A Bidirectional DC-DC Energy Feedback System Based on Multi-Source Integration and Its Control Method

Abstract

The present invention provides a bidirectional DC-DC energy feedback system based on multi-source integration and its control method, which relates to the technical field of energy recovery. The present invention includes an input source module for obtaining multiple input sources; a bidirectional DC-DC power conversion module adaptively connected to the input source module and the energy storage grid interface module to achieve bidirectional controllable energy transmission; a multi-source energy management module is adapted to be connected to a control unit, and the real-time data collected by the sensor network is input into an AI optimization unit for dynamic parameter adjustment and then transmitted to the control unit; the control unit integrates a digital signal processor DSP and an FPGA; the DSP is used to execute a dynamic control strategy; the FPGA is used to generate a high-frequency PWM drive signal; the energy storage grid interface module is used to achieve standardized power interaction between load ends. The present invention can improve energy utilization efficiency, reduce harmonic pollution, and enhance dynamic response ability, and is applicable to new energy vehicle braking energy recovery, renewable energy energy storage, and industrial test equipment scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control and energy recovery, and more particularly, to a bidirectional DC-DC energy feedback system based on multi-source integration and its control method. Background Art

[0002] Bidirectional DC-DC energy feedback systems are widely used in scenarios such as regenerative braking energy recovery in new energy vehicles, renewable energy storage systems, and regenerative power supply for industrial test equipment. However, traditional DC-DC converters usually adopt a unidirectional energy transfer design, which cannot effectively recover the excess energy generated by the load, resulting in energy waste and low system efficiency. For example, during the braking of electric vehicles or the discharging of industrial test equipment, a large amount of energy is dissipated in the form of heat, causing not only an efficiency loss of about 30%-40%, but also a significant increase in heat dissipation costs and carbon emissions, which has an adverse impact on both the environment and the economy.

[0003] Although some bidirectional DC-DC converters in the prior art can achieve energy feedback, there are still many technical bottlenecks to be solved. Firstly, due to high-frequency switching losses and unreasonable topology design, the overall efficiency of existing systems is generally less than 90%, making it difficult to meet the requirements of efficient energy management. Secondly, the dynamic response performance is poor, and voltage fluctuations are easily induced during mode switching, affecting the system stability, especially in complex application scenarios. In addition, there are often harmonic interference problems in the energy fed back to the power grid, and the grid-connected power quality does not meet international standards, which limits its wide application in practical engineering. More importantly, the existing systems have a low level of intelligence and lack adaptive energy management strategies. They usually rely on fixed control strategies and are difficult to adapt to the actual needs of dynamic changes in multi-source inputs.

[0004] Therefore, there is an urgent need for an energy feedback system with high efficiency, fast response, low harmonic pollution, and intelligent decision-making capabilities. In view of this, the applicant has put forward this application after studying the existing technologies to overcome the deficiencies of the prior art and promote the further development and application of energy feedback technologies. Summary of the Invention

[0005] The present invention aims to provide a bidirectional DC-DC energy feedback system based on multi-source integration and its control method to solve the problems of low bidirectional energy transfer efficiency, unstable mode switching, and grid-connected power quality in traditional technologies, and to achieve efficient energy recovery and intelligent management.

[0006] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] A bidirectional DC-DC energy feedback system based on multi-source integration, comprising:

[0008] An input source module for obtaining multiple input sources;

[0009] The bidirectional DC-DC power conversion module includes a bidirectional DC-DC converter and a power switch, which are adaptively connected to the input source module and the energy storage grid interface module for bidirectional controllable energy transmission.

[0010] The multi-source energy management module includes a sensor network and an AI optimization unit, which are adaptively connected to the control unit for inputting the real-time data collected by the sensor network into the AI optimization unit for dynamic parameter adjustment and then transmitting it to the control unit.

[0011] The control unit integrates a digital signal processor DSP and an FPGA; the DSP is configured to communicate with the sensor network to execute a dynamic control strategy; the FPGA is configured to communicate with the bidirectional DC-DC power conversion module to generate a high-frequency PWM drive signal.

[0012] The energy storage grid interface module is adaptively connected to the bidirectional DC-DC power conversion module and the load end for realizing standardized power interaction between the load ends.

[0013] Preferably, the bidirectional DC-DC converter adopts a dual-active-bridge topology structure; the power switch is made of a wide-bandgap semiconductor material, including a silicon carbide MOSFET or a gallium nitride HEMT.

[0014] Preferably, the input source includes photovoltaic, battery and supercapacitor; the load end includes the power grid and the energy storage device; the input source and the load end are collected and monitored in real time through the sensor network.

[0015] Preferably, the objective function of the dynamic control strategy The formula is:

[0016] ;

[0017] Among them, is the tracking error at the k-th time step; , are weight matrices; k represents the time step index, that is, the discrete time step within the prediction horizon; N represents the length of the prediction horizon, that is, the time range for the controller to optimize the future state and input; represents the control input change amount at the k-th time step; T represents the transpose.

[0018] Preferably, the sensor network includes a Hall voltage sensor, a precision shunt resistor and an NTC thermistor;

[0019] Among them, the Hall voltage sensor is used to monitor and collect the input voltage;

[0020] The precision shunt resistor is used to measure and collect the output current;

[0021] The NTC thermistor is used to monitor and collect the temperature of key devices.

[0022] Preferably, the AI optimization unit is used to predict the power demand through a pre-trained LSTM network for the data collected by the sensor network, and perform reinforcement learning using a reward function to dynamically adjust the energy feedback ratio and transmit it to the control unit;

[0023] The formula of the reward function is:

[0024] ;

[0025] Wherein, is the reward function; is the grid electricity price revenue; is the battery health state; 、 are the corresponding weight factors respectively.

[0026] Preferably, the energy storage grid interface includes a full-bridge PWM inverter circuit, a digital phase-locked loop and an islanding protection unit;

[0027] Wherein, the output stage of the full-bridge PWM inverter circuit integrates an LCL filter; the LCL filter includes an inductor and a capacitor, which are used to suppress the switching frequency harmonics;

[0028] The digital phase-locked loop is used to achieve phase synchronization with the grid, and the frequency tracking range of the phase synchronization is 45Hz to 65Hz;

[0029] The islanding protection unit is used to cut off the energy feedback when the grid is abnormal.

[0030] The present invention also provides a bidirectional DC-DC energy feedback control method, including:

[0031] Obtain the input source state and load demand collected by the sensor network in real time, and calculate the SOC of the battery in the input source; wherein, the load demand includes the electricity price of the grid and the load power demand;

[0032] According to the input source state and load demand, perform working mode judgment and switching in combination with a preset logic, and the preset logic is:

[0033] When the data of the input source meets the set power threshold or the electricity price of the grid is less than the set low valley threshold, switch the working mode to the charging mode and execute the charging strategy;

[0034] When the battery SOC is greater than the set threshold or the load power demand is lower than the set rated power, the working mode is switched to the feedback mode, and the energy feedback strategy is executed;

[0035] Among them, the charging strategy is: according to the input source state collected in real time, the charging current and voltage of the battery are dynamically controlled by the DSP and FPGA to adjust the bidirectional DC-DC converter;

[0036] The energy feedback strategy is:

[0037] The sensor network samples the grid feedback current and voltage in real time and performs FFT analysis on harmonics to identify the harmonic amplitude;

[0038] According to the identified harmonic amplitude, the FPGA is used to adjust the PWM to compensate for harmonics until the total harmonic distortion error is less than the set threshold;

[0039] After the PWM adjusts the harmonics, combined with the dynamic frequency fluctuation of the grid, the phase-locked loop is used to synchronize the grid frequency to reduce the harmonic pollution of the feedback current and feed the energy back to the grid;

[0040] The real-time data collected by the sensor network is input into the pre-trained LSTM network, and the power prediction value for a future period of time is output;

[0041] According to the power prediction value, combined with the reward function, the optimal feedback ratio is calculated and selected in real time through deep reinforcement learning;

[0042] According to the optimal feedback ratio, the output parameters of the bidirectional DC-DC power conversion module are adjusted to realize the bidirectional DC-DC energy feedback control.

[0043] The present invention also provides a computer-readable storage medium, including computer-readable instructions stored on the computer-readable storage medium. When the computer-readable instructions are executed by the processor of the device where the computer-readable storage medium is located, the above-mentioned bidirectional DC-DC energy feedback control method is realized.

[0044] To sum up, compared with the prior art, the present invention has the following beneficial effects:

[0045] Through the efficient power conversion technology of the bidirectional DC-DC power conversion module and the intelligent energy management strategy of the control unit and the multi-source energy management module, the present invention realizes the efficient conversion and utilization of energy, significantly improves the overall energy utilization efficiency of the system, and reduces energy waste.

[0046] The present invention has the ability to comprehensively manage and coordinate multiple energy sources (such as batteries, supercapacitors, renewable energy, etc.). It can intelligently allocate energy according to the parameter characteristics, states, and corresponding demands of different energy sources, realizing the complementary advantages of multiple energy sources and improving the comprehensive utilization efficiency of energy.

[0047] The present invention adopts a target function dynamic control algorithm and FPGA to generate PWM signals to optimize dynamic performance. At the same time, it combines LSTM network AI technology to predict load demands and optimize feedback strategies. When abnormal conditions or faults occur in the energy supply, it can quickly respond, automatically adjust the energy allocation strategy, ensure the stable operation of the system, and improve the reliability and fault tolerance of the system. The present invention can significantly improve energy utilization efficiency, reduce harmonic pollution, and enhance dynamic response ability, meeting the requirements of green energy development and being applicable to scenarios such as regenerative braking energy recovery in new energy vehicles, renewable energy energy storage, and industrial test equipment.

[0048] The present invention can either feed back the energy in the energy storage system to the power grid to support the stable operation of the power grid, or obtain energy from the power grid to charge the energy storage system, realizing flexible energy scheduling and optimal configuration. During the energy interaction process, advanced power electronic technologies and reward function control methods are adopted to effectively reduce harmonic pollution, improve power quality, and ensure the stable operation of the power grid and the normal operation of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, so they should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 FIG. 15 is an overall architecture diagram of a bidirectional DC-DC energy feedback system based on multi-source integration provided by the first embodiment of the present invention.

[0051] Figure 2 FIG. 19 is a schematic diagram of energy flow provided by the first embodiment of the present invention.

[0052] Figure 3 FIG. 23 is a flowchart of regenerative braking energy recovery provided by the first embodiment of the present invention.

[0053] Figure 4 FIG. 27 is a configuration diagram of the energy feedback system of industrial test equipment provided by the first embodiment of the present invention.

[0054] Figure 5 FIG. 31 is a schematic flowchart of a bidirectional DC-DC energy feedback control method provided by the second embodiment of the present invention.

[0055] Figure 6 It is a schematic flowchart of the energy feedback strategy provided by the second embodiment of the present invention.

[0056] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. Specific Embodiments

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention. Therefore, the detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0058] Embodiment 1

[0059] As Figure 1 shown, a bidirectional DC-DC energy feedback system based on multi-source integration includes: an input source module, a control unit, a bidirectional DC-DC converter module, an energy storage grid interface module, and a multi-source energy management module.

[0060] The input source module is used to obtain multiple input sources; the input sources include photovoltaic, battery, and supercapacitor;

[0061] The bidirectional DC-DC power conversion module includes a bidirectional DC-DC converter and a power switch, and is adaptively connected to the input source module and the energy storage grid interface module (for example, the input end is electrically connected to the input source module through a pre-charge circuit, and the output end is electrically connected to the energy storage grid interface module) to achieve bidirectional controllable energy transmission;

[0062] The control unit integrates a digital signal processor DSP and an FPGA; the DSP is configured to communicate with a sensor network (such as through an SPI / CAN interface) to execute a dynamic control strategy; the FPGA is configured to communicate with the bidirectional DC-DC power conversion module (such as through an optical fiber isolation driver to connect to the power switch) to generate a high-frequency PWM drive signal;

[0063] The energy storage grid interface module is adaptively connected to the bidirectional DC-DC power conversion module and the load end, and is used to achieve standardized power interaction between the load ends; the load ends include a power grid and an energy storage device.

[0064] The multi-source energy management module, including a sensor network and an AI optimization unit, is used to adaptively connect to a control unit, and input the real-time data of the input source and the load collected by the sensor network into the AI optimization unit for dynamic parameter adjustment and then transmit it to the control unit.

[0065] Furthermore, the bidirectional DC-DC converter adopts a dual-active-bridge topology structure; the power switch is made of wide-bandgap semiconductor materials, including silicon carbide MOSFET or gallium nitride HEMT. Among them, the switching frequency of the silicon carbide MOSFET is adapted to the dual-active-bridge topology.

[0066] In a preferred embodiment, as Figure 2 shown, the input source module uses an 800V high-voltage power battery pack as the input energy source, with the model of NCM 811 battery cells and a capacity of 120kWh. This battery pack realizes bidirectional energy transfer through a dual-active-bridge (DAB) converter, supporting dynamic switching between the forward charging mode and the reverse energy feedback mode. The design of the DAB converter (i.e., the bidirectional DC-DC converter) adopts a high-frequency switching topology structure, and the power device (i.e., the power switch) selects silicon carbide (SiC) MOSFET, such as the specific model C3M0075120K, and the switching frequency can be set to 100kHz. Such a design reduces the conduction loss by 40%-60% compared with traditional silicon-based devices, and the measured overall efficiency reaches 96.2%.

[0067] In addition, the transformer turns ratio in the DAB topology can be designed as 1:1, and ferrite magnetic core materials are used to reduce hysteresis loss and eddy current loss to further improve the system energy efficiency.

[0068] The control unit is composed of a digital signal processor (DSP) and a field programmable gate array (FPGA). Among them, the DSP executes the model predictive control (MPC) algorithm to optimize the switching timing and duty cycle in real time.

[0069] MPC balances the control performance and constraint conditions within the prediction time domain by optimizing the objective function to minimize the state tracking error (such as output voltage fluctuation) and the control cost (such as switching loss).

[0070] The objective function is as follows:

[0071] ;

[0072] where is the tracking error at the k-th time step; , are weight matrices, , used to adjust the weight of the tracking error. The larger Q is, the more inclined the controller is to quickly eliminate the error; , which is used to adjust the weight of the control input change. The larger the R, the more the controller tends to smooth control actions;

[0073] k represents the time step index, that is, the discrete time step within the prediction horizon; N represents the length of the prediction horizon, that is, the time range for the controller to optimize the future state and input; represents the control input change at the k-th time step; T represents the transpose, which is used to convert a column vector into a row vector for matrix multiplication operations.

[0074] Specifically, first establish the discrete-time state-space model of the bidirectional DC-DC converter, and recursively calculate the predicted state sequence based on the current state; substitute the predicted state into the objective function, and combine the constraint conditions (such as duty cycle range constraint, voltage / current limit, switching frequency limit, dead time, etc.) to transform it into a quadratic programming (QP) problem. Take the first control quantity of the optimization result and update the prediction iteratively. The DSP transmits the optimization result (such as duty cycle) to the FPGA through the AXI bus; the FPGA feeds back the real-time state to the DSP to form a closed-loop control.

[0075] In the charging mode, the MPC optimizes the charging current to quickly track the SOC target (such as 90%); the FPGA generates a constant-current PWM to limit the battery temperature rise.

[0076] In the feedback mode, the MPC predicts the grid harmonics, adjusts the PWM to compensate for the THD, combines the LSTM to predict the electricity price, and dynamically optimizes the feedback power.

[0077] Such as Figure 3 As shown, when the vehicle is in a decelerating state, the braking energy generated by the drive motor is fed back to the battery pack through the DAB converter. The control unit calculates the optimal control strategy based on the current system state data (including input voltage, output current, and temperature, etc.) collected by the sensor network to ensure the maximization of the energy transfer efficiency.

[0078] The FPGA generates a PWM drive signal with a resolution of up to 0.1 ns and supports dynamic adjustment of the dead time in the range of 10 ns to 100 ns, thereby effectively suppressing the voltage spike phenomenon during the switching process. In terms of dynamic performance, the measured value of the mode switching time is 78 μs, the voltage fluctuation is only 1.8%, and the overshoot is less than 5%, meeting the high dynamic response requirements.

[0079] The multi-source energy management module includes a sensor network and an AI optimization unit. The system status is monitored in real time through the sensor network, which includes a Hall voltage sensor, a precision shunt resistor, and an NTC thermistor. The measurement range of the Hall voltage sensor can be set from 0 to 1000V as required, and the accuracy is set to 0.2% for monitoring the input voltage. The temperature drift of the precision shunt resistor can be set to less than 50 ppm / °C for measuring the output current. The resolution of the NTC thermistor can be set to 0.1°C as required for monitoring the temperature of key devices.

[0080] The AI optimization unit is used to predict the power demand through a pre-trained LSTM network for the data collected by the Hall voltage sensor, the precision shunt resistor, and the NTC thermistor, and uses a reward function for reinforcement learning to dynamically adjust the energy feedback ratio.

[0081] The data of these sensors are transmitted to the AI optimization unit, which integrates long short-term memory network (LSTM) and deep reinforcement learning (DRL) algorithms.

[0082] Specifically, the collected data such as voltage, current, and temperature (e.g., data with a time window of 1 hour) are filtered to eliminate noise and interference. Then the filtered data are normalized to facilitate the training and prediction of the LSTM network. The normalized data are constructed into a time series in chronological order as the input of the LSTM network.

[0083] In this embodiment, the LSTM network includes an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer receives the time series data. The LSTM layer is used to capture the long-term dependencies in the time series. The fully connected layer is used to map the output of the LSTM layer to the prediction result. The output layer gives the predicted value of the power demand. The LSTM network is trained using historical data, and loss functions such as mean squared error (MSE) are used to evaluate the prediction performance of the model. The generalization ability of the model is verified through methods such as cross-validation to ensure that the model can also give accurate prediction results for unseen data.

[0084] The real-time collected and preprocessed data are input into the trained LSTM network to obtain the predicted value of the future power demand (e.g., the next 5 minutes), and the prediction error can be controlled within less than 5% to guide the system pre-regulation strategy.

[0085] In this embodiment, the deep reinforcement learning (DRL) algorithm defines the state space of the system, including the current power demand prediction value, the energy feedback ratio, the battery state (such as power, temperature, etc.). The action space of the system is defined, that is, the adjustment range of the energy feedback ratio. For example, the energy feedback ratio can be adjusted to any value within the range of [0,1], or defined as several discrete levels. And a reward function is designed to evaluate the effects of the system taking different actions in different states.

[0086] In this embodiment, the DRL algorithm uses the grid electricity price revenue and the battery health state as the reward function R to dynamically adjust the energy feedback ratio. The reward function can comprehensively consider multiple factors such as energy utilization efficiency, system stability, and battery life. When the energy feedback ratio is adjusted properly so that the system can efficiently utilize energy and maintain stable operation, a positive reward is given; otherwise, a negative reward is given.

[0087] For example, giving priority to energy feedback during peak grid electricity price periods can increase the economy of the system by 10%-15%. For example, in a scenario where the grid electricity price is 0.8 yuan / kWh, the daily average revenue increases by 12.7%.

[0088] The formula for the reward function R is:

[0089] ;

[0090] where is the grid electricity price revenue; is the battery health state; and are the corresponding weight factors respectively.

[0091] Then, a suitable reinforcement learning algorithm can be selected for training so that it can learn the strategy of taking optimal actions in different states, such as Q-learning, Deep Q Network (DQN), etc.

[0092] The energy storage grid interaction interface module includes a full-bridge PWM inverter circuit, a digital phase-locked loop (PLL), and an islanding protection mechanism. The output stage of the full-bridge PWM inverter circuit integrates an LCL filter. The inductance value of the LCL filter can be 2mH, and the capacitance value can be 10μF, which is used to suppress switching frequency harmonics and ensure that the total harmonic distortion (THD) of the grid-connected current is less than a set threshold, such as 3%. The digital phase-locked loop (PLL) is used to achieve phase synchronization with the grid. The synchronization error is set to be less than 0.5°, and the frequency tracking range of phase synchronization is set to 45Hz to 65Hz to meet the requirements of the IEEE 1547-2018 standard. The islanding protection mechanism can cut off the feedback path within 10ms when detecting grid abnormalities to comply with the UL1741 standard and ensure the safe operation of the system.

[0093] In addition, this embodiment further includes: in terms of liquid cooling heat dissipation design, a liquid cooling controller is directly attached to the ceramic substrate to ensure that the operating temperature of the power device is maintained within a reasonable range. After 12 hours of full-load operation, the measured temperature rise is 14.2°C, significantly improving the system reliability.

[0094] In another preferred embodiment, the application of the present invention in an industrial test device is demonstrated.

[0095] The hardware configuration of the industrial test device is as follows:

[0096] Input source: 48V lithium-ion battery;

[0097] Bidirectional DC-DC topology: Synchronous Buck-Boost non-isolated design, GaN HEMT switching frequency 1MHz;

[0098] Communication interface: Wi-Fi module is connected to the host monitoring platform.

[0099] Test energy feedback: When the electronic load discharges, 80% of the energy is fed back to the power grid, and the Python monitoring platform displays the energy efficiency data in real time;

[0100] Heat dissipation management: When the temperature > 85°C, air cooling is started, and the temperature rise is suppressed to < 12°C.

[0101] Fault protection: When overcurrent is triggered, the FPGA cuts off the PWM signal within 2μs and sends an alarm message through RS485.

[0102] As Figure 4 shown, the input source module uses a 48V lithium-ion battery pack, and the bidirectional DC-DC converter module uses a synchronous Buck-Boost topology design. The power device selects gallium nitride (GaN) HEMT, with the specific model GS61008P, and the switching frequency is increased to 1MHz. The measured overall efficiency is 95.8%. During the discharge process of the electronic load, 80% of the energy is fed back to the power grid through the synchronous Buck-Boost converter, and the remaining energy is stored in the battery pack.

[0103] The control unit generates PWM signals through a DSP+FPGA hardware platform, and combines with the PID algorithm to achieve precise control, ensuring a stable and reliable energy transmission process. The communication interface uses a Wi-Fi module to connect to the host monitoring platform, and displays the energy efficiency data in real time, facilitating remote monitoring and management by users.

[0104] In terms of heat dissipation management, the air-cooled radiator starts automatically according to the feedback data of the temperature sensor. When the temperature exceeds 85°C, the air-cooled system starts to work, suppressing the temperature rise within 12°C to ensure the long-term stable operation of the system. The fault protection mechanism is implemented through the RS485 communication interface. When an overcurrent fault is detected, the FPGA cuts off the PWM signal within 2μs and sends an alarm message to the monitoring platform through RS485 to remind the user to handle it in time.

[0105] In terms of efficient energy management of the present invention, the overall efficiency of the system reaches more than 95%. The measured values are 96.2% (800V input) and 95.8% (48V input) respectively, with an efficiency improvement of 15% - 20% compared to the traditional silicon-based solution. Secondly, in terms of dynamic performance, the mode switching time is less than 100μs, the measured value is 78μs, the voltage fluctuation is 1.8%, and the overshoot is less than 5%. Thirdly, in terms of green compatibility, the THD of the grid-connected current is less than 3%, the measured value is 2.6%, the grid synchronization error is less than 0.5°, and the temperature rise of liquid-cooled heat dissipation is less than 15°C. Finally, in terms of intelligent decision-making, the load prediction error of LSTM is less than 5%, and the DRL dynamic feedback strategy improves the economy by 10% - 15%. In addition, the system has high reliability. The MTBF of the dual MCU redundant monitoring is greater than 100,000 hours, and the island protection response time is less than 10ms.

[0106] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0107] The present invention includes a bidirectional DC-DC power conversion module, a control unit, a multi-source energy management module, and an energy storage grid interaction interface module. The system realizes efficient energy conversion through a high-frequency switching topology and wide-bandgap semiconductor devices, and uses an objective function dynamic control algorithm and FPGA to generate PWM signals to optimize dynamic performance. At the same time, it combines LSTM network AI technology to predict load demand and optimize the feedback strategy. The present invention can significantly improve the energy utilization efficiency, reduce harmonic pollution, and enhance the dynamic response ability, and is applicable to new energy vehicle braking energy recovery, renewable energy energy storage, and industrial test equipment scenarios.

[0108] The present invention has the ability to comprehensively manage and coordinate multiple energy sources (such as batteries, supercapacitors, renewable energy, etc.). It can intelligently allocate energy according to the parameter characteristics, states, and demands of different energy sources, realize the complementary advantages of multiple energy sources, and improve the comprehensive utilization efficiency of energy. At the same time, when abnormal conditions or faults occur in the energy supply, it can quickly respond, automatically adjust the energy distribution strategy, ensure the stable operation of the system, and improve the reliability and fault tolerance of the system.

[0109] The present invention can not only feed back the energy in the energy storage system to the power grid to support the stable operation of the power grid, but also obtain energy from the power grid to charge the energy storage system, realizing flexible scheduling and optimal allocation of energy. During the energy interaction process, advanced power electronic technologies and reward function control algorithms are adopted to effectively reduce harmonic pollution, improve power quality, and ensure the stable operation of the power grid and the normal operation of equipment.

[0110] By combining hardware design, control algorithms, and intelligent management strategies, the present invention solves the key technical problems in traditional technologies and provides an efficient solution for the regenerative power supply scenarios of new energy vehicle braking energy recovery, renewable energy storage systems, and industrial test equipment. The present invention fully embodies the advantages of efficient energy management, excellent dynamic performance, green compatibility, and intelligent decision-making.

[0111] Embodiment 2

[0112] As Figure 5 、 Figure 6 shown, the second embodiment of the present invention also provides a bidirectional DC-DC energy feedback control method, including:

[0113] Obtain the input source status (such as battery voltage, current, temperature) and load demand (such as grid voltage, current, frequency, electricity price) collected by the sensor network in real time, and calculate the SOC of the battery in the input source; wherein, the load demand includes the electricity price of the power grid and the load power demand;

[0114] According to the input source status and load demand, combined with the preset logic, perform working mode judgment and switching, that is:

[0115] When the data of the input source meets the set power threshold or the electricity price of the power grid is less than the set low valley threshold, switch the working mode to the charging mode and execute the charging strategy;

[0116] When the battery SOC is greater than the set threshold or the load power demand is lower than the set rated power, switch the working mode to the feedback mode and execute the energy feedback strategy;

[0117] Among them, the charging strategy is: according to the input source status collected in real time, dynamically control the bidirectional DC-DC converter through DSP and FPGA to adjust the charging current and voltage of the battery;

[0118] The energy feedback strategy is:

[0119] The sensor network samples the feedback current and voltage of the power grid in real time and performs FFT (Fast Fourier Transform) analysis on harmonics to identify the harmonic amplitude;

[0120] According to the identified harmonic amplitude, the PWM is adjusted by the FPGA to compensate for harmonics until the total harmonic distortion error is less than the set threshold (e.g., 3%);

[0121] After the PWM adjustment, combined with the dynamic frequency fluctuation of the power grid, a phase-locked loop is used to synchronize the power grid frequency to reduce the harmonic pollution of the feedback current and feed the energy back to the power grid;

[0122] The real-time data (such as voltage, current, frequency, electricity price, SOC, etc.) collected by the sensor network is input into the pre-trained LSTM network, and the power prediction value for a future period (e.g., 5 minutes) is output;

[0123] According to the power prediction value, combined with the reward function, the optimal feedback ratio is calculated and selected in real time through deep reinforcement learning;

[0124] According to the optimal feedback ratio, the output parameters of the bidirectional DC-DC power conversion module are adjusted to achieve bidirectional DC-DC energy feedback control.

[0125] In this embodiment, the battery SOC can be calculated based on the ampere-hour integration method or the extended Kalman filter (EKF) algorithm, combined with the battery temperature and charge / discharge current.

[0126] The DSP calculates the optimal charging current and voltage according to the input source status (voltage, current, temperature) collected in real time. The FPGA generates a PWM signal to control the bidirectional DC-DC converter to achieve constant current-constant voltage (CC-CV) charging.

[0127] Specifically, it can be divided into mode switching control, harmonic suppression and grid synchronization, and intelligent optimization strategies.

[0128] In the mode switching control, the system dynamically switches the working mode according to the input source status and load demand. When the photovoltaic input is sufficient or the grid electricity price is at a low valley, the system enters the charging mode, preferentially charging the battery and adopting a constant current / constant voltage control strategy; when the battery SOC is greater than 90% or the load demand drops suddenly, the system switches to the feedback mode, and the maximum power point tracking (MPPT) technology is used to achieve energy feedback, and the measured efficiency reaches 99%; in the seamless switching mode, the FPGA hard real-time control ensures that there is no voltage interruption during the switching process, and the transition time is less than 50 μs.

[0129] In the harmonic suppression and grid synchronization link, the system samples the feedback current at a rate 10 times the switching frequency through FFT spectrum analysis, identifies the second and third harmonic amplitudes, and adjusts the PWM duty cycle through active damping control to compensate for the harmonic components, so that the total harmonic distortion (THD) is less than 3%; then, according to the dynamic frequency fluctuation of the power grid, the PLL parameters are adjusted by using adaptive phase-locking technology to ensure that the synchronization error is less than 0.5°.

[0130] The intelligent optimization strategy predicts the power demand in the next 5 minutes through an LSTM network, combines the DRL algorithm to calculate the optimal feedback ratio in real time, and adjusts the output parameters (such as voltage, current, frequency) of the bidirectional DC-DC converter according to the optimal feedback ratio. The present invention can preferentially feedback energy during the peak period of the grid electricity price, significantly improving the system economy and energy efficiency.

[0131] Embodiment III

[0132] The third embodiment of the present invention also provides a computer-readable storage medium, which includes computer-readable instructions stored on the computer-readable storage medium. When the computer-readable instructions are executed by the processor of the device where the computer-readable storage medium is located, the above-mentioned bidirectional DC-DC energy feedback control method is implemented.

[0133] In several embodiments provided by the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are only illustrative. For example, the flowcharts in the drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0134] In addition, the functional modules in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0135] When the above-mentioned functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this 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 for causing a computer device (which may be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs. It should be noted that in this article, the terms "including", "comprising", or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article, or device including the said element.

[0136] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0137] It should be understood that the term "and / or" used herein is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0138] Depending on the context, the word "if" as used herein can be interpreted as "when", "while", "in response to determining", or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detected (stated condition or event)" can be interpreted as "when determined", "in response to determining", "when detecting (stated condition or event)", or "in response to detecting (stated condition or event)".

[0139] The "first / second" mentioned in the embodiments is only used to distinguish similar objects and does not represent a specific order for the objects. It can be understood that the "first / second" can be interchanged in a specific order or sequence when permitted. It should be understood that the objects distinguished by the "first / second" can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0140] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A bidirectional DC-DC energy feedback system based on multi-source integration, characterized in that: include: Input source module, used to obtain various input sources; A bidirectional DC-DC power conversion module, comprising a bidirectional DC-DC converter and a power switch, adapted to be connected to the input source module and the energy storage grid interface module, for bidirectional controllable transmission of energy; A multi-source energy management module, including a sensor network and an AI optimization unit, is adapted to be connected to the control unit and is used to input the real-time data collected by the sensor network into the AI ​​optimization unit for dynamic parameter adjustment and then transmit it to the control unit; A control unit, integrating a digital signal processor DSP and an FPGA; the DSP is configured to communicate with the sensor network to execute a dynamic control strategy; the FPGA is configured to communicate with the bidirectional DC-DC power conversion module to generate a high-frequency PWM drive signal; An energy storage grid interface module, adapted to be connected to the bidirectional DC-DC power conversion module and the load end, for realizing standardized power interaction between the load ends; The bidirectional DC-DC converter adopts a dual active bridge topology; the power switch is made of wide bandgap semiconductor materials, including silicon carbide MOSFET or gallium nitride HEMT; The objective function of the dynamic control strategy The formula is: ； in, is the tracking error at the kth time step; , is the weight matrix; k represents the time step index, i.e., the discrete time step in the prediction time domain; N represents the prediction time domain length, i.e., the time range for the controller to optimize the future state and input; represents the change in the control input at the kth time step; T represents transposition; The sensor network includes a Hall voltage sensor, a precision shunt resistor and an NTC thermistor; Wherein, the Hall voltage sensor is used to monitor and collect input voltage; The precision shunt resistor is used to monitor and collect the output current; The NTC thermistor is used to monitor and collect the temperature of key components; The AI ​​optimization unit is used to use the data collected by the sensor network to predict power demand through a pre-trained LSTM network, and use a reward function to perform reinforcement learning to dynamically adjust the energy feedback ratio, and transmit it to the control unit; The formula of the reward function is: ； in, is the reward function; The revenue from the power grid price; Battery health status; , are the corresponding weight factors respectively.

2. A bidirectional DC-DC energy feedback system based on multi-source integration according to claim 1, characterized in that: The input source includes photovoltaics, batteries and supercapacitors; the load end includes a power grid and an energy storage device; the input source and the load end are collected and monitored in real time through the sensor network.

3. The bidirectional DC-DC energy feedback system based on multi-source integration according to claim 1 is characterized in that: The energy storage grid interface includes a full-bridge PWM inverter circuit, a digital phase-locked loop and an island protection unit; Wherein, the output stage of the full-bridge PWM inverter circuit integrates an LCL filter; the LCL filter comprises an inductor and a capacitor, and is used to suppress switching frequency harmonics; The digital phase-locked loop is used to achieve phase synchronization with the power grid, and the frequency tracking range of phase synchronization is 45Hz~65Hz; The island protection unit is used to cut off energy feedback when the power grid is abnormal.

4. A bidirectional DC-DC energy feedback control method, applied to a bidirectional DC-DC energy feedback system based on multi-source integration as described in any one of claims 1 to 3, characterized in that: include: Obtain the input source status and load demand collected in real time by the sensor network, and calculate the SOC of the battery in the input source; wherein the load demand includes the electricity price of the power grid and the load power demand; According to the input source status and load requirements, the working mode is judged and switched in combination with the preset logic. The preset logic is: When the data of the input source meets the set power threshold or the electricity price of the power grid is less than the set valley threshold, the working mode is switched to the charging mode and the charging strategy is executed; When the battery SOC is greater than the set threshold or the load power demand is lower than the set rated power, the working mode is switched to the regenerative mode and the energy regenerative strategy is executed; The charging strategy is: according to the input source state collected in real time, the bidirectional DC-DC converter is dynamically controlled by DSP and FPGA to adjust the charging current and voltage of the battery; The energy feedback strategy is: The sensor network samples the grid feedback current and voltage in real time and performs FFT analysis on harmonics to identify harmonic amplitudes; According to the identified harmonic amplitude, the PWM compensation harmonic is adjusted through FPGA until the total harmonic distortion error is less than the set threshold; After PWM adjusts the harmonics, it uses a phase-locked loop to synchronize the grid frequency in combination with the frequency fluctuation dynamics of the grid to reduce the feedback current harmonic pollution and feed energy back to the grid; The real-time data collected by the sensor network is input into the pre-trained LSTM network to output the power prediction value for a period of time in the future; According to the power prediction value, combined with the reward function, deep reinforcement learning is used to calculate and select the optimal reward ratio in real time; According to the optimal feedback ratio, the output parameters of the bidirectional DC-DC power conversion module are adjusted to achieve bidirectional DC-DC energy feedback control.

5. A bidirectional DC-DC energy feedback control method according to claim 4, characterized in that: When the charging current and voltage are adjusted by dynamically controlling the bidirectional DC-DC converter through DSP and FPGA, the objective function is used for control. The formula is: ； in, is the tracking error at the kth time step; , is the weight matrix; k represents the time step index, i.e., the discrete time step in the prediction time domain; N represents the prediction time domain length, i.e., the time range for the controller to optimize the future state and input; represents the change in the control input at the kth time step; T represents transpose.

6. A bidirectional DC-DC energy feedback control method according to claim 4, characterized in that: The formula of the reward function is: ； in, is the reward function; The revenue from the power grid price; Battery health status; , are the corresponding weight factors respectively.

Images