharging power, modules with low SOC receive appropriate charging, and modules with high SOC discharge reasonably. This avoids system performance degradation and safety hazards caused by excessive SOC differences, thereby achieving SOC balancing control in a multi-energy storage inverter grid-connected system and improving the reliability and energy utilization efficiency of the home energy storage system.

[0079] For example, after the system has been running for a period of time, the photovoltaic power generation increases due to enhanced sunlight. At this time, by collecting data in real time and recalculating, it is found that the ideal charging power of module A can be reduced to 1.5kW, the ideal discharging power of module B can be increased to 1.2kW, and the ideal charging power of module C remains unchanged. The system will immediately use these new ideal power values ​​as setpoints and adjust the duty cycle of the DC-DC converter modules so that the charging and discharging power of each module is quickly adjusted to the new ideal values, ensuring that the system always operates in an optimal state.

[0080] In one embodiment of this example, such as Figure 4 As shown, S1 can be further refined to include the following steps:

[0081] S11. Construct a GRU-LSTM hybrid neural network model based on the MATLAB Neural Network Toolbox. The GRU-LSTM hybrid neural network model includes an input layer, a Dropout layer, an LSTM layer, a GRU layer, a fully connected layer, a weight layer, and an output layer.

[0082] It's worth noting that the MATLAB Neural Network Toolbox provides a wealth of mature functions and tools for building, training, and simulating various neural network models. It features a user-friendly interface and a powerful programming interface, making it easy to design complex neural network structures and significantly reducing development difficulty and time.

[0083] The role of each layer in the model:

[0084] Input layer: Responsible for receiving various input data related to the operation of the energy storage inverter, such as the real-time charging and discharging power and SOC of the inverter's own battery, real-time PV power, backup load power, system average SOC, and total system PV power. This data is the foundation for the model's learning and prediction.

[0085] Dropout layer: During training, the output of a subset of neurons is randomly set to zero to prevent overfitting. By introducing randomness, the model avoids over-reliance on certain neurons, thereby improving its generalization ability and enabling it to perform well even with unseen data.

[0086] LSTM layer: Long Short-Term Memory (LSTM) network layers can handle long-term dependencies in sequential data. In energy storage inverter systems, operating data is often a sequence that changes over time. LSTM layers can remember information from a relatively long period of time, helping the model better understand the dynamic changes of the system.

[0087] GRU layer: The gated recurrent unit layer structure is simpler and more computationally efficient than the LSTM layer, while also capturing short-term dependencies in sequence data. When used in conjunction with the LSTM layer, it can fully leverage the advantages of both to learn the features of sequence data more comprehensively.

[0088] Fully connected layer: Integrates and transforms the features extracted from the previous layers, mapping high-dimensional features to a low-dimensional space through linear transformation, in preparation for subsequent output.

[0089] Weight layer: Used to adjust the connection weights between layers. These weights are continuously optimized during training to make the model's output closer to the true value.

[0090] Output layer: Outputs the ideal charging and discharging power of each energy storage inverter. This is the final result that the model wants to predict, providing a basis for subsequent system control.

[0091] S12. Train the GRU-LSTM hybrid neural network model.

[0092] In one embodiment of this example, step S12 can be further refined to include the following steps:

[0093] S121. In a laboratory environment, several energy storage inverters are set up. The maximum PV power and backup load power are kept constant. The real-time charging and discharging power of the energy storage inverter's own battery, SOC, real-time PV power, backup load power, system average SOC, and total system PV power are recorded at a time of five minutes. These data are used as inputs for training and testing the GRU-LSTM hybrid neural network model.

[0094] It should be noted that several energy storage inverters are set up in the laboratory, such as... Figure 1 As shown, keeping the maximum PV power and backup load power constant, relevant data are recorded at five-minute intervals. The purpose of this is to obtain the operating status data of the energy storage inverter at different times under relatively stable external conditions.

[0095] The system records real-time charging and discharging power of the energy storage inverter's own battery, as well as data on SOC, real-time PV power, backup load power, system average SOC, and total system PV power. This data covers the system's input (PV power, backup load power), internal state (battery SOC, charging and discharging power), and overall state (system average SOC, total PV power), comprehensively reflecting the system's operating status and providing rich information for model training.

[0096] S122. Mark the ideal charge and discharge power data of each energy storage inverter and use it as data output for training and testing of the GRU-LSTM hybrid neural network model.

[0097] It should be noted that the ideal charging and discharging power is determined based on the overall operating goals and optimization strategies of the system, and is the target value that the model needs to learn.

[0098] The labeling function is to clarify the prediction target of the model, enabling the model to continuously adjust its parameters during the learning process so that the output result is as close as possible to the ideal charging and discharging power, thereby achieving accurate prediction.

[0099] S123. Randomly divide the acquired data into training data and test data.

[0100] It should be noted that the acquired data is randomly divided into training data and test data. The training data is used to train the model, allowing it to learn the patterns and regularities in the data; the test data is used to verify the model's performance and evaluate its performance on unseen data.

[0101] Random partitioning can avoid human bias in the data, ensure that the training and testing data are representative, and more objectively reflect the training effect and generalization ability of the model.

[0102] S124. Import the training data into the GRU-LSTM hybrid neural network model and train the GRU-LSTM hybrid neural network model until the GRU-LSTM hybrid neural network model converges.

[0103] In one embodiment of this example, step S124 can be further refined to include the following steps:

[0104] S1241. Select the Adam stochastic gradient descent algorithm to update the network weights.

[0105] It should be noted that the Adam stochastic gradient descent algorithm is chosen to update the network weights. The Adam algorithm combines the advantages of momentum gradient descent and adaptive learning rate, and can automatically adjust the learning rate based on the historical gradient information of each parameter. It can converge faster during training and has good robustness.

[0106] Compared to the traditional stochastic gradient descent algorithm, the Adam algorithm can handle non-stationary objective functions and problems with sparse gradients more effectively, and can achieve better results in neural network training.

[0107] S1242. Import the training data into the GRU-LSTM hybrid neural network model for a set number of training rounds until the GRU-LSTM hybrid neural network model converges, and set the initial learning rate to 0.0002.

[0108] It should be noted that this step involves importing the training data into the GRU-LSTM hybrid neural network model for a set number of training rounds, such as 5000 rounds, until the model converges. The selection of the number of training rounds needs to be based on a comprehensive consideration of factors such as model complexity, data size, and training effect.

[0109] The initial learning rate is set to 0.0002. The learning rate controls the step size of parameter updates in each iteration. Setting the initial learning rate too large will cause the model to oscillate and fail to converge during training, while setting it too small will make the training process slow and the convergence time long. 0.0002 is a suitable initial learning rate value that has been verified by experience and practice.

[0110] S125. Import the test data into the GRU-LSTM hybrid neural network model, verify the converged GRU-LSTM hybrid neural network model, and confirm that the evaluation result of the GRU-LSTM hybrid neural network model has reached the expected target when the maximum error MAX of the GRU-LSTM hybrid neural network model is less than or equal to 5% and the average error MAE is less than or equal to 3%.

[0111] It should be noted that the test data is imported into the converged GRU-LSTM hybrid neural network model to validate the model. The maximum error (MAX) and the average error (MAE) are calculated by comparing the model's predicted output with the true values ​​in the test data.

[0112] When the model's maximum error (MAX) is less than or equal to 5% and the mean error (MAE) is less than or equal to 3%, the model's evaluation results are considered to have met the expected goals. This indicates that the model has good prediction accuracy on the test data and can meet the needs of practical applications. If the error exceeds this range, further adjustments to the model structure or training parameters are needed, followed by retraining and validation.

[0113] In one embodiment of this example, after step S121, the method further includes the following steps:

[0114] S121.5 Modify the maximum power value of different PVs and the backup load power, and return to perform the experiment described in S121 to obtain more data for training and testing the GRU-LSTM hybrid neural network model.

[0115] It should be noted that by changing these key parameters, the operation of the energy storage inverter under different operating conditions can be simulated, thereby obtaining more diverse data.

[0116] Significance of data diversity: More data samples enable the model to learn a wider range of system operating patterns, improve the model's adaptability to different working conditions, enhance the model's robustness and generalization ability, and avoid large errors when the model faces unseen working conditions.

[0117] In one embodiment of this example, after step S123, the method further includes the following steps:

[0118] S123.5. Map the training data uniformly to the interval [0, 1] to obtain the normalized training data.

[0119] It should be noted that the physical quantities and value ranges of different features may vary greatly. If normalization is not performed, features with larger value ranges may play a dominant role in the training of the model, while the role of features with smaller value ranges will be ignored.

[0120] Normalization can give all features the same scale, allowing the model to treat all features equally during training, improving training efficiency and convergence speed, and also helping to improve the model's prediction accuracy.

[0121] In one embodiment of this example, in step S125, the formula for calculating the maximum error MAX is:

[0122] ;

[0123] The formula for calculating the average error (MAE) is as follows:

[0124] .

[0125] Although this application uses terms such as neural network model and SOC frequently, the possibility of using other terms is not excluded. These terms are used merely for the convenience of describing and explaining the essence of the invention; interpreting them as any additional limitation would contradict the spirit of the invention.

[0126] This invention provides a neural network-based SOC equalization control method for a multi-energy storage inverter grid-connected system. By utilizing the real-time learning capability of the GRU-LSTM hybrid neural network model to study the current operating conditions and future trends of each energy storage inverter, it generates online charging and discharging power commands that match the global state of the system. This proactively suppresses the continuous differentiation of the state of charge among modules, significantly reducing the risk of bus voltage loss caused by the early exit of low-SOC modules and ensuring the continuity of power supply in off-grid mode. Furthermore, it breaks the traditional limitation of "self-generation and self-consumption," enabling flexible allocation of photovoltaic energy among modules. This allows low-SOC units to receive timely support from high-SOC units or surplus photovoltaic power, avoiding energy idleness and waste. Consequently, it improves the energy utilization efficiency and operational reliability of the entire home energy storage system and further enhances the power supply continuity and user experience in home energy storage scenarios.

[0127] Example 2

[0128] Figure 6 This is a schematic diagram of the structure of a computer device provided in Embodiment 2 of the present invention. Figure 6 A block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present invention is shown. Figure 6 The computer device 12 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0129] like Figure 6 As shown, the computer device 12 is represented in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0130] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0131] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0132] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 6 Not shown; usually referred to as a "hard drive"). Although Figure 6 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0133] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0134] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with the computer device 12, and / or with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although... Figure 6 As not shown, it can be used in conjunction with computer device 12 with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0135] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks provided in this embodiment of the invention.

[0136] Example 3

[0137] Embodiment 3 of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network as provided in all embodiments of the present application.

[0138] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0139] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0140] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0141] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0142] Finally, it should be noted that although the above embodiments have been described in the text and drawings of this application, this should not limit the scope of patent protection of this application. Any technical solutions that are based on the essential concept of this application and utilize the content described in the text and drawings of this application, resulting in equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this application.

Claims

1. A SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks, characterized in that, The method includes: S1. Construct and train the GRU-LSTM hybrid neural network model; S2. Apply the GRU-LSTM hybrid neural network model to the real-time operating conditions, and calculate the ideal charging and discharging power of each energy storage inverter online by collecting input parameters. S3. Based on the ideal charging and discharging power of each energy storage inverter, achieve SOC balance control of the multi-energy storage inverter grid-connected system.

2. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks according to claim 1, characterized in that, S3 includes: S31. The ideal charging and discharging power of each energy storage inverter is used as the given value for the battery charging and discharging loop. By adjusting the duty cycle of the DC-DC conversion module, each energy storage inverter is made to operate in an ideal state in real time.

3. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks according to claim 1, characterized in that, S1 includes: S11. Construct a GRU-LSTM hybrid neural network model based on the MATLAB Neural Network Toolbox. The GRU-LSTM hybrid neural network model includes an input layer, a Dropout layer, an LSTM layer, a GRU layer, a fully connected layer, a weight layer, and an output layer. S12. Train the GRU-LSTM hybrid neural network model.

4. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network according to claim 3, characterized in that, S12 includes: S121. In a laboratory environment, several energy storage inverters are set up. The maximum PV power value and the backup load power are kept constant. Record the real-time charging and discharging power of the energy storage inverter's own battery, SOC, real-time PV power, backup load power, system average SOC, and total system PV power at a time of five minutes. These data are used as inputs for training and testing the GRU-LSTM hybrid neural network model. S122. Mark the ideal charge and discharge power data of each energy storage inverter and use it as the data output for training and testing the GRU-LSTM hybrid neural network model; S123. Randomly divide the acquired data into training data and test data; S124. Import the training data into the GRU-LSTM hybrid neural network model and train the GRU-LSTM hybrid neural network model until the GRU-LSTM hybrid neural network model converges. S125. Import the test data into the GRU-LSTM hybrid neural network model, verify the converged GRU-LSTM hybrid neural network model, and confirm that the evaluation result of the GRU-LSTM hybrid neural network model has reached the expected target when the maximum error MAX of the GRU-LSTM hybrid neural network model is less than or equal to 5% and the average error MAE is less than or equal to 3%.

5. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network according to claim 4, characterized in that, Following S121, the method further includes: S121.5 Modify the maximum power value of different PVs and the backup load power, and return to perform the experiment described in S121 to obtain more data for training and testing the GRU-LSTM hybrid neural network model.

6. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network according to claim 4, characterized in that, Following S123, the method further includes: S123.

5. Map the training data uniformly to the interval [0, 1] to obtain the normalized training data.

7. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks according to claim 4, characterized in that, S124 includes: S1241. Select the Adam stochastic gradient descent algorithm to update the network weights; S1242. Import the training data into the GRU-LSTM hybrid neural network model for a set number of training rounds until the GRU-LSTM hybrid neural network model converges, and set the initial learning rate to 0.0002.

8. The SOC equalization control method for a multi-energy storage inverter grid-connected system based on a neural network according to claim 4, characterized in that, In step S125, the formula for calculating the maximum error MAX is: ； The formula for calculating the average error (MAE) is as follows: 。 9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks as described in any one of claims 1-8.

10. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that, The computer-executable instructions are executed by a computer processor to implement the SOC equalization control method for a multi-energy storage inverter grid-connected system based on neural networks as described in any one of claims 1-8.

Images