Physical information neural network controller implementation method for hybrid energy storage system and related device

Abstract

The invention relates to the technical field of physical information neural network training, in particular to a hybrid energy storage system-oriented physical information neural network controller implementation method and a related device. The method comprises the following steps: acquiring a plurality of sampling data sets acquired by aiming at the hybrid energy storage system; iterative training is conducted on the physical information neural network controller through the multiple sampling data sets, the trained physical information neural network controller is obtained, and training of each iteration round is conducted through a single sampling data set. Through the physical information neural network controller obtained after training, the prediction accuracy of the predicted duty ratio of the battery side and the predicted duty ratio of the super-capacitor side can be improved, and then the hybrid energy storage system is controlled more efficiently.

Description

Technical Field

[0001] This application relates to the field of physical information neural network training technology, and in particular to a method and related apparatus for implementing a physical information neural network controller for hybrid energy storage systems. Background Technology

[0002] In microgrid operation, energy storage systems are the core component for power regulation and energy buffering. Hybrid Energy Storage Systems (HESS) integrate different types of energy storage units to achieve complementary and synergistic energy and power densities, effectively addressing the performance limitations of single energy storage systems and providing reliable support for stable operation and optimized scheduling of microgrids. However, random fluctuations in local loads can easily lead to unstable DC bus voltage and power supply-demand imbalances, posing a severe challenge to the power quality of HESS. Therefore, the performance of HESS depends on efficient control strategies. However, existing solutions struggle to achieve efficient control of HESS, resulting in poor performance and low energy efficiency. Summary of the Invention

[0003] To address the aforementioned issues, this application provides a method and related apparatus for implementing a physical information neural network controller for hybrid energy storage systems, which can efficiently control hybrid energy storage systems.

[0004] According to one aspect of the embodiments of this application, a method for implementing a physical information neural network controller for a hybrid energy storage system is proposed, applied to a physical information neural network controller, wherein the physical information neural network controller includes a first fully connected neural network and a second fully connected neural network, and the method includes: Obtain multiple sample datasets collected for the hybrid energy storage system; The physical information neural network controller is iteratively trained using the multiple sampled datasets to obtain the trained physical information neural network controller. Each iteration of training is performed using a single sampled dataset. The training in a single iteration includes the following steps: Determine the label validation set corresponding to the sampled dataset; The sampled dataset is input into the first fully connected neural network to obtain a first output value output by the first fully connected neural network. The first output value includes the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side of the hybrid energy storage system. The first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side are determined based on the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side. The sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side are input into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network. Based on the first base duty cycle, the first modified duty cycle, the second base duty cycle, and the second modified duty cycle, determine the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side; The loss value of the physical information neural network controller in the current iteration is determined based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set, so as to train the next iteration based on the loss value of the physical information neural network controller in the current iteration.

[0005] In the above scheme, the sampling dataset includes the DC bus voltage of the hybrid energy storage system, the output current on the battery side, the output current on the supercapacitor side, the inductor impedance parameters on the battery side, and the inductor impedance parameters on the supercapacitor side; determining the first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side based on the predicted current reference values ​​on the battery side and the predicted current reference values ​​on the supercapacitor side includes: The first base duty cycle is determined based on the DC bus voltage of the hybrid energy storage system, the output current of the battery side, the inductance parameters of the battery side, and the predicted current reference value of the battery side. The second base duty cycle is determined based on the DC bus voltage of the hybrid energy storage system, the output current of the supercapacitor side, the inductance parameters of the supercapacitor side, and the predicted current reference value of the supercapacitor side.

[0006] In the above scheme, the step of inputting the sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network includes: The predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side are concatenated as features into the sampled dataset to obtain the sampled dataset after feature fusion. The sampled dataset after feature fusion is input into the second fully connected neural network to obtain the first corrected duty cycle and the second corrected duty cycle.

[0007] In the above scheme, determining the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side based on the first base duty cycle, the first corrected duty cycle, the second base duty cycle, and the second corrected duty cycle includes: The first prediction duty cycle is determined based on the preset feature fusion coefficient, the first base duty cycle, and the first modified duty cycle; The second predicted duty cycle is determined based on the preset feature fusion coefficient, the second base duty cycle, and the second modified duty cycle.

[0008] In the above scheme, determining the loss value of the physical information neural network controller in the current iteration based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set includes: The loss value of the physical information neural network controller in the current iteration is determined based on the difference between the first predicted duty cycle and the first actual duty cycle, and the difference between the second predicted duty cycle and the second actual duty cycle.

[0009] In the above scheme, determining the loss value of the physical information neural network controller in the current iteration based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set includes: Determine the predicted DC bus voltage, the predicted output current on the battery side, and the predicted output current on the supercapacitor side of the hybrid energy storage system; The difference between the predicted DC bus voltage of the hybrid energy storage system and the actual DC bus voltage in the tag verification set is taken as the first difference. The difference between the predicted output current value on the battery side and the actual output current value on the battery side in the tag verification set is used as the second difference. The difference between the predicted output current value of the supercapacitor side and the actual output current value of the supercapacitor side in the tag verification set is used as the third difference value. The difference between the first predicted duty cycle and the first actual duty cycle is taken as the fourth difference. The difference between the second predicted duty cycle and the second actual duty cycle is taken as the fifth difference. The first difference, the second difference, the third difference, the fourth difference, and the fifth difference are used as the loss values ​​of the physical information neural network controller in the current iteration.

[0010] According to one aspect of the embodiments of this application, a physical information neural network controller implementation device for hybrid energy storage systems is proposed, applied to a physical information neural network controller, wherein the physical information neural network controller includes a first fully connected neural network and a second fully connected neural network, and the device includes: The acquisition unit is used to acquire multiple sampled datasets collected for the hybrid energy storage system; The training unit is used to iteratively train the physical information neural network controller using the multiple sampled datasets to obtain the trained physical information neural network controller. Each iteration of training is performed using a single sampled dataset. The training in a single iteration includes the following steps: Determine the label validation set corresponding to the sampled dataset; The sampled dataset is input into the first fully connected neural network to obtain a first output value output by the first fully connected neural network. The first output value includes the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side of the hybrid energy storage system. The first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side are determined based on the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side. The sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side are input into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network. Based on the first base duty cycle, the first modified duty cycle, the second base duty cycle, and the second modified duty cycle, determine the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side; The loss value of the physical information neural network controller in the current iteration is determined based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set, so as to train the next iteration based on the loss value of the physical information neural network controller in the current iteration.

[0011] According to one aspect of the embodiments of this application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the physical information neural network controller implementation method for hybrid energy storage systems as described above.

[0012] According to one aspect of the embodiments of this application, a computer program product is provided, the computer program product including a computer program, the computer program being read and executed by a processor of an electronic device, causing the electronic device to execute the physical information neural network controller implementation method for hybrid energy storage systems as described above.

[0013] The beneficial effects of this application are as follows: This application constructs a physical information neural network controller with innovative logic, utilizing its first and second fully connected neural networks for data processing. Specifically, its implementation logic is as follows: By acquiring multiple sampling datasets collected for the hybrid energy storage system, and using these collected datasets for multiple iterations of training, a trained physical information neural network controller can be obtained. Each iteration of training calculates the loss value for the current iteration by determining the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side, combined with the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side from the tag verification set. The next iteration can then be trained based on the loss value. Therefore, the physical information neural network controller obtained by this application after training can improve the prediction accuracy of the predicted duty cycle on the battery side and the supercapacitor side, thereby enabling more efficient control of the hybrid energy storage system. Attached Figure Description

[0014] Figure 1 This is a system architecture diagram of the physical information neural network controller implementation method for hybrid energy storage systems provided in this application embodiment. Figure 2 A flowchart illustrating the implementation method of the physical information neural network controller for a hybrid energy storage system provided in this application embodiment; Figure 3 This is a structural diagram of a hybrid energy storage system provided in an embodiment of this application; Figure 4 An architecture diagram of the physical information neural network controller provided in the embodiments of this application; Figure 5 A block diagram of a physical information neural network controller implementation device for hybrid energy storage systems provided in this application embodiment; Figure 6 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a server provided in an embodiment of this application. Detailed Implementation

[0015] To enable those skilled in the art to better understand the solutions of this application, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] It should be noted that while some processes described in the specification, claims, and accompanying drawings include multiple steps appearing in a specific order, it should be clearly understood that these steps may not be performed in the order they appear herein, or may be performed in parallel. The step numbers are merely used to distinguish different steps and do not themselves represent any execution order. Furthermore, descriptions such as "first," "second," or "objective" in this document are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. "Multiple" in this document refers to at least two.

[0017] It is worth noting that in the specific embodiments of this application, sampling datasets, label verification sets, and other related data are involved. When the above embodiments of this application are applied to specific products or technologies, permission or consent from the target object is required, and the collection, use, and processing of related data must comply with relevant laws, regulations, and standards. For example, when an embodiment of this application needs to obtain sampling datasets, label verification sets, and other related data, separate permission or consent from the target object can be obtained through pop-ups or redirection to a confirmation page. After obtaining the separate permission or consent from the target object, the necessary sampling datasets, label verification sets, and other related data for the normal operation of the embodiments of this application can then be obtained.

[0018] Please see Figure 1 , Figure 1 This is a system architecture diagram of the multi-vehicle platooning method provided in this application embodiment. It includes a terminal 140, an Internet connection 130, a gateway 120, a server 110, etc.

[0019] Terminal 140 can take various forms, including desktop computers, laptops, PDAs (personal digital assistants), mobile phones, vehicle terminals, and dedicated terminals. Furthermore, it can be a single device or a collection of multiple devices. For example, multiple desktop computers can be interconnected via a local area network, sharing a single monitor to work collaboratively, forming a single terminal 140. Terminal 140 can communicate with the Internet 130 via wired or wireless means to exchange data.

[0020] Server 110 refers to a computer system capable of providing certain services to terminal 140. Compared to ordinary terminal 140, server 110 has higher requirements in terms of stability, security, and performance. Server 110 can be a single high-performance computer in a network platform, a cluster of multiple high-performance computers, a portion of a single high-performance computer (e.g., a virtual machine), or a combination of portions of multiple high-performance computers (e.g., virtual machines). Server 110 can also communicate with the Internet 130 via wired or wireless means to exchange data.

[0021] Gateway 120, also known as an internetwork connector or protocol converter, is a computer system or device that acts as a translator, enabling network interconnection at the transport layer. It bridges the gap between two systems using different communication protocols, data formats, languages, or even completely different architectures. Gateways can also provide filtering and security functions. Messages sent from terminal 140 to server 110 are forwarded to the corresponding server 110 via gateway 120. Messages sent from server 110 to terminal 140 are also forwarded to the corresponding terminal 140 via gateway 120.

[0022] The following provides a detailed description of the specific implementation methods of the embodiments of this application: Please see Figure 2 , Figure 2 This is a flowchart illustrating the implementation method of the physical information neural network controller for hybrid energy storage systems provided in this application embodiment. Figure 2 The illustrated implementation method of the physical information neural network controller for hybrid energy storage systems includes: Step 210: Obtain multiple sampling datasets collected for the hybrid energy storage system; Step 220: Iteratively train the physical information neural network controller using the multiple sampled datasets to obtain the trained physical information neural network controller. The training of each iteration round is performed using a single sampled dataset. The training in a single iteration includes the following steps: Determine the label validation set corresponding to the sampled dataset; The sampled dataset is input into the first fully connected neural network to obtain a first output value output by the first fully connected neural network. The first output value includes the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side of the hybrid energy storage system. The first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side are determined based on the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side. The sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side are input into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network. Based on the first base duty cycle, the first modified duty cycle, the second base duty cycle, and the second modified duty cycle, a first predicted duty cycle corresponding to the battery side and a second predicted duty cycle corresponding to the supercapacitor side are determined. Based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set, the loss value of the physical information neural network controller in the current iteration round is determined, so as to train the next iteration round based on the loss value of the physical information neural network controller in the current iteration round.

[0023] The following is a detailed explanation of steps 210-220: In step 210, the multiple sampling datasets collected from the hybrid energy storage system can be used... Let k represent the sampled dataset x collected at sampling time k. It can be used to represent the collected dataset of a hybrid energy storage system at multiple sampling times.

[0024] In step 220, the acquired dataset is used for training through multiple iterations to obtain the trained physical information neural network controller. Each iteration calculates the loss value by determining the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side, combined with the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side from the label verification set. The next iteration can then be trained based on this loss value. Therefore, the physical information neural network controller obtained after training in this application can improve the prediction accuracy of the predicted duty cycle on both the battery and supercapacitor sides, thereby enabling more efficient control of the hybrid energy storage system.

[0025] Please refer to Figure 3 As shown, Figure 3 This is an architecture diagram of a hybrid energy storage system, also known as HESS. It illustrates the HESS topology, which includes a hybrid energy storage system composed of batteries and supercapacitors, along with bidirectional DC-DC converters and DC loads. The batteries and supercapacitors are connected to the DC bus via their respective bidirectional DC-DC converters. and These are the output voltages on the battery side and the supercapacitor side, respectively. and These are the output currents on the battery side and the supercapacitor side, respectively, and also the inductor currents of their corresponding bidirectional DC-DC converters. and These are the inductors for the bidirectional DC-DC converters on the battery side and the supercapacitor side, respectively. and It is the impedance of a non-ideal inductor. and It is the MOSFET (Metal Oxide Semiconductor Field Effect Transistor) of the battery-side converter, and its control signal and Complementary duty cycle signals and It is modulated using pulse width modulation (PWM) technology, in which... .

[0026] Similarly, and It is the MOSFET of the converter on the supercapacitor side, and the corresponding control signal is and Duty cycle is and C is a filter capacitor connected in parallel to the DC bus, and the voltage of the DC bus is... R and These are the load resistance and load current of HESS, respectively.

[0027] Based on the circuit topology of HESS, the following differential equation can be obtained: ; Through a bidirectional DC-DC converter, the hybrid energy storage system can discharge and charge, and its state is determined by the inductor current. Decision, in which subscript ,when When the signal is positive, energy flows from the hybrid energy storage system to the DC bus; conversely, when the signal is negative, the hybrid energy storage system is charged.

[0028] The Physics-Informed Neural Network (PINN) controller requires offline training using labeled data before being applied to a power electronic converter. It uses sensors to collect voltage and current information from the hybrid energy storage system's circuitry as input (the collected dataset), and then infers online the duty cycle of the switching MOSFETs (i.e., the first predicted duty cycle for the battery side and the second predicted duty cycle for the supercapacitor side) as output, thereby achieving closed-loop control of the power electronic converter (hybrid energy storage system). The PINN controller, also known as the physical information neural network controller, incorporates prior physical knowledge while continuing data-driven design. As a supervised learning method, its training process requires a certain amount of labeled data as labels, i.e., the label verification set described in this application. In this application, the reference voltage is measured... With bus voltage The voltage deviation between the two sides is used to generate a total current reference value by the voltage loop controller, and the current reference values ​​for the battery side and the supercapacitor side are obtained through power distribution. and Then, the current loop controller tracks the current reference value to generate the signal duty cycle of the corresponding converter MOSFETs on the battery side and the supercapacitor side. (First Real Duty Cycle) and (Second true duty cycle), and modulated into a control signal through PWM technology to achieve the control purpose. Therefore, the purpose of this application is to obtain the first predicted duty cycle corresponding to the battery side as close as possible to the first true duty cycle through physical information neural network controller reasoning, and the second predicted duty cycle corresponding to the supercapacitor side as close as possible to the second true duty cycle.

[0029] like Figure 4 As shown, Figure 4 This is the architecture diagram of the physical information neural network controller proposed in this application. fnn1 is the first fully connected neural network, and fnn2 is the second fully connected neural network. This application uses the Simulink platform in MATLAB to simulate and obtain the dataset required by the PINN controller. Active Disturbance Rejection Control (ADRC) is used as the voltage loop controller, a low-pass filter (LPF) as the power distribution module, and a PI controller as the current loop controller. The discrete data acquisition step size is [missing information]. .

[0030] Collect the input vector of the dataset ,Label , can be represented as: ; in, For the time of data collection, and It is the length of the input vector and the output vector. As an instance of a dataset.

[0031] Based on this, combined Figure 4 As shown, Input into PINN, output (First Prediction Duty Cycle) and (Second prediction duty cycle), used to compare with the label validation set Among (First Real Duty Cycle) and By comparing the (second true duty cycle) to obtain the loss value of the current iteration, the backpropagation gradient can be trained directly based on the loss value to update the weight parameters of the first fully connected neural network and the second fully connected neural network, so that the first predicted duty cycle and the second predicted duty cycle output in the next iteration are closer to the first true duty cycle and the second true duty cycle.

[0032] The following combination Figure 4 Detailed explanation of the structure and reasoning process of PINN (Physical Information Neural Network Controller): To align with HESS control, a power distribution module was designed, utilizing a fully connected neural network. To fit the current reference value, as shown below: ; in, , They are respectively for When the input vector is The predicted reference current values ​​for both the battery side and the supercapacitor side. for Trainable parameters, and This represents the weights and biases of the neurons. The activation function is the LeakyReLU function, and its expression is shown in (7).

[0033] ; To achieve better control performance, and unlike the commonly used loss function-based PINN method, this application embeds a HESS circuit model into the network structure to output the basic duty cycle. (First basic duty cycle) and (Second basic duty cycle), while using a second fully connected neural network. Output corrected duty cycle (First Corrected Duty Cycle) and (First corrected duty cycle). Based on the HESS model and incorporating features learned by neural networks, PINN is able to leverage the learning capabilities of machine learning while preserving physical constraints.

[0034] The forward Euler method is used to discretize equations (1) and (2) for power distribution modules. Received (Predicted current reference value on the battery side) and (The predicted current reference value on the supercapacitor side) is used as the current reference trajectory, and then the basic duty cycle is calculated based on the HESS circuit model. and This is equivalent to embedding physical knowledge into the structure of a neural network. The specific calculation formula is shown below: ; Meanwhile, the predicted current reference value and It will be used as a feature and concatenated into the input vector of PINN. In the middle, through Calculate the corrected duty cycle and This achieves feature fusion, as shown in the following formula: ; in, This indicates the concatenation of column vectors. for Trainable parameters. The activation function is the hyperbolic tangent function tanh, and its expression is shown in (11): ; Regarding the model proposed in this application and Each of its output layers has a practical physical meaning. Considering that the actual duty cycle has upper and lower limits, it is necessary to use an activation function to compare the basic duty cycle obtained from the HESS circuit model with the duty cycle obtained from the HESS circuit model. The obtained modified duty cycle is used as a constraint, and this application employs a hyperbolic tangent function with upper and lower limits. As an activation function, its expression is as follows: ; Will and By summing according to formula (13), the predicted duty cycle of PINN can be obtained. (First Prediction Duty Cycle) and (Second prediction duty cycle).

[0035] ; in, is the adjustment coefficient for feature fusion.

[0036] The following is a detailed introduction to the loss function of PINN: Regarding the design of the loss function, this application retains the data-driven part of traditional neural networks while incorporating physical consistency constraints and adding the control objective to the loss function, thereby enabling the model to achieve better control performance.

[0037] The mean squared error loss of the data-driven part can be expressed by formulas (15) and (16), with the aim of fitting the duty cycle and reference current between the PINN output vector and the tag validation set: ; in, It refers to the batch size during the training process.

[0038] In order for PINN to satisfy the physical laws, by combining formulas (1) to (3), the predicted duty cycle is used. , To obtain the predicted current (Predicted output current on the battery side) (Predicted output current on the supercapacitor side) and voltage (Predicted DC bus voltage) can be obtained as follows: ; The mean square loss of the error in the physical consistency part is given by formulas (20)~(21), with the aim of making the predictive duty cycle that plays a control role satisfy the actual physical laws.

[0039] ; The mean square error of the control section is given by formula (22), with the aim of making the predicted voltage closer to the voltage reference value.

[0040] ; Ultimately, the optimization goals for PINN are as follows: ; in, For the total loss, This serves as a moderating factor for each type of loss.

[0041] Because the PINN proposed in this application relies on the circuit differential equations of Hess for calculation, the preprocessing method of normalizing the dataset would cause the predicted state of the circuit to lose its physical meaning, which obviously contradicts the original intention of PINN. Therefore, the data in this application is not preprocessed. However, this results in differences in the magnitude of various losses, making it more difficult to set the adjustment factor. Therefore, this application preprocesses the variables before each loss calculation. Normalization ,gather Includes loss function All variables are normalized, although this prolongs the model's training time, it does not affect the model's performance in applications. The specific method of normalization is as follows: Before training begins, the variables are first obtained. Maximum value in the collected dataset and minimum value Then, before each loss calculation, use formula (24) to adjust the variables. Normalize.

[0042] ; In summary, this application has the following beneficial effects: This application proposes a PINN structure design framework that differs from traditional dual-loop controllers. It has good versatility for the control method of hybrid energy storage systems. This framework can be extended to hybrid energy storage systems with other different energy storage units. Furthermore, the control objective is incorporated into the loss function design of PINN, resulting in better control performance.

[0043] The PINN controller proposed in this application has better performance than traditional controllers. For load fluctuations faced by hybrid energy storage systems, the PINN controller has a smaller overshoot and a shorter settling time.

[0044] The PINN controller proposed in this application has fewer parameters than traditional neural network structures, thus enabling faster inference time. Please see Figure 5 , Figure 5 This is a schematic diagram of a physical information neural network controller implementation device for hybrid energy storage systems provided in an embodiment of this application. This device is applied to a computer device and may include: The acquisition unit 301 is used to acquire multiple sampled datasets collected for the hybrid energy storage system; Training unit 302 is used to iteratively train the physical information neural network controller using the multiple sampled datasets to obtain the trained physical information neural network controller. The training of each iteration round is performed using a single sampled dataset. The training in a single iteration includes the following steps: Determine the label validation set corresponding to the sampled dataset; The sampled dataset is input into the first fully connected neural network to obtain a first output value output by the first fully connected neural network. The first output value includes the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side of the hybrid energy storage system. The first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side are determined based on the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side. The sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side are input into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network. Based on the first base duty cycle, the first modified duty cycle, the second base duty cycle, and the second modified duty cycle, determine the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side; The loss value of the physical information neural network controller in the current iteration is determined based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set, so as to train the next iteration based on the loss value of the physical information neural network controller in the current iteration.

[0045] Reference Figure 6 , Figure 6 To implement the structural block diagram of a portion of the terminal 140 in this application embodiment, the terminal 140 includes: a radio frequency (RF) circuit 710, a memory 715, an input unit 730, a display unit 740, a sensor 750, an audio circuit 760, a wireless fidelity (WiFi) module 770, a processor 780, and a power supply 790, among other components. Those skilled in the art will understand that... Figure 6 The terminal 140 structure shown does not constitute a limitation on a mobile phone or computer, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0046] The RF circuit 710 can be used to receive and transmit signals during information transmission or calls. In particular, it receives downlink information from the base station and processes it with the processor 780; in addition, it transmits uplink data to the base station.

[0047] The memory 715 can be used to store software programs and modules. The processor 780 executes various functional applications of the terminal and performs processing by running the software programs and modules stored in the memory 715 and the physical information neural network controller for the hybrid energy storage system.

[0048] The input unit 730 can be used to receive input numeric or character information, and to generate key signal inputs related to the terminal's settings and function control. Specifically, the input unit 730 may include a touch panel 731 and other input devices 732.

[0049] The display unit 740 can be used to display input or provided information, as well as various menus of the terminal. The display unit 740 may include a display panel 741.

[0050] Audio circuitry 760, speaker 761, and microphone 762 provide an audio interface.

[0051] In this embodiment, the processor 780 included in the terminal 140 can execute the physical information neural network controller implementation method for hybrid energy storage systems described in the previous embodiment.

[0052] The terminal 140 in this application embodiment includes, but is not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, and aircraft. This application embodiment can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving.

[0053] Figure 7 This is a partial structural block diagram of a server 110 implementing an embodiment of this application. The server 110 can vary significantly due to different configurations or performance characteristics, and may include one or more central processing units (CPUs) 822 (e.g., one or more processors) and memory 832, and one or more storage media 830 (e.g., one or more mass storage devices) for storing application programs 842 or data 844. The memory 832 and storage media 830 can be temporary or persistent storage. The program stored in the storage media 830 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the server 110. Furthermore, the CPU 822 may be configured to communicate with the storage media 830 and execute the series of instruction operations in the storage media 830 on the server 110.

[0054] Server 110 may also include one or more power supplies 826, one or more wired or wireless network interfaces 850, one or more input / output interfaces 858, and / or one or more operating systems 841, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0055] The central processing unit 822 in server 110 can be used to execute the physical information neural network controller implementation method for hybrid energy storage systems according to the embodiments of this application.

[0056] This application also provides a computer-readable storage medium for storing program code, which is used to execute the physical information neural network controller implementation method for hybrid energy storage systems in the foregoing embodiments.

[0057] This application also provides a computer program product, which includes a computer program. A processor of a computer device reads and executes the computer program, causing the computer device to perform the above-described method for implementing a physical information neural network controller for a hybrid energy storage system.

[0058] Furthermore, the terms “comprising” and “including”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0059] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0060] It should be understood that in the description of the embodiments of this application, "multiple" means two or more, "greater than", "less than", "exceeding" etc. are understood to exclude the number itself, and "above", "below", "within" etc. are understood to include the number itself.

[0061] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0062] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0063] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0064] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.

[0066] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0067] The above is a detailed description of the embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for implementing a physical information neural network controller for hybrid energy storage systems, characterized in that, Applied to a physical information neural network controller, the physical information neural network controller comprising a first fully connected neural network and a second fully connected neural network, the method includes: Obtain multiple sample datasets collected for the hybrid energy storage system; The physical information neural network controller is iteratively trained using the multiple sampled datasets to obtain the trained physical information neural network controller. Each iteration of training is performed using a single sampled dataset. The training in a single iteration includes the following steps: Determine the label validation set corresponding to the sampled dataset; The sampled dataset is input into the first fully connected neural network to obtain a first output value output by the first fully connected neural network. The first output value includes the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side of the hybrid energy storage system. The first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side are determined based on the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side. The sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side are input into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network. Based on the first base duty cycle, the first modified duty cycle, the second base duty cycle, and the second modified duty cycle, determine the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side; The loss value of the physical information neural network controller in the current iteration is determined based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set, so as to train the next iteration based on the loss value of the physical information neural network controller in the current iteration.

2. The method for implementing a physical information neural network controller for hybrid energy storage systems according to claim 1, characterized in that, The sampling dataset includes the DC bus voltage of the hybrid energy storage system, the output current on the battery side, the output current on the supercapacitor side, the inductor impedance parameters on the battery side, and the inductor impedance parameters on the supercapacitor side; determining the first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side based on the predicted current reference values ​​on the battery side and the predicted current reference values ​​on the supercapacitor side includes: The first base duty cycle is determined based on the DC bus voltage of the hybrid energy storage system, the output current of the battery side, the inductance parameters of the battery side, and the predicted current reference value of the battery side. The second base duty cycle is determined based on the DC bus voltage of the hybrid energy storage system, the output current of the supercapacitor side, the inductance parameters of the supercapacitor side, and the predicted current reference value of the supercapacitor side.

3. The method for implementing a physical information neural network controller for hybrid energy storage systems according to claim 2, characterized in that, The step of inputting the sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network includes: The predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side are concatenated as features into the sampled dataset to obtain the sampled dataset after feature fusion. The sampled dataset after feature fusion is input into the second fully connected neural network to obtain the first corrected duty cycle and the second corrected duty cycle.

4. The method for implementing a physical information neural network controller for hybrid energy storage systems according to claim 3, characterized in that, The determination of the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side based on the first base duty cycle, the first corrected duty cycle, the second base duty cycle, and the second corrected duty cycle includes: The first prediction duty cycle is determined based on the preset feature fusion coefficient, the first base duty cycle, and the first modified duty cycle; The second predicted duty cycle is determined based on the preset feature fusion coefficient, the second base duty cycle, and the second modified duty cycle.

5. The method for implementing a physical information neural network controller for hybrid energy storage systems according to claim 4, characterized in that, The step of determining the loss value of the physical information neural network controller in the current iteration based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set includes: The loss value of the physical information neural network controller in the current iteration is determined based on the difference between the first predicted duty cycle and the first actual duty cycle, and the difference between the second predicted duty cycle and the second actual duty cycle.

6. The method for implementing a physical information neural network controller for hybrid energy storage systems according to claim 1, characterized in that, The step of determining the loss value of the physical information neural network controller in the current iteration based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set includes: Determine the predicted DC bus voltage, the predicted output current on the battery side, and the predicted output current on the supercapacitor side of the hybrid energy storage system; The difference between the predicted DC bus voltage of the hybrid energy storage system and the actual DC bus voltage in the tag verification set is taken as the first difference. The difference between the predicted output current value on the battery side and the actual output current value on the battery side in the tag verification set is used as the second difference. The difference between the predicted output current value of the supercapacitor side and the actual output current value of the supercapacitor side in the tag verification set is used as the third difference value. The difference between the first predicted duty cycle and the first actual duty cycle is taken as the fourth difference. The difference between the second predicted duty cycle and the second actual duty cycle is taken as the fifth difference. The first difference, the second difference, the third difference, the fourth difference, and the fifth difference are used as the loss values ​​of the physical information neural network controller in the current iteration.

7. A device for implementing a physical information neural network controller for hybrid energy storage systems, characterized in that, Applied to a physical information neural network controller, the physical information neural network controller including a first fully connected neural network and a second fully connected neural network, the device includes: The acquisition unit is used to acquire multiple sampled datasets collected for the hybrid energy storage system; The training unit is used to iteratively train the physical information neural network controller using the multiple sampled datasets to obtain the trained physical information neural network controller. Each iteration of training is performed using a single sampled dataset. The training in a single iteration includes the following steps: Determine the label validation set corresponding to the sampled dataset; The sampled dataset is input into the first fully connected neural network to obtain a first output value output by the first fully connected neural network. The first output value includes the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side of the hybrid energy storage system. The first base duty cycle corresponding to the battery side and the second base duty cycle corresponding to the supercapacitor side are determined based on the predicted current reference value on the battery side and the predicted current reference value on the supercapacitor side. The sampled dataset, the predicted current reference value on the battery side, and the predicted current reference value on the supercapacitor side are input into the second fully connected neural network to obtain the first corrected duty cycle corresponding to the battery side and the second corrected duty cycle corresponding to the supercapacitor side output by the second fully connected neural network. Based on the first base duty cycle, the first modified duty cycle, the second base duty cycle, and the second modified duty cycle, determine the first predicted duty cycle corresponding to the battery side and the second predicted duty cycle corresponding to the supercapacitor side; The loss value of the physical information neural network controller in the current iteration is determined based on the first predicted duty cycle, the second predicted duty cycle, and the first true duty cycle corresponding to the battery side and the second true duty cycle corresponding to the supercapacitor side in the tag verification set, so as to train the next iteration based on the loss value of the physical information neural network controller in the current iteration.

8. An electronic 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 physical information neural network controller implementation method for hybrid energy storage systems as described in any one of claims 1 to 6.

9. A computer program product, the computer program product comprising a computer program, characterized in that, The computer program is read and executed by the processor of the electronic device, causing the electronic device to perform the physical information neural network controller implementation method for hybrid energy storage systems as described in any one of claims 1 to 6.

Images