Large model high computing power server power supply side waste heat electric energy recovery system

Abstract

The invention relates to the technical field of direct-current power supply and distribution and electric energy management, and discloses a large-model high-computing-power server power supply side waste heat electric energy recovery system which is characterized in that a waste heat power generation unit converts heat into direct-current electric energy; the recovery power conversion unit performs power conversion and forms recovery electric energy; the anti-reverse-flow isolation unit blocks reverse flow of power from the direct current bus to the recovery power conversion unit; the recovery control unit is used for acquiring data information, acquiring a reverse current detection result and a grid-connected end voltage detection result, and executing disconnection isolation, pre-charging grid connection and recovery; when the first condition is met, the anti-backflow isolation unit is disconnected, and output of the recycled electric energy to the direct-current bus is stopped; in the recovery process, the redundancy margin is determined, and the target recovery power output with the maximum net income is selected; and estimating open-circuit voltage and equivalent internal resistance, and adjusting an optimization step length and a preset reverse current limit value. According to the invention, stable and controllable grid-connected operation of waste heat recovery is obtained on the premise that the reliability of power supply is not reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of direct current power supply and power management, in particular to a large model high computing power server power side waste heat electric energy recovery system. BACKGROUND

[0002] In the high computing power data center of large model training and inference, the server usually adopts redundant power modules to power the computing and storage load, and gradually evolves into a power supply mode with a direct current bus as the core at the cabinet / column level or even the park level, so as to reduce multi-stage conversion loss and improve power supply reliability. Under high power density operation, the redundant power module and its power side devices will generate continuous and considerable waste heat. At present, this part of heat is mostly taken away and dissipated by air cooling / liquid cooling systems, so that the energy efficiency is improved more by improving the refrigeration efficiency rather than directly recovering electric energy. Therefore, the energy recovery idea of converting the power side waste heat into direct current electric energy and recycling it at the direct current bus side has gradually attracted attention.

[0003] In the prior art, the focus is usually on data center waste heat power generation itself, and the direct current output of waste heat power generation is not controlled in a closed loop according to the engineering constraints of the direct current bus connection system. For example, the patent document CN207995635U discloses a data center cooling waste heat power generation system, which utilizes the temperature difference between the cabinet hot air and the cooling water to generate electricity through a thermoelectric component, and connects the generated electric energy to the electric device (such as an exhaust fan) to realize energy utilization, but it does not give a system-level control mechanism, and there are still some problems: when the recovered electric energy is to be connected to the direct current bus or reused with the bus side load in parallel, there is a lack of anti-backflow isolation and reverse current threshold closed loop, which is easy to form reverse flow and bus disturbance risk in the process of bus voltage fluctuation, load transition or hot plug / redundancy mode switching; there is a lack of pre-charging and grid connection strategy and stability criterion before grid connection, which makes it difficult to suppress grid transient inrush and port voltage difference impact; there is a lack of recovery power selection logic with redundancy margin as a hard constraint, which erodes the available power supply capacity of the redundant power module when pursuing recovery power, thereby affecting the N+1 redundancy requirement; there is a lack of net income optimization that takes into account the recovery electric energy income and the cooling energy consumption cost increased to meet the power temperature constraints, which makes it difficult to stably obtain real energy saving income under the condition of heat-electricity-cooling coupling; there are problems of unstable optimization or mismatched limit setting when the waste heat source and internal resistance change with working conditions.

[0004] Therefore, it is necessary to design a large model high computing power server power side waste heat electric energy recovery system to solve the problems existing in the current technology. SUMMARY

[0005] In view of this, the present application proposes a large model high computing power server power side waste heat electric energy recovery system, aiming to solve the problem of how to realize the safe grid-connected recovery of power side waste heat power generation direct current output under the working conditions of hot plug, redundancy mode switching, alarm and bus fluctuation in the server power supply and distribution network containing a direct current bus, while avoiding the problems of reverse power injection, grid-connected impact, erosion of redundancy capability and unsustainable recovery benefits under thermal constraints.

[0006] The present application proposes a large model high computing power server power side waste heat electric energy recovery system, comprising: A waste heat power generation unit is configured to convert the waste heat on the power side of the redundant power supply module into direct current electric energy and output a waste heat power generation end voltage and a waste heat power generation end current. A recovery power conversion unit is configured to perform power conversion on the direct current electric energy and form recovery electric energy. A reverse injection isolation unit is configured to block the reverse injection of the direct current bus to the recovery power conversion unit when the recovery electric energy is incorporated into the direct current bus and output a reverse current detection result. A recovery control unit is configured to collect the redundancy working mode, hot plug request, alarm, direct current bus voltage, load and cooling energy consumption information of the server management controller, and collect the reverse current detection result output by the reverse injection isolation unit and the grid-connected end voltage detection result output by the recovery power conversion unit, and perform disconnection isolation, pre-charging grid connection and recovery according to the state machine. When the hot plug request, redundancy working mode switching, alarm, direct current bus voltage fluctuation exceeding the preset direct current bus voltage fluctuation limit value or reverse current exceeding the preset reverse current limit value is detected, the reverse injection isolation unit is disconnected and the output of recovery electric energy to the direct current bus is stopped. Before grid connection, the grid-connected end is pre-charged, and after the pre-charging of the grid-connected end to the difference between the grid-connected end voltage and the direct current bus voltage is less than the preset grid-connected voltage difference threshold value, the recovery is entered. In the recovery, the redundancy margin is determined based on the redundancy working mode, the available power supply capacity of the redundant power supply module and the load peak value. When the redundancy margin meets the redundancy requirement and meets the power temperature constraint, the target recovery power output with the maximum net benefit is selected, the net benefit being the recovery electric energy minus the cooling energy consumption increased to meet the power temperature constraint. The voltage and current response of the waste heat power generation unit is collected with limited disturbance, the open circuit voltage and equivalent internal resistance are estimated, and the optimization step and the preset reverse current limit value are adjusted.

[0007] Further, when the recovery control unit collects the redundancy working mode, hot plug request, alarm, direct current bus voltage, load and cooling energy consumption of the server management controller and performs disconnection isolation, pre-charging grid connection and recovery according to the state machine, it comprises: The recovery control unit reads the redundant operating mode, hot plug request, alarm, DC bus voltage, load and cooling energy consumption from the server management controller within a preset collection period, time marks and consistency checks the read results to form state machine inputs; when the consistency check fails, the recovery control unit outputs an open isolation control signal to keep the anti-reflow isolation unit open and to make the recovery power conversion unit in a shutdown state; when the state machine inputs meet the entering conditions of the pre-charge grid-connected state, the recovery control unit outputs a pre-charge control signal to keep the anti-reflow isolation unit open and controls the recovery power conversion unit to perform pre-charge on the grid-connected end; when the state machine inputs meet the entering conditions of the recovery state, the recovery control unit outputs a recovery enable signal to make the anti-reflow isolation unit in a closed state and to make the recovery power conversion unit output recovery electric energy to the DC bus.

[0008] Further, when the hot plug request, the redundant operating mode switching, the alarm, the DC bus voltage fluctuation exceeding the preset DC bus voltage fluctuation limit or the reverse current exceeding the preset reverse current limit are detected, the method comprises: The recovery control unit outputs an open control signal to make the anti-reflow isolation unit open, controls the recovery power conversion unit to stop outputting recovery electric energy to the DC bus, and performs controlled discharge on the grid-connected end in the open state of the anti-reflow isolation unit to make the grid-connected end voltage drop to a preset safety voltage and then keep shutdown.

[0009] Further, the recovery control unit further comprises, within a preset confirmation duration, the hot plug request, the redundant operating mode switching, the alarm, the DC bus voltage fluctuation and the reverse current, and the method further comprises: The recovery control unit continuously determines and only executes the open anti-reflow isolation unit and stops outputting recovery electric energy to the DC bus when any abnormal condition continuously meets the preset confirmation duration; when the abnormal condition disappears but does not continuously meet the preset confirmation duration, the recovery control unit keeps the current state without switching.

[0010] Further, the recovery control unit pre-charges before grid connection, and the method comprises: The recovery control unit keeps the anti-reflow isolation unit open, controls the recovery power conversion unit to pre-charge the grid-connected end in a current-limiting manner, and compares the difference between the grid-connected end voltage and the DC bus voltage in real time within a preset pre-charge duration; when the difference continuously meets the condition of being less than a preset grid-connected voltage difference threshold for a preset stable duration, the recovery control unit allows entering the recovery; when the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold still without reaching the preset stable duration at the end of the preset pre-charge duration, the recovery control unit outputs an open isolation control signal and makes the recovery power conversion unit shutdown.

[0011] Further, the recovery control unit determines the redundancy margin, and the method comprises: The recovery control unit determines a load peak value according to the load within a preset statistical period; determines a redundant power module participating in power supply according to the redundant operation mode, and sets the available power supply capacity of the redundant power module in an alarm state to zero; takes a margin of the available power supply capacity of the redundant power module participating in power supply relative to the load peak value as a redundancy margin, and only allows output of the target recovery power when the redundancy margin is not less than a preset redundancy margin threshold.

[0012] Further, when the recovery control unit determines the target recovery power, the method comprises: The recovery control unit generates a plurality of candidate recovery powers within a preset decision period; respectively acquires recovery electric energy corresponding to each candidate recovery power, and respectively determines cooling energy consumption increased to meet the power supply temperature constraint; takes net income obtained by deducting the cooling energy consumption from the recovery electric energy, and selects a candidate recovery power with the maximum net income from the candidate recovery powers as the target recovery power.

[0013] Further, the power supply temperature constraint comprises a power supply temperature limit value and a temperature rise rate limit value, and the recovery control unit acquires a power supply temperature and a temperature rise rate in recovery, and limits the target recovery power to be not greater than a preset upper limit of recovery or sets the target recovery power to zero when the power supply temperature reaches the power supply temperature limit value or the temperature rise rate reaches the temperature rise rate limit value.

[0014] Further, when the recovery control unit collects the voltage and current responses of the waste heat power generation unit with limited disturbance, estimates the open circuit voltage and the equivalent internal resistance, and adjusts the optimization step and the preset reverse current limit value, the method comprises: The recovery control unit applies a preset amplitude of disturbance to the recovery power conversion unit within a preset disturbance period, and collects responses of the waste heat power generation end voltage and the waste heat power generation end current; when the direct current bus voltage fluctuation does not exceed a preset direct current bus voltage fluctuation limit value, estimates the open circuit voltage and the equivalent internal resistance according to the responses, and adjusts the optimization step and the preset reverse current limit value based on the open circuit voltage and the equivalent internal resistance; when the direct current bus voltage fluctuation exceeds the preset direct current bus voltage fluctuation limit value, stops the limited disturbance and adjusts the optimization step and the preset reverse current limit value to a preset conservative value.

[0015] Further, when the recovery control unit adjusts the optimization step and the preset reverse current limit value based on the open circuit voltage and the equivalent internal resistance, the method further comprises: The recovery control unit obtains open circuit voltage estimation results and equivalent internal resistance estimation results respectively in adjacent two disturbance periods; variation quantity determination is performed on the open circuit voltage estimation results and the equivalent internal resistance estimation results to obtain a stability flag; when the stability flag meets a preset stability condition, corresponding optimization step size and preset reverse current limit value are selected from a preset gear table according to a preset internal resistance interval to which the equivalent internal resistance estimation results belong, and the optimization step size and the preset reverse current limit value are updated according to the gear; when the stability flag does not meet the preset stability condition, the optimization step size and the preset reverse current limit value are gradually reduced to a preset conservative gear, wherein the gradual reduction is only allowed to reduce one gear per disturbance period.

[0016] Compared with the prior art, the beneficial effects of the present application are that: through anti-backfill isolation and reverse current threshold determination, rapid blocking and shutdown are realized under high-risk working conditions such as hot plug, redundancy mode switching, alarm and bus fluctuation, reducing the power supply risk caused by reverse filling and bus disturbance; through pre-charging before grid connection and voltage difference threshold criterion, the process of incorporating the recovery power conversion unit into the DC bus has verifiable voltage consistency, inhibits grid transient inrush current and port impact, and improves grid stability and equipment life; by introducing redundancy margin constraints and temperature constraints, the available power supply capacity and power supply thermal safety boundary are used as hard conditions in the recovery decision, avoiding the erosion of recovery power on the redundancy capability or triggering over-temperature degradation; further taking the cooling energy consumption increased to meet the temperature constraint as the net income target, the control strategy can select the truly energy-saving target recovery power under the thermal-electric-cold coupling, avoiding the appearance of pseudo-revenue with surface recovery but overall energy consumption rising; with the help of limited disturbance, the voltage and current response of the waste heat power generation end are collected and the open circuit voltage and equivalent internal resistance are estimated, and the optimization step size and current limit value are dynamically adjusted, so that the system can still maintain optimization stability, limit value matching and recovery efficiency when the waste heat source intensity and internal resistance change with the working condition. BRIEF DESCRIPTION OF DRAWINGS

[0017] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not considered as limiting the application thereto. Moreover, like reference numerals are used to designate like parts throughout the specification and drawings. In the drawings: Figure 1 The functional block diagram of the large model high computing power server power side waste heat electric energy recovery system provided by the embodiments of the present application is shown. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood, and the scope of the present disclosure can be accurately conveyed to those skilled in the art. It should be noted that the embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0019] The conventional existing data center waste heat power generation system, when recycling electric energy and connecting to the DC bus or parallelly connecting with the bus side load for reuse, has the risk of reverse power injection and bus disturbance, the impact of grid transient inrush current and port voltage difference, the erosion of redundant power supply module available power supply capacity, the difficulty in obtaining real energy saving benefits, and the instability of optimization or mismatch of limit value setting.

[0020] In this regard, see Figure 1As shown, the present application proposes a large model high computing power server power supply side waste heat electric energy recovery system, applied to a server power supply and distribution network containing a direct current bus. The system includes a waste heat power generation unit, a recovery power conversion unit, an anti-backflow isolation unit, and a recovery control unit. The waste heat power generation unit is used to convert the waste heat on the power supply side of the redundant power supply module into direct current electric energy and output a waste heat power generation end voltage and a waste heat power generation end current. The recovery power conversion unit is used to perform power conversion on the direct current electric energy and form recovery electric energy. The anti-backflow isolation unit is used to block the reverse flow of the direct current bus to the recovery power conversion unit when the recovery electric energy is incorporated into the direct current bus and output a reverse current detection result. The recovery control unit is configured to collect the redundant working mode of the server management controller, the hot plug request, the alarm, the direct current bus voltage, the load and cooling energy consumption information, and collect the reverse current detection result output by the anti-backflow isolation unit and the grid-connected end voltage detection result output by the recovery power conversion unit, and perform disconnection isolation, pre-charging grid connection and recovery according to the state machine. When the hot plug request, the redundant working mode switching, the alarm, the direct current bus voltage fluctuation exceeding the preset direct current bus voltage fluctuation limit value, or the reverse current exceeding the preset reverse current limit value is detected, the anti-backflow isolation unit is disconnected and the output of the recovery electric energy to the direct current bus is stopped. Pre-charge the grid-connected end before grid connection, and after the pre-charge of the grid-connected end to the difference between the grid-connected end voltage and the direct current bus voltage is less than the preset grid-connected voltage difference threshold, enter the recovery. In the recovery, the redundant margin is determined based on the redundant working mode, the available power supply capacity of the redundant power supply module, and the load peak value, and when the redundant margin meets the redundant requirement and meets the power supply temperature constraint, the target recovery power output with the maximum net benefit is selected, the net benefit is the recovery electric energy minus the cooling energy consumption increased to meet the power supply temperature constraint; and the voltage and current response of the waste heat power generation unit is collected with limited disturbance, the open circuit voltage and equivalent internal resistance are estimated, and the optimization step and the preset reverse current limit value are adjusted.

[0021] Specifically, the server management controller can be a server BMC, a cabinet-level management controller, or a management entity formed by both; the cooling energy consumption information at least includes fan power consumption, pump power consumption, or cooling actuator power consumption related to redundant power module heat dissipation, and can be directly collected by the management controller or forwarded to the recovery control unit via the cooling controller / cabinet controller; in the net income calculation, the increased cooling energy consumption to meet the power supply temperature constraint can be defined as the incremental cooling actuator power consumption caused by the candidate recovery power relative to the baseline recovery power, to reflect the marginal influence of the recovery strategy on the cooling side. The DC bus voltage fluctuation can be defined as the deviation amplitude of the DC bus voltage instantaneous value relative to the window average value or relative to the last sampling value within a preset voltage statistical window, and when the deviation amplitude exceeds the preset DC bus voltage fluctuation limit, it is determined that the bus is disturbed. The preset reverse current limit is used as a limit parameter variable of the recovery control unit in the present application, and its initial value can be given by a configuration file or factory calibration, and can be updated during operation according to the equivalent internal resistance estimation result and gear table mapping obtained by the limited disturbance; the updated limit is still used as the effective reverse current limit for abnormality determination and anti-backflow protection in the current period.

[0022] Specifically, the waste heat power generation unit is used to convert the waste heat generated on the power supply side of the redundant power module into DC electric energy. It usually contains a thermoelectric conversion device, such as a thermoelectric generator, which works on the principle of generating an electric potential using a temperature difference.

[0023] The recovery power conversion unit is used to adjust and stabilize the voltage or current level of the DC electric energy output by the waste heat power generation unit, so that it can be safely integrated into the DC bus. Usually a DC-DC converter is used to achieve this.

[0024] The anti-backflow isolation unit is used to prevent the electric energy on the DC bus side from flowing back into the recovery power conversion unit when the recovery electric energy is integrated into the DC bus, thereby protecting the stable operation of the system. It includes a controlled isolation switch and a reverse current detection circuit, where the controlled isolation switch is a relay, a contactor, or a semiconductor power switch; optionally, a unidirectional conduction device or an ideal diode control circuit can be connected in series or parallel to enhance the anti-backflow resistance.

[0025] The recovery control unit is the core intelligent part of the entire system, responsible for collecting various operating data and managing and controlling the waste heat electric energy recovery process according to the preset logic and state machine. It is usually implemented by a microcontroller or a digital signal processor, and runs the corresponding control algorithm.

[0026] The DC bus is the main DC power bus in the server power distribution network, connecting multiple power modules and loads to provide a stable DC voltage.

[0027] Redundant power module, a plurality of power modules configured to ensure high reliability in a server power supply system, wherein a part of the power modules are in a redundant standby state under normal load.

[0028] Server management controller, a unit inside the server for monitoring and managing hardware status, operating parameters and executing control instructions, such as baseboard management controller (BMC).

[0029] State machine is a mathematical model used to describe the behavior of a system in different states. In this system, it is used to manage the operation modes of waste heat power recovery, such as disconnection, pre-charging, grid connection and recovery.

[0030] The large model high-performance server power side waste heat power recovery system provided in this embodiment is characterized by the cooperative action of various functional units to realize waste heat power recovery and safe grid connection.

[0031] Specifically, the waste heat power generation unit is responsible for converting the waste heat generated by the power side of the redundant power module into direct current power and outputting the waste heat power generation end voltage and current. For example, a thermoelectric generator array can be used to directly convert heat energy into electricity by sensing the temperature of the power module shell or radiator surface. The output voltage and current will change with the temperature of the heat source and the load.

[0032] The recovery power conversion unit is used to convert the direct current power output by the waste heat power generation unit into power, and form a recovery power suitable for merging into the direct current bus. For example, this unit can be a simple step-up or step-down DC-DC converter, and its output voltage is set by the recovery control unit based on the detection results of the direct current bus voltage and the grid connection end voltage in a closed loop, and the grid connection end voltage is gradually adjusted in a current limiting manner to meet the grid connection voltage difference threshold during the pre-charging and grid connection stage.

[0033] The anti-backflow isolation unit is used to block the reverse flow of power from the direct current bus to the recovery power conversion unit when the recovery power is merged into the direct current bus, and outputs the reverse current detection result. For example, this unit can use a combination of controlled isolation switches and unidirectional conduction devices to form an anti-backflow structure, where the unidirectional conduction device is used to enhance the reverse flow blocking, and the controlled isolation switch is used to realize disconnection / switching under the control of the state machine, and the reverse current detection result is output by the reverse current detection circuit. At the same time, by connecting a current sensor in series in the diode loop, it can be detected whether there is a reverse current.

[0034] The recovery control unit is configured to collect the redundancy working mode of the server management controller, the hot plug request, the alarm, the DC bus voltage, the load and cooling energy consumption information, and collect the reverse current detection result output by the anti-backflow isolation unit and the grid-connected end voltage detection result output by the recovery power conversion unit. According to the collected information, the control unit performs the operations of disconnecting the isolation, pre-charging and grid connection, and recovery, etc. according to the preset logic flow. For example, the recovery control unit can periodically poll the server management controller to obtain these information, and directly judge and control according to these raw data.

[0035] When the hot plug request, the redundancy working mode switching, the alarm, the DC bus voltage fluctuation exceeding the preset DC bus voltage fluctuation limit value, or the reverse current exceeding the preset reverse current limit value is detected, the recovery control unit will disconnect the anti-backflow isolation unit and stop outputting the recovery electric energy to the DC bus. For example, as soon as any of the above abnormal conditions is detected, the recovery control unit immediately sends an instruction to disconnect the anti-backflow isolation unit and turn off the output of the recovery power conversion unit.

[0036] Before grid connection, the system will pre-charge the grid-connected end. Specifically, the recovery control unit keeps the anti-backflow isolation unit disconnected, controls the recovery power conversion unit to pre-charge the grid-connected end in a current-limiting manner, and compares the difference between the grid-connected end voltage and the DC bus voltage in real time within a preset pre-charge time; when the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold for a preset stable time, the recovery control unit allows the recovery; when the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold for the preset stable time has not been reached at the end of the preset pre-charge time, the recovery control unit outputs a disconnecting isolation control signal and makes the recovery power conversion unit stop.

[0037] During the recovery process, the recovery control unit determines the redundancy margin based on the redundancy working mode, the available power supply capacity of the redundant power supply module, and the load peak value. When the redundancy margin meets the redundancy requirement and meets the power supply temperature constraint, the system will select the target recovery power with the maximum net benefit for output. The net benefit is defined as the recovery electric energy minus the cooling energy consumption increased to meet the power supply temperature constraint. For example, the recovery control unit determines the load peak value according to the load within a preset statistical period; determines the redundant power supply modules participating in power supply according to the redundancy working mode, and counts the available power supply capacity of the redundant power supply module in the alarm state as zero; the redundancy margin is the margin of the available power supply capacity of the redundant power supply module participating in power supply relative to the load peak value, and only when the redundancy margin is not less than the preset redundancy margin threshold, the target recovery power is allowed to be output.

[0038] In addition, the recovery control unit also collects the voltage and current response of the waste heat power generation unit in a limited disturbance manner, estimates the open circuit voltage and equivalent internal resistance based on this, and adjusts the optimization step and the preset reverse current limit value. For example, the recovery control unit applies a preset amplitude disturbance to the recovery power conversion unit within a preset disturbance period, and collects the response of the waste heat power generation end voltage and the waste heat power generation end current; when the DC bus voltage fluctuation does not exceed the preset DC bus voltage fluctuation limit value, the open circuit voltage and the equivalent internal resistance are estimated according to the response, and the optimization step and the preset reverse current limit value are adjusted based on the open circuit voltage and the equivalent internal resistance; when the DC bus voltage fluctuation exceeds the preset DC bus voltage fluctuation limit value, the limited disturbance is stopped and the optimization step and the preset reverse current limit value are adjusted to the preset conservative value.

[0039] The large model high computing power server power side waste heat electric energy recovery system provided by the embodiment blocks reverse filling electricity, reduces bus disturbance risk through the cooperative work of the anti-backfilling isolation unit and the recovery control unit. Through the pre-charging strategy before grid connection, the grid transient inrush current and port voltage difference impact are suppressed. By optimizing the selection of recovery power based on redundancy margin and net income, the redundancy requirement is ensured, and real energy saving income is obtained under the condition of heat-electricity-cold coupling. At the same time, by estimating the waste heat source characteristics and dynamically adjusting the control parameters through limited disturbance, the optimization stability and limit value matching of the system when the working condition changes are improved, thereby improving the overall energy efficiency and power supply reliability of the large model high computing power data center.

[0040] In the large model high computing power server power side waste heat electric energy recovery system applied to a server power supply network containing a DC bus, the recovery control unit needs to collect various running information of the server management controller in real time, and perform disconnection isolation, pre-charging grid connection and recovery and other operations accordingly. However, in the actual operation process, due to unstable data transmission or abnormal state of the server management controller itself, the collected information may be delayed, wrong or inconsistent, and if these unreliable data are directly used for state judgment and control, it may cause system malfunction, even cause safety hazards, and affect the efficiency of electric energy recovery and the stability of the system.

[0041] To this end, the application further proposes that the recycling control unit collects the redundancy working mode, hot plug request, alarm, DC bus voltage, load and cooling energy consumption of the server management controller, and performs the specific ways of disconnection isolation, pre-charging and grid connection and recycling according to the state machine. Specifically, the recycling control unit reads the redundancy working mode, hot plug request, alarm, DC bus voltage, load and cooling energy consumption from the server management controller within a preset collection period, and performs time tagging and consistency checking on the read results to form the state machine input. When the consistency check fails, the recycling control unit outputs a disconnection isolation control signal to keep the anti-backflow isolation unit disconnected and to keep the recycling power conversion unit in a shutdown state. When the state machine input meets the entering condition of the pre-charging grid connection state, the recycling control unit outputs a pre-charging control signal to keep the anti-backflow isolation unit disconnected and controls the recycling power conversion unit to perform pre-charging on the grid connection end. When the state machine input meets the entering condition of the recycling state, the recycling control unit outputs a recycling enable signal to keep the anti-backflow isolation unit in the connected state and to make the recycling power conversion unit output recycling power to the DC bus.

[0042] In detail, the recycling control unit actively obtains the latest running data from the server management controller by setting a preset collection period, such as every 100 milliseconds or 500 milliseconds, to ensure the real-time nature of the data. In order to ensure the reliability of the data, the recycling control unit will time tag each set of data read to record its collection time for subsequent timing analysis and data validity judgment. At the same time, the recycling control unit also performs strict consistency checking on the read results. This checking can include but is not limited to cyclic redundancy check to detect transmission errors, data range check to ensure that the data is within a reasonable physical interval, and logic consistency check to verify whether the logical relationship between different data items is reasonable. For example, if the system reports a shutdown state, but the load data shows high load, the data is considered inconsistent. The data after time tagging and consistency checking is used as the input of the state machine.

[0043] When the recycling control unit detects that the consistency check fails, it is considered that the currently collected data is unreliable or abnormal. In order to ensure the safety of the system, the recycling control unit will immediately output a disconnection isolation control signal. The signal will force the anti-backflow isolation unit to remain in the disconnected state, thereby physically cutting off the connection between the recycling system and the DC bus, preventing any potential misoperation or fault from affecting the main power supply network. At the same time, the recycling control unit also makes the recycling power conversion unit in a shutdown state to stop any power output or pre-charging operation, ensuring that the system is in the safest standby mode.

[0044] When the state machine input received by the recovery control unit passes the check and meets the entering condition of the pre-charging grid-connection state, for example, the system is in a stable standby state, there is no alarm, the DC bus voltage is stable, and there is no hot plug request, etc., the recovery control unit outputs a pre-charging control signal. At this stage, the anti-backflow isolation unit is still kept open to avoid the impact of the pre-charging process on the DC bus. The recovery control unit controls the recovery power conversion unit to perform a pre-charging operation on the grid-connection end to gradually increase the voltage of the grid-connection end to approach the DC bus voltage.

[0045] Further, when the state machine input meets the entering condition of the recovery state, for example, the pre-charging grid-connection has been successfully completed (the difference between the grid-connection end voltage and the DC bus voltage is less than a preset threshold), and the system is stable and the redundancy meets the requirements, etc., the recovery control unit outputs a recovery enable signal. At this time, the anti-backflow isolation unit is instructed to be in the closed state to establish an electrical connection between the recovery system and the DC bus. Then, the recovery power conversion unit starts to output recovery electric energy to the DC bus to realize the effective utilization of the waste heat electric energy.

[0046] Through the above technical solutions, the recovery control unit improves the real-time, accuracy and reliability of obtaining information from the server management controller by introducing a preset acquisition period, a time tag and a consistency check mechanism. The problem of system malfunction and safety hazard caused by unreliable data is solved. When detecting data anomalies, the system can quickly and safely enter the open isolation and shutdown state, avoiding potential fault propagation. At the same time, on the premise that the data is reliable, the recovery control unit can orderly perform the pre-charging grid-connection and recovery operation according to the accurate state machine input, ensuring the smoothness and efficiency of the electric energy recovery process. This fine data processing and state control strategy not only enhances the operation safety of the entire waste heat electric energy recovery system, but also optimizes the efficiency and stability of the electric energy recovery, maximizing the recycling value of the server power side waste heat.

[0047] In some embodiments of the present application, when the system detects abnormal conditions such as a hot plug request, a redundancy operating mode switching, an alarm, a DC bus voltage fluctuation exceeding a preset DC bus voltage fluctuation limit, or a reverse current exceeding a preset reverse current limit, etc., the recovery control unit will open the anti-backflow isolation unit and stop outputting recovery electric energy to the DC bus. However, after only performing the open isolation and stop outputting operation, there may still be a high residual voltage at the grid-connection end of the recovery power conversion unit, which not only may prolong the time required for the system to recover to a safe state, but also may bring potential safety hazards to subsequent maintenance operations or re-grid connection.

[0048] In response to the detection of a hot plug request, a redundant operating mode switch, an alarm, a DC bus voltage fluctuation exceeding a preset DC bus voltage fluctuation limit, or a reverse current exceeding a preset reverse current limit, the recovery control unit outputs a disconnection control signal to disconnect the anti-reflow isolation unit, controls the recovery power conversion unit to stop outputting the recovery electric energy to the DC bus, and performs controlled discharge on the grid-connected end in a disconnected state of the anti-reflow isolation unit to make the grid-connected end voltage drop to a preset safety voltage and then remain stopped.

[0049] Specifically, the preset safety voltage can be determined according to the DC bus rated voltage and safety standards / maintenance requirements. The recovery control unit is the core intelligent component of the entire recovery system, responsible for monitoring system status and issuing instructions. When detecting the above abnormal conditions, the recovery control unit will immediately generate and send a disconnection control signal. The signal instructs the anti-reflow isolation unit to perform its core function, i.e., physically or electrically disconnect the connection between the recovery power conversion unit and the DC bus. The anti-reflow isolation unit can be a relay, contactor, or isolation circuit composed of semiconductor switches, which functions to ensure that the recovery electric energy does not continue to be injected into the DC bus under abnormal conditions, and to prevent the DC bus voltage from acting in reverse on the recovery power conversion unit. Synchronously with the disconnection of the anti-reflow isolation unit, the recovery control unit also sends instructions to the recovery power conversion unit to stop its power conversion function, so that it no longer converts the DC electric energy generated by the waste heat power generation unit into recovery electric energy and attempts to output to the DC bus. This is usually achieved by turning off the power switch tube inside the recovery power conversion unit, ensuring that no current flows from the recovery power conversion unit to the DC bus.

[0050] On this basis, after the anti-backflow isolation unit is disconnected, that is, the recovery power conversion unit is physically isolated from the DC bus, the recovery control unit further instructs the recovery power conversion unit to perform controlled discharge on the grid-connected end (that is, the side originally connected to the DC bus). The purpose of controlled discharge is to actively, quickly and safely reduce the residual voltage of the grid-connected end. The way to achieve controlled discharge can include: the recovery power conversion unit is integrated with a controllable discharge resistor, when discharge is needed, the recovery control unit connects the resistor to the grid-connected end circuit to dissipate the residual energy through the resistor; or the inverter part of the recovery power conversion unit can be controlled in a specific mode to dissipate the energy of the grid-connected end through the internal circuit, or to transfer it to the internal energy storage element (such as a capacitor) in a very short time and then dissipate it; or the system can be configured with an independent external discharge circuit controlled by the recovery control unit, which is connected to the grid-connected end for discharge after isolation. This controlled discharge is different from simple natural discharge (through leakage current or parasitic resistance), which is an active and controllable process that can speed up the voltage drop. The preset safety voltage is a voltage threshold determined according to system design and safety standards, which is usually much lower than the working voltage of the DC bus and harmless to personnel contact or equipment safety. During the execution of controlled discharge, the recovery control unit continuously monitors the voltage of the grid-connected end until it drops and stabilizes below the preset safety voltage. Once this condition is met, the recovery control unit will keep the recovery power conversion unit in a stopped state to ensure that the system is in a completely safe and stable state, ready for subsequent troubleshooting, maintenance or restart.

[0051] Through the above technical solution, when the system detects an abnormal situation that requires the recovery function to be disconnected, the recovery control unit not only disconnects the anti-backflow isolation unit and stops the power output of the recovery power conversion unit in time, but further actively instructs the recovery power conversion unit to perform controlled discharge on the grid-connected end after the anti-backflow isolation unit is disconnected. This controlled discharge mechanism can quickly reduce the residual voltage of the grid-connected end to the preset safety voltage, thereby avoiding the safety risks that may be caused by high voltage remaining in the grid-connected end, and shortening the time required for the system to recover from an abnormal state to a safe stopped state. This not only improves the overall safety of the system, but also provides a safer environment for subsequent maintenance operations, and lays the foundation for the system to quickly and smoothly re-grid when conditions permit, avoiding potential shocks or failures caused by voltage remaining, thereby improving the reliability and maintainability of the entire large model high-performance server power side waste heat electric energy recovery system.

[0052] In some embodiments of the application described above, when the recovery control unit detects a hot plug request, a redundancy operating mode switch, an alarm, a DC bus voltage fluctuation exceeding a preset DC bus voltage fluctuation limit, or a reverse current exceeding a preset reverse current limit, the anti-backflow isolation unit will be immediately triggered to disconnect and stop recovering electrical energy. However, in actual operating environment, these abnormal conditions may only be transient or temporary disturbances, and if the system responds immediately to each transient disturbance and performs a disconnection operation, it may cause frequent interruptions of the electrical energy recovery process, reducing the efficiency and stability of the system.

[0053] To this end, the application further proposes that the recovery control unit continuously determines the hot plug request, the redundancy operating mode switch, the alarm, the DC bus voltage fluctuation, and the reverse current within a preset confirmation duration, and only when any abnormal condition continuously meets the preset confirmation duration, the anti-backflow isolation unit is disconnected and the recovery of electrical energy to the DC bus is stopped; when the abnormal condition disappears but does not continuously meet the preset confirmation duration, the recovery control unit remains in the current state without switching.

[0054] Specifically, when the recovery control unit receives abnormal signals such as hot plug request, redundancy operating mode switch, alarm, DC bus voltage fluctuation, or reverse current, it does not immediately trigger a disconnection operation, but starts an internal timer or counting mechanism to continuously monitor the status of these abnormal signals. For example, the recovery control unit can periodically sample these abnormal conditions and record their duration or frequency of occurrence.

[0055] The preset confirmation duration is a pre-set time threshold for judging the persistence of abnormal conditions. The recovery control unit will only issue a disconnection control signal to disconnect the anti-backflow isolation unit and instruct the recovery power conversion unit to stop outputting recovered electrical energy to the DC bus when it continuously and uninterruptedly detects an abnormal condition (e.g., DC bus voltage fluctuation exceeding a preset DC bus voltage fluctuation limit) that meets the abnormal state and its duration reaches or exceeds the preset confirmation duration. The preset confirmation duration can be flexibly configured according to the system's tolerance to transient disturbances, system response speed, and desired stability requirements, for example, set to tens of milliseconds to hundreds of milliseconds.

[0056] If the recovery control unit detects an abnormal condition, but the abnormal condition disappears or returns to normal before reaching the preset confirmation duration, such as a short voltage spike or a transient reverse current, the recovery control unit will not perform a disconnection operation, but will continue to maintain the current electrical energy recovery state. This means that the anti-backflow isolation unit remains connected and the recovery power conversion unit continues to output recovered electrical energy to the DC bus.

[0057] By the technical solution, the recovery control unit can continuously determine the abnormal condition, and introduces a preset confirmation time length as a trigger condition of the disconnection operation. Frequent and unnecessary system disconnection caused by instantaneous or temporary abnormal disturbance is avoided, thereby improving the operation stability and continuity of the electric energy recovery system. The system no longer overreacts to occasional transient events, can maintain the electric energy recovery state for a longer time, maximizes the use of the power side waste heat of the redundant power module, and improves the overall energy utilization efficiency. At the same time, the frequent switching of the system state is reduced, and the potential impact on the DC bus stability is also reduced.

[0058] In some embodiments of the present application, the recovery control unit pre-charges the grid-connected end before grid connection, so that the difference between the grid-connected end voltage and the DC bus voltage is less than the preset grid-connected voltage difference threshold, and then enters the recovery. However, in actual operation, there may be voltage fluctuations or transient instability during the pre-charging process. If only the instantaneous voltage difference is used as the basis for grid connection, grid connection impact or system instability may occur. How to ensure the stability and safety of the pre-charging process and avoid misjudgment of the grid connection condition due to transient fluctuations is a technical problem to be solved.

[0059] To this end, the recovery control unit pre-charges before grid connection, the recovery control unit keeps the anti-backflow isolation unit disconnected, controls the recovery power conversion unit to pre-charge the grid-connected end in a current-limiting manner, and compares the difference between the grid-connected end voltage and the DC bus voltage in real time within a preset pre-charging time length; when the difference continuously satisfies the condition of being less than the preset grid-connected voltage difference threshold for a preset stable time length, the recovery control unit allows the recovery; when the difference continuously satisfies the condition of being less than the preset grid-connected voltage difference threshold for the preset stable time length, the recovery control unit outputs a disconnection isolation control signal and stops the recovery power conversion unit.

[0060] Specifically, the recovery control unit keeps the anti-backflow isolation unit disconnected, which aims to ensure that the grid-connected end of the recovery power conversion unit is electrically isolated from the DC bus during the pre-charging process. This can prevent the voltage or current on the DC bus from causing unnecessary interference or impact on the recovery power conversion unit during the pre-charging phase, thereby ensuring the safety of the pre-charging process. The recovery control unit sends a control instruction to the anti-backflow isolation unit to keep it in a disconnected state, for example, by controlling the on-off of a relay, contactor or semiconductor switch.

[0061] Meanwhile, the recovery control unit controls the recovery power conversion unit to pre-charge the grid-connected end in a current-limited manner. The purpose of current-limited pre-charge is to limit the size of the pre-charge current, avoiding the generation of excessive inrush current when the grid-connected end capacitor is charging, thereby protecting the power devices inside the recovery power conversion unit and the DC bus. The recovery control unit can send a control signal to the recovery power conversion unit, making its internal DC-DC converter or other power conversion circuit work in a current-limited mode, for example, by adjusting the duty cycle of the switching tube, so that the output current does not exceed the preset limit.

[0062] On this basis, the recovery control unit compares the difference between the grid-connected end voltage and the DC bus voltage in real time within the preset pre-charge duration. The recovery control unit continuously collects real-time data of the grid-connected end voltage and the DC bus voltage, and calculates the difference between the two through the internal processor. This real-time comparison mechanism enables the recovery control unit to dynamically monitor the progress of the pre-charge process, ensuring the accuracy of voltage matching. Voltage acquisition can convert analog voltage signals to digital signals through an analog-to-digital converter for processing by the recovery control unit. When the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold for a preset stable duration, the recovery control unit allows entry into recovery. The concepts of "continuous satisfaction" and "preset stable duration" are introduced to ensure that the grid-connected end voltage not only reaches instantaneous matching with the DC bus voltage, but also that this matching state can last for a period of time, thereby avoiding false positives due to transient fluctuations. A timer or counter can be set inside the recovery control unit, which starts timing or counting when the voltage difference meets the condition. Only when the timing or counting reaches the preset stable duration is the pre-charge considered successful, and an instruction is issued to allow entry into recovery.

[0063] In addition, when the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold without reaching the preset stable duration at the end of the preset pre-charge duration, the recovery control unit outputs an isolation control signal to disconnect and stops the recovery power conversion unit. This is an important safety mechanism to handle pre-charge failures or abnormal situations. If the pre-charge process fails to achieve a stable voltage matching state within the specified time, the system will determine that the pre-charge has failed and immediately take measures to disconnect the isolation and stop the work of the recovery power conversion unit to prevent potential malfunctions or damage to the system.

[0064] By the technical solution, the recovery control unit can avoid grid connection impact caused by excessive voltage difference or unstable pre-charging process when pre-charging the grid connection end before grid connection. By keeping the anti-backflow isolation unit disconnected, the electrical isolation safety of the pre-charging process is ensured. The grid connection end is pre-charged in a current limiting manner, which suppresses the pre-charging current and protects the recovery power conversion unit and the DC bus. The difference between the grid connection end voltage and the DC bus voltage is compared in real time, and the system is allowed to enter recovery only after the difference continuously meets the condition of being less than the preset grid connection voltage difference threshold for a preset stable time length, which ensures the voltage matching degree and system stability during grid connection and avoids transient impact. In addition, when the preset pre-charging time length ends and the voltage stability condition is not met, the system can timely output a disconnection isolation control signal and make the recovery power conversion unit stop, thereby avoiding a long period of unstable pre-charging state and improving the reliability and safety of the system.

[0065] In some embodiments of the present application, the recovery control unit determines the redundancy margin based on the redundancy working mode, the available power supply capacity of the redundant power supply module, and the load peak value during recovery. However, if the redundancy margin evaluation is inaccurate, the system may not be able to cope with load fluctuations or power supply module failures when recovering electrical energy, thereby affecting the reliability of server power supply.

[0066] To this end, the present application further proposes a way for the recovery control unit to determine the redundancy margin. Specifically, the recovery control unit determines the load peak value according to the load within a preset statistical period. The load peak value can be obtained by continuously monitoring the overall load of the server (for example, by sampling the current and voltage of the DC bus or each server rack). Within the preset statistical period (for example, configurable for 1 minute, 5 minutes, or 10 minutes), the recovery control unit records and identifies the highest load value in the period as the load peak value. This is intended to accurately capture the maximum power demand of the server system in the near future and provide an accurate benchmark for subsequent redundancy evaluation.

[0067] On this basis, the recovery control unit determines the redundant power supply modules participating in power supply according to the redundancy working mode, and counts the available power supply capacity of the redundant power supply module in an alarm state as zero. The redundancy working mode (such as N+1, N+M, or 2N mode) is provided by the server management controller, indicating how many power supply modules are considered active power supply modules. For each redundant power supply module identified as participating in power supply, its rated power output or actual available power is taken into account. However, if a module is in an alarm state (such as over-temperature, over-current, failure, or communication loss), its available power supply capacity will be considered as zero regardless of its rated capacity. This processing method ensures that the evaluation of the total available power of the system is based on the actual healthy and available power supply modules, avoiding overestimation of the power supply capacity.

[0068] Subsequently, the recovery control unit takes the margin of the available power supply capability of the redundant power supply modules participating in power supply relative to the load peak as the redundancy margin. The redundancy margin is calculated by accumulating the available power supply capabilities of all redundant power supply modules participating in power supply and in a non-alarm state to obtain the total actual available power supply capability of the system, and then comparing the total available power supply capability with the previously determined load peak to calculate the difference or ratio therebetween. The margin quantifies the buffer capability of the power supply system remaining after meeting the current peak load demand.

[0069] Finally, the recovery control unit allows the output of the target recovery power only when the redundancy margin is not less than a preset redundancy margin threshold. The preset redundancy margin threshold is a configurable safety value (e.g., 500 W or 10% of the total capacity). The recovery control unit compares the calculated redundancy margin with the threshold. Only when the redundancy margin reaches or exceeds the threshold, the waste heat electric energy recovery system is allowed to output the target recovery power to the DC bus. If the redundancy margin is lower than the threshold, the recovery system will stop or limit the output of recovered electric energy to ensure that the stability and redundancy of the main power supply system are not affected.

[0070] Through the above technical solutions, the recovery control unit can accurately capture the load peak of the server within a preset statistical period, and accurately identify the redundant power supply modules actually participating in power supply and their available power supply capabilities in combination with the redundancy working mode and real-time alarm information. Based on this, the system can calculate the real redundancy margin, and take it as the key basis for judging whether to allow the output of the target recovery power. When the redundancy margin is lower than the preset threshold, the recovery system will stop or limit the recovery of electric energy, thereby avoiding the further occupation of system margin by the recovery system due to excessive recovery of electric energy when the main power supply system is insufficient in redundancy or has potential risks. The reliability and stability of server power supply are improved, and it is ensured that the key operation of the large model high-performance server is not negatively affected while recovering waste heat electric energy.

[0071] In some embodiments of the present application described above, it is proposed to allow the output of the target recovery power when the redundancy margin meets the redundancy requirement. However, in actual operation, how to accurately determine the target recovery power to maximize the recovery benefit while managing the possible increase in cooling energy consumption due to the recovery of electric energy is a problem to be solved. If only the recovery of electric energy is considered without taking into account the additional cooling cost, it may lead to poor overall system benefits.

[0072] To this end, the application further proposes a specific method for the recovery control unit to determine the target recovery power, comprising: the recovery control unit generates a plurality of candidate recovery powers within a preset decision period; the recovery electric energy corresponding to each candidate recovery power is obtained respectively, and the cooling energy consumption increased to meet the power supply temperature constraint is determined respectively; the net benefit is obtained by deducting the cooling energy consumption from the recovery electric energy, and the candidate recovery power with the maximum net benefit is selected from the candidate recovery powers as the target recovery power.

[0073] Specifically, the recovery control unit generates a plurality of candidate recovery powers within a preset decision period. The preset decision period refers to the time interval for the recovery control unit to make a decision on the target recovery power, which can be set to several seconds or tens of seconds, for example. Within this period, the recovery control unit generates a series of possible recovery power values according to the current system operating state, such as the available power range of the waste heat power generation unit, the DC bus voltage, etc. These candidate recovery powers can be a set of pre-set discrete values, or continuous or quasi-continuous values generated within a certain power range according to a specific algorithm (e.g., based on linear interpolation, historical data analysis or prediction model). The purpose of generating multiple candidate powers is to make comprehensive comparison and evaluation, so as to select the optimal recovery strategy from them.

[0074] Subsequently, the recovery control unit obtains the recovery electric energy corresponding to each candidate recovery power respectively. The recovery electric energy refers to the electric energy actually converted and output to the DC bus by the recovery power conversion unit under a certain candidate recovery power. This can be calculated or estimated by the efficiency curve of the recovery power conversion unit, the characteristics of the waste heat power generation unit and the selected candidate recovery power value. For example, if the candidate recovery power represents an instantaneous power value, the recovery electric energy can be understood as the total amount of electric energy generated at this power level within the preset decision period.

[0075] At the same time, the recovery control unit also needs to determine the cooling energy consumption increased to meet the power supply temperature constraint respectively. When the recovery power increases, the amount of waste heat on the power side of the redundant power module may also increase, which in turn causes the power temperature to rise. In order to ensure that the power temperature is maintained within the preset safe range (i.e., to meet the power temperature constraint), the cooling system may need to increase the operating intensity, thereby generating additional cooling energy consumption. The recovery control unit needs to establish a correlation model between the recovery power and the cooling energy consumption, which can be constructed by experimental data, simulation analysis or empirical formula. The model can take into account various factors such as ambient temperature, cooling system efficiency, heat dissipation characteristics of the power module, etc. For example, according to the size of the recovery power, the temperature rise of the power module can be predicted, and then the additional fan speed or cooling liquid flow required to offset the temperature rise is calculated, and the operating load of the cooling system is converted into the corresponding electric energy consumption.

[0076] On this basis, the recovery control unit deducts the cooling energy consumption from the recovered electric energy to obtain a net benefit. The net benefit is used to uniformly quantify the candidate recovery power in a preset decision period, and the net benefit is defined as the equivalent electric energy benefit obtained by deducting the cooling energy consumption increased to meet the power supply temperature constraint from the recovered electric energy corresponding to the candidate recovery power, wherein the recovered electric energy and the cooling energy consumption are represented in electric energy units; when economic evaluation is needed, the price or energy efficiency coefficient can be introduced to convert the equivalent electric energy benefit into a cost index without changing the above definition.

[0077] Finally, the recovery control unit selects the candidate recovery power with the maximum net benefit from the candidate recovery power as the target recovery power. After calculating the net benefits corresponding to all candidate recovery powers, the recovery control unit compares and selects the candidate recovery power with the maximum net benefit as the target recovery power in the current preset decision period. The selected target recovery power is then output by the recovery power conversion unit.

[0078] Through the above technical solution, the recovery control unit can systematically evaluate the potential benefits under different recovery power levels in a preset decision period. By quantitatively comparing the value of recovered electric energy with the increased cooling energy consumption for maintaining the power supply temperature constraint and calculating the net benefit, the system can avoid the problem of excessively high cooling cost caused by blindly pursuing high recovery power. This decision mechanism based on maximum net benefit ensures that the output target recovery power can bring the best economic benefit and energy efficiency to the entire server power supply network under the premise of meeting the redundancy requirement and power supply temperature constraint, thereby improving the overall operation performance and benefit of the waste heat electric energy recovery system.

[0079] In some embodiments of the present application, a large model high-performance server power side waste heat electric energy recovery system is proposed. The recovery power conversion unit converts the direct current electric energy generated by the waste heat power generation unit into recovery electric energy, and the recovery control unit controls the recovery to realize the recovery and utilization of electric energy. During the recovery process, the recovery control unit determines the redundancy based on the redundancy working mode, the available power supply capacity of the redundancy power module, and the load peak value, and selects the target recovery power with the maximum net benefit to output when the redundancy requirement is met and the power supply temperature constraint is met. However, in actual operation, if the definition and processing mechanism of the power supply temperature constraint are not fine enough, the power module may face the risk of overheating when recovering electric energy, affecting its long-term stable operation and reliability, especially in high load or high ambient temperature conditions. Simply considering the cooling energy consumption may not be enough to ensure the temperature safety of the power module.

[0080] To this end, the application further proposes a specific implementation of the power supply temperature constraint, which includes a power supply temperature limit and a temperature rise rate limit. The power supply temperature limit refers to the maximum temperature that the power supply module can withstand under normal operating conditions, and exceeding this limit may lead to performance degradation, shortened life, or even damage. The temperature rise rate limit refers to the maximum allowed temperature rise per unit time of the power supply module, and an excessively fast temperature rise rate may indicate abnormal heating or cooling system failure, which also needs to be limited to protect the equipment. These limits are usually provided by the power supply module manufacturer or set according to actual operating experience and reliability requirements.

[0081] To implement these constraints, the recovery control unit acquires the power supply temperature and temperature rise rate in real time during the recovery process. This can be achieved by deploying temperature sensors (such as thermistors, thermocouples, etc.) inside or near the power supply module. The sensors convert the collected analog temperature signals into digital signals and transmit them to the recovery control unit. The recovery control unit can periodically collect these temperature data and calculate the temperature rise rate by calculating the temperature difference between adjacent time points. For example, a moving average or difference algorithm can be used to smooth the temperature data and calculate the rate to reduce the impact of transient fluctuations. When the recovery control unit detects that the power supply temperature reaches the power supply temperature limit or the temperature rise rate reaches the temperature rise rate limit, it will take immediate action. Specifically, the recovery control unit will limit the target recovery power to be no greater than a preset recovery upper limit or set the target recovery power to zero. When the power supply temperature reaches the preset power supply temperature limit, it indicates that the power supply module is already in a high-temperature state and needs to be immediately taken to prevent further temperature rise. Similarly, when the temperature rise rate reaches the preset temperature rise rate limit, even if the current temperature has not yet reached the limit, it indicates a potential overheating risk and needs to be intervened in advance. In this case, the recovery control unit will immediately adjust its output strategy to limit the target recovery power below a preset recovery upper limit to reduce the heat generation of the power supply module, or further set the target recovery power to zero to completely stop recovering electrical energy, thereby maximizing the temperature load of the power supply module to ensure its safe operation. The preset recovery upper limit can be a fixed value or a dynamically adjusted value based on the current environmental temperature, load conditions, etc.

[0082] By the technical solution, the definition of the power supply temperature constraint is further refined, not only considering the power supply temperature limit value, but also introducing the temperature rise rate limit value, so that the running thermal state of the power module can be more comprehensively and timely evaluated. The recovery control unit obtains the power supply temperature and the temperature rise rate in real time, compares them with the preset limit value, and once any one reaches or exceeds the limit value, measures to limit the target recovery power are taken immediately, including limiting it below the preset recovery upper limit or directly setting it to zero. This active temperature management strategy avoids the risk of performance degradation, shortened life or even damage of the power module due to overheating, and improves the safety and reliability of the entire waste heat electric energy recovery system. It enables the system to maximize the net benefit of electric energy recovery while ensuring that the power module always operates in a safe thermal working condition, thereby ensuring the long-term stable operation of the server power supply network, especially in complex working conditions with high load or large environmental temperature fluctuations, its protection of the power module is more prominent.

[0083] In some embodiments of the present application, the voltage and current responses of the waste heat power generation unit are collected by limited disturbance, and the open circuit voltage and equivalent internal resistance are estimated, and then the optimization step and the preset reverse current limit value are adjusted to optimize the recovery efficiency and system stability. However, in actual operation, the characteristics of the waste heat power generation unit may change with temperature, aging and other factors, and the system operating environment (such as the DC bus voltage) may fluctuate, if not addressed, it may lead to inaccurate parameter estimation, and then affect the rationality of the optimization step and the reverse current limit value, and even introduce additional disturbance when the system is unstable, affecting the stable operation of the entire server power supply network.

[0084] To this end, the application further proposes a more refined parameter estimation and adjustment strategy. Specifically, the recovery control unit applies a preset amplitude disturbance to the recovery power conversion unit within a preset disturbance period and collects the responses of the waste heat power generation end voltage and the waste heat power generation end current. The preset disturbance period refers to a fixed time interval, for example, every few seconds or tens of seconds, the recovery control unit will start a disturbance process once, and the period is set according to the dynamic response characteristics of the waste heat power generation unit and the real-time requirements of the system for parameter updating. The preset amplitude disturbance usually refers to the recovery control unit changing the control instruction of the recovery power conversion unit (for example, the set value of the output current or voltage) to make the operating point of the recovery power conversion unit slightly and controllably deviate. The amplitude of this disturbance needs to be small enough to avoid affecting the DC bus voltage, but large enough to clearly observe the changes in the responses of the waste heat power generation end voltage and the waste heat power generation end current. The recovery control unit monitors the output voltage and current of the waste heat power generation unit in real time through the built-in voltage sensor and current sensor. When the recovery power conversion unit is disturbed, the voltage and current of the waste heat power generation unit will change accordingly. These changes are accurately recorded as the basis for subsequent parameter estimation.

[0085] When the DC bus voltage fluctuation does not exceed the preset DC bus voltage fluctuation limit, the recovery control unit estimates the open circuit voltage and the equivalent internal resistance according to the response, and adjusts the optimization step and the preset reverse current limit based on the open circuit voltage and the equivalent internal resistance. When the DC bus voltage fluctuation is small and within the normal stable range, it can be considered that the external environment has little effect on the characteristic estimation of the waste heat power generation unit, and the parameter estimation is reliable at this time. Based on the collected response data of the waste heat power generation end voltage and the waste heat power generation end current, the recovery control unit can use various algorithms to estimate the open circuit voltage and the equivalent internal resistance of the waste heat power generation unit, for example, the least squares method can be used to linearly fit the voltage-current data at multiple disturbance points, and the intercept of the fitted straight line is the open circuit voltage, and the negative value of the slope is the equivalent internal resistance. The optimization step refers to the amplitude of adjusting the operating point of the recovery power conversion unit in each iteration of the maximum power point tracking (MPPT) algorithm. If the equivalent internal resistance is large, it means that the output characteristic curve of the waste heat power generation unit is steep, and a smaller optimization step can be used to avoid overshoot and oscillation; if the equivalent internal resistance is small, the curve is flat, and the optimization step can be appropriately increased to speed up the convergence. The preset reverse current limit is the maximum reverse current allowed by the anti-backflow isolation unit, and its setting needs to consider the internal resistance characteristics of the waste heat power generation unit. When the equivalent internal resistance is small, the waste heat power generation unit is more likely to produce large current changes when the voltage fluctuates, and a more stringent reverse current limit may be needed to protect the system.

[0086] When the DC bus voltage fluctuation exceeds the preset DC bus voltage fluctuation limit, the recovery control unit stops the limited disturbance and adjusts the optimization step and the preset reverse current limit to a preset conservative value. The DC bus voltage fluctuation exceeding the preset DC bus voltage fluctuation limit indicates that the system may be in an unstable state or subject to a larger external disturbance. In this case, the recovery control unit immediately stops applying any active disturbance to the recovery power conversion unit, so that it works in a relatively stable state, avoiding additional impact on the already unstable DC bus. At the same time, in order to ensure safe and stable operation, the recovery control unit adjusts the optimization step to a smaller value to reduce the aggressiveness of the maximum power point tracking algorithm, avoiding oscillation caused by rapid adjustment. The preset reverse current limit is also adjusted to a more conservative value to maximize the prevention of reverse power flow from the DC bus to the recovery power conversion unit, thereby protecting the equipment and maintaining system stability.

[0087] Through the above technical solution, the recovery control unit can realize real-time perception and estimation of the characteristics of the waste heat power generation unit in a controlled manner. Specifically, when the system is stable and the DC bus voltage fluctuation is within an acceptable range, the recovery control unit can accurately collect the voltage and current response of the waste heat power generation unit by applying a limited disturbance of a preset amplitude, and accurately estimate the current open-circuit voltage and equivalent internal resistance of the waste heat power generation unit. These real-time updated parameters enable the recovery control unit to adaptively adjust the optimization step of the maximum power point tracking, thereby improving the efficiency and speed of the recovery power conversion unit in tracking the maximum power point, and ensuring optimal electric energy recovery effect under different operating conditions. At the same time, based on the estimated equivalent internal resistance, the preset reverse current limit is dynamically adjusted to more accurately match the actual characteristics of the waste heat power generation unit, prevent reverse power flow, and improve the safety margin of the system. Further, when the DC bus voltage fluctuation is detected to exceed the preset limit, the recovery control unit can stop the disturbance and switch to the preset conservative optimization step and reverse current limit in time, avoiding introducing additional disturbance when the system is unstable, thereby ensuring stable operation of the entire server power supply network and improving the intelligence, adaptability and reliability of the waste heat electric energy recovery system.

[0088] In some embodiments of the present application, the recovery control unit estimates the open-circuit voltage and equivalent internal resistance of the waste heat power generation unit by applying a limited disturbance to the recovery power conversion unit and collecting the voltage and current response of the waste heat power generation unit, and adjusts the optimization step and the preset reverse current limit based on these estimated results. However, in actual operation, these estimated results may be affected by environmental noise, load changes or measurement errors, etc., resulting in unstable or inaccurate estimated values. If the optimization step and the preset reverse current limit are directly adjusted based on unstable estimated values, it may cause frequent fluctuations in system control parameters, affecting recovery efficiency and system stability.

[0089] To this end, the application further proposes that when the recovery control unit adjusts the optimization step size and the preset reverse current limit value based on the open circuit voltage and the equivalent internal resistance, it includes: the recovery control unit obtains open circuit voltage estimation results and equivalent internal resistance estimation results in adjacent two perturbation periods respectively; the change amount of the open circuit voltage estimation results and the equivalent internal resistance estimation results is determined to obtain a stability flag; when the stability flag meets a preset stability condition, according to the preset internal resistance interval to which the equivalent internal resistance estimation result belongs, the corresponding optimization step size gear and the preset reverse current limit value gear are selected from the preset gear table, and the optimization step size and the preset reverse current limit value are updated according to the gear; when the stability flag does not meet the preset stability condition, the optimization step size and the preset reverse current limit value are gradually lowered to a preset conservative gear, wherein the gradual lowering allows only one gear to be lowered in each perturbation period.

[0090] Specifically, the recovery control unit applies limited perturbation to the waste heat power generation unit in two consecutive perturbation periods, and collects the voltage and current responses thereof. Through these response data, the recovery control unit calculates the estimation values of the open circuit voltage and the equivalent internal resistance of the waste heat power generation unit respectively by using a preset estimation algorithm, such as the least square method or the Kalman filter. These estimation results of adjacent periods are stored for subsequent comparison and stability judgment.

[0091] Subsequently, the recovery control unit determines the change amount of the open circuit voltage estimation results and the equivalent internal resistance estimation results obtained in adjacent two perturbation periods. The determination process usually involves calculating the absolute difference or the relative difference between adjacent estimation values, and comparing these differences with preset threshold values. If the change amount of the estimation values of the open circuit voltage and the equivalent internal resistance in adjacent periods are both less than the respective preset threshold values, it is considered that the estimation results are stable, and a “stable” flag is generated; otherwise, if the change amount of any estimation value exceeds the threshold value, it is considered that the estimation results are unstable, and an “unstable” flag is generated. This stability flag is a key basis for subsequent dynamic adjustment of control parameters.

[0092] When the stability flag meets the preset stability condition, it indicates that the characteristic estimation results of the current waste heat power generation unit are reliable. At this time, the recovery control unit will map the equivalent internal resistance estimation result to a preset internal resistance interval. The recovery control unit internally pre-stores a “gear table”, which associates different equivalent internal resistance intervals with specific optimization step size gears and preset reverse current limit value gears. For example, when the equivalent internal resistance is low, a larger optimization step size can be selected to speed up the maximum power point tracking; when the equivalent internal resistance is high, a smaller optimization step size can be selected to improve the tracking accuracy and stability. The recovery control unit queries the current equivalent internal resistance estimation result, selects the corresponding optimization step size gear and preset reverse current limit value gear from the gear table, and updates the actual applied optimization step size and preset reverse current limit value accordingly.

[0093] However, when the stability flag does not satisfy the preset stability condition, i.e., the estimation result is determined to be unstable, the recovery control unit will adopt a conservative strategy. At this time, the system will not make aggressive parameter adjustments based on unstable estimation values, but will gradually reduce the current optimization step and the preset reverse current limit to the preset conservative gear. This gradual reduction mechanism means that in each disturbance period, the optimization step and the preset reverse current limit are at most reduced by one gear. For example, the optimization step is gradually reduced, and the preset reverse current limit is gradually tightened. The preset conservative gear is a parameter setting that has been tested and verified by the system to ensure safe and stable operation of the system in high uncertainty conditions.

[0094] Through the above technical solution, the recovery control unit can judge the stability of the open-circuit voltage and equivalent internal resistance estimation result of the waste heat power generation unit. When the estimation result is stable, the system can dynamically select the most suitable optimization step and preset reverse current limit according to the actual equivalent internal resistance characteristics of the waste heat power generation unit, thereby achieving more accurate and efficient maximum power point tracking and optimizing the protection strategy of the system during normal operation. When the estimation result is unstable, the system adopts a conservative strategy of gradually reducing the control parameters, avoiding the frequent fluctuations of the control parameters and the instability of the system caused by inaccurate estimation values, and improving the robustness and safety of the recovery system. This adaptive parameter adjustment mechanism enables the system to maintain efficient and stable operation when facing environmental changes or measurement uncertainty, ensuring the quality of recovered electric energy and the reliability of grid connection.

[0095] The above technical solution will be further described through a more specific example as follows: In a large data center, multiple large model high-performance server cabinets are deployed, which are powered by N+1 redundant power supply modules and are uniformly powered and distributed through a DC bus. During long-term high-load operation of the servers, the redundant power supply modules and their power supply side devices generate a large amount of waste heat. In order to recover this part of energy, the data center deploys a power supply side waste heat electric energy recovery system.

[0096] The system first converts the waste heat generated by the server redundant power supply modules into DC electric energy through a waste heat power generation unit. For example, the waste heat power generation unit can include a thermoelectric conversion module, which starts to work when the temperature of the heat dissipation fin of the server power supply module reaches a certain value, converts the temperature difference into electric energy, and outputs a waste heat power generation end voltage and a waste heat power generation end current.

[0097] The DC power generated by the waste heat power generation unit is then sent to the recovery power conversion unit. The recovery power conversion unit is a DC-DC converter that performs power conversion such as voltage boosting or voltage reduction on the input DC power to match its voltage and current characteristics with the DC bus, thereby forming the recovery power.

[0098] The anti-backflow isolation unit plays a key role when the recovery power is ready to be integrated into the DC bus. This unit is a power electronic switch or diode array with unidirectional conduction characteristics, and its main function is to block the reverse flow of the DC bus to the recovery power conversion unit when the recovery power is integrated into the DC bus. At the same time, the anti-backflow isolation unit also monitors and outputs the reverse current detection results in real time, providing a basis for subsequent control decisions. It solves the problem of lack of anti-backflow isolation and reverse current threshold closed loop in the prior art, which is prone to form reverse flow and bus disturbance risks during bus voltage fluctuations, load transitions, or hot plug / redundancy mode switching.

[0099] The core of the entire system is the recovery control unit. The control unit is configured to collect multiple aspects of information in real time: Data from the server management controller, including redundancy operating mode (e.g. N+1, N+M), hot plug request (e.g. power module being pulled out or inserted), alarm information (e.g. power module failure, over-temperature), DC bus voltage, server load, and cooling energy consumption information.

[0100] Reverse current detection results from the anti-backflow isolation unit output.

[0101] Grid-connected end voltage detection results from the recovery power conversion unit output.

[0102] The recovery control unit reads these information from the server management controller within a preset collection period, and performs time tagging and consistency checking on the read results to form the state machine input. When the consistency checking fails, the recovery control unit will output a disconnect isolation control signal, causing the anti-backflow isolation unit to remain disconnected and the recovery power conversion unit to be in a shutdown state, ensuring system safety.

[0103] The recovery control unit executes three main states of "disconnect isolation", "pre-charge grid connection" and "recovery" according to these collected information according to the state machine.

[0104] Disconnect isolation state: when the recovery control unit detects any of the following abnormal conditions, the system will immediately enter the disconnect isolation state: The server management controller issues a hot plug request (e.g. an operator is performing maintenance on the server power module).

[0105] Redundancy operating mode switching (e.g. from N+1 to N+2).

[0106] The server management controller issues an alarm (e.g., a certain power module fails).

[0107] The DC bus voltage fluctuation exceeds a preset DC bus voltage fluctuation limit (e.g., the bus voltage transiently drops or rises by more than 5%).

[0108] The reverse current detected by the anti-reflow isolation unit exceeds a preset reverse current limit.

[0109] To avoid misjudgment, the recycling control unit continuously judges these abnormal conditions, and only when any abnormal condition continuously meets the preset confirmation duration (e.g., 500 milliseconds) is it executed to disconnect the anti-reflow isolation unit and control the recycling power conversion unit to stop outputting recycling power to the DC bus. If the abnormal condition disappears within the preset confirmation duration, the recycling control unit will remain in the current state without switching to avoid frequent actions.

[0110] Once it is confirmed that it needs to be disconnected, the recycling control unit outputs a disconnection control signal to make the anti-reflow isolation unit disconnected, and controls the recycling power conversion unit to stop outputting recycling power to the DC bus. To further ensure safety, the recycling control unit will perform controlled discharge on the grid-connected end of the recycling power conversion unit in the anti-reflow isolation unit disconnected state, so that the grid-connected end voltage drops to a preset safety voltage (e.g., 12V) and remains stopped. This solves the problem of lack of anti-reflow isolation and reverse current threshold closed loop in the prior art, which is prone to form reverse flow and bus disturbance risks during bus voltage fluctuation, load transition or hot plug / redundancy mode switching.

[0111] Pre-charge grid-connected state: When the state machine input meets the entering condition of the pre-charge grid-connected state (e.g., all abnormal conditions have been removed, and the system is in a standby grid-connected state), the recycling control unit outputs a pre-charge control signal to make the anti-reflow isolation unit remain disconnected, and controls the recycling power conversion unit to perform pre-charge on the grid-connected end.

[0112] The recycling control unit keeps the anti-reflow isolation unit disconnected and controls the recycling power conversion unit to pre-charge the grid-connected end in a current-limiting manner (e.g., limit the pre-charge current to 1A). During this process, the recycling control unit compares the difference between the grid-connected end voltage and the DC bus voltage in real time within a preset pre-charge duration (e.g., 10 seconds). When the difference continuously meets the condition of being less than a preset grid-connected voltage difference threshold (e.g., 0.5V) for a preset stable duration (e.g., 2 seconds), the recycling control unit determines that the grid-connected condition is mature and allows entering the recycling state.

[0113] If the preset pre-charge time expires, and the difference between the grid-connected terminal voltage and the DC bus voltage continuously meets the condition of being less than the preset grid-connected voltage difference threshold but has not yet reached the preset stabilization time, the recovery control unit will output a disconnection isolation control signal and shut down the recovery power conversion unit to avoid forced grid connection due to pre-charge failure. This pre-charge strategy suppresses grid-connected transient inrush current and port voltage difference impacts, solving the problem of the lack of pre-charge grid connection strategy and stability criteria in existing technologies.

[0114] Recovery status: When the state machine input meets the entry conditions for the recovery status (i.e., pre-charging and grid connection are successful), the recovery control unit outputs a recovery enable signal, which puts the anti-backflow isolation unit on and enables the recovery power conversion unit to output recovered electrical energy to the DC bus.

[0115] During the recycling process, the recycling control unit will make a series of optimization decisions: Redundancy Margin Determination: The recovery control unit determines the peak load based on the server load within a preset statistical period (e.g., 1 minute). Simultaneously, based on the redundancy operating mode provided by the server management controller, it determines the redundant power modules currently participating in power supply. For redundant power modules in an alarm state, their corresponding available power supply capacity is counted as zero. Then, the margin of the total available power supply capacity of the participating redundant power modules relative to the peak load is used as the redundancy margin. The recovery control unit only allows the output of the target recovery power when the redundancy margin is not less than a preset redundancy margin threshold (e.g., the total power supply capacity is at least 1.2 times the peak load), ensuring that redundancy requirements such as N+1 are not eroded. This solves the problem in existing technologies of lacking recovery power selection logic with redundancy margin as a hard constraint.

[0116] Target Recovered Power Selection: The recovery control unit generates multiple candidate recovered power values ​​within a preset decision period (e.g., 10 seconds). For each candidate recovered power value, the system acquires its corresponding recovered electrical energy and determines the additional cooling energy consumption that may be required to meet power supply temperature constraints (including power supply temperature limits and temperature rise rate limits). For example, if increasing the recovered power value leads to an increase in power module temperature, the system estimates the electrical energy required to additionally turn on the fan or increase the liquid cooling flow rate. Then, the recovery control unit calculates the net benefit by subtracting the cooling energy consumption from the recovered electrical energy value and selects the candidate recovered power value with the largest net benefit as the target recovered power value. During recovery, the recovery control unit acquires the power supply temperature and temperature rise rate in real time. When the power supply temperature reaches the power supply temperature limit (e.g., 85°C) or the temperature rise rate reaches the temperature rise rate limit (e.g., 5°C / minute), the target recovered power value is limited to no more than a preset recovery upper limit or the target recovered power value is set to zero to protect the power module. This net benefit optimization strategy solves the problem in existing technologies where it is difficult to stably obtain real energy-saving benefits under thermal-electrical-cold coupling conditions.

[0117] Optimization parameter adjustment: In order to adapt to the problem that the waste heat source and internal resistance change with the working condition, the recovery control unit collects the voltage and current responses of the waste heat power generation unit through limited disturbance. Within a preset disturbance period (for example, 30 seconds), the recovery control unit applies a preset amplitude disturbance (for example, a slight fluctuation of ±5% based on the current output power) to the recovery power conversion unit, and collects the responses of the waste heat generation end voltage and the waste heat generation end current. When the DC bus voltage fluctuation does not exceed the preset DC bus voltage fluctuation limit, the recovery control unit estimates the open circuit voltage and equivalent internal resistance of the waste heat power generation unit according to the response data.

[0118] The recovery control unit obtains the open circuit voltage estimation result and the equivalent internal resistance estimation result in the adjacent two disturbance periods respectively, and judges the change amount of the results to obtain a stability flag. When the stability flag meets the preset stability condition (for example, the change rate of the estimation result is less than 1% for three times in succession), the recovery control unit selects the corresponding optimization step size and the preset reverse current limit value from the preset gear table according to the preset internal resistance interval to which the equivalent internal resistance estimation result belongs, and updates the optimization step size and the preset reverse current limit value according to the gear. For example, when the equivalent internal resistance is large, the optimization step size may be reduced, and the reverse current limit value may be tightened. When the stability flag does not meet the preset stability condition, the recovery control unit gradually reduces the optimization step size and the preset reverse current limit value to the preset conservative gear, wherein the gradual reduction is only allowed to reduce one gear per disturbance period, avoiding aggressive adjustment.

[0119] If the DC bus voltage fluctuation exceeds the preset DC bus voltage fluctuation limit during the limited disturbance, the recovery control unit will immediately stop the limited disturbance, and adjust the optimization step size and the preset reverse current limit value to the preset conservative value, so as to avoid further disturbance to the bus. This mechanism of dynamically adjusting the optimization step size and the reverse current limit value solves the problem that the optimization is unstable or the limit value setting is not matched in the prior art when the waste heat source and the internal resistance change with the working condition.

[0120] Through the above-mentioned cooperation, the power side waste heat electric energy recovery system of the large model high computing power server can safely, efficiently and intelligently recover the power side waste heat electric energy in a complex power supply and distribution environment, while ensuring the stability and redundancy of server power supply, and optimizing the overall energy efficiency.

[0121] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them, although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by the same, without departing from the spirit and scope of the present application. Any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A waste heat and electrical energy recovery system for a large-scale, high-computing-power server power supply side, applied to a server power distribution network including a DC bus, characterized in that... include: The waste heat power generation unit is used to convert the waste heat on the power supply side of the redundant power module into DC power and output the waste heat power generation terminal voltage and waste heat power generation terminal current. A power conversion unit is used to convert the DC power into recovered power. The anti-reverse current isolation unit is used to block the reverse current injection from the DC bus to the recovery power conversion unit and output the reverse current detection result when the recovered electrical energy is connected to the DC bus. The recycling control unit is configured to collect information on the redundant operating mode, hot-plug requests, alarms, DC bus voltage, load and cooling energy consumption of the server management controller, and to collect the reverse current detection results output by the anti-backflow isolation unit and the grid-connected voltage detection results output by the recycling power conversion unit, and to perform disconnection isolation, pre-charging grid connection and recycling according to the state machine. When a hot-plug request, redundant working mode switching, alarm, DC bus voltage fluctuation exceeding the preset DC bus voltage fluctuation limit or reverse current exceeding the preset reverse current limit is detected, the anti-backflow isolation unit is disconnected and the output of recovered electrical energy to the DC bus is stopped; the grid-connected end is pre-charged before grid connection, and the grid-connected end is pre-charged until the difference between the grid-connected end voltage and the DC bus voltage is less than the preset grid-connected voltage difference threshold before entering the recovery process; During the recovery process, the redundancy margin is determined based on the redundant operating mode, the available power supply capacity of the redundant power modules, and the peak load. When the redundancy margin meets the redundancy requirements and the power supply temperature constraint, the target recovered power output with the greatest net benefit is selected. The net benefit is the recovered electrical energy minus the cooling energy consumption increased to meet the power supply temperature constraint. The voltage and current response of the waste heat power generation unit is collected with constrained disturbances to estimate the open circuit voltage and equivalent internal resistance, and the optimization step size and preset reverse current limit are adjusted.

2. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 1, characterized in that, The recycling control unit collects the redundant operating mode, hot-swap requests, alarms, DC bus voltage, load and cooling energy consumption of the server management controller, and performs disconnection isolation, pre-charging grid connection and recycling according to the state machine, including: The recovery control unit reads the redundant operating mode, hot-plug requests, alarms, DC bus voltage, load, and cooling energy consumption from the server management controller within a preset acquisition period, and performs time stamping and consistency verification on the reading results to form state machine input. When the consistency verification fails, the recovery control unit outputs a disconnection isolation control signal to keep the anti-backflow isolation unit disconnected and to put the recovery power conversion unit in a shutdown state. When the state machine input meets the entry conditions for pre-charging grid connection, the recovery control unit outputs a pre-charging control signal to keep the anti-backflow isolation unit disconnected and to control the recovery power conversion unit to perform pre-charging on the grid connection end. When the state machine input meets the entry conditions for recovery state, the recovery control unit outputs a recovery enable signal to put the anti-backflow isolation unit in an on state and to make the recovery power conversion unit output recovered electrical energy to the DC bus.

3. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 1, characterized in that, When a hot-plug request, redundant operating mode switching, alarm, DC bus voltage fluctuation exceeding the preset DC bus voltage fluctuation limit, or reverse current exceeding the preset reverse current limit is detected, including: The recovery control unit outputs a disconnection control signal to disconnect the anti-backflow isolation unit, controls the recovery power conversion unit to stop outputting recovered electrical energy to the DC bus, and performs controlled discharge on the grid-connected terminal while the anti-backflow isolation unit is disconnected, so that the grid-connected terminal voltage drops to a preset safe voltage and remains shut down.

4. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 3, characterized in that, The recycling control unit also includes the following functions within a preset confirmation time: hot-plug requests, redundant operating mode switching, alarms, DC bus voltage fluctuations, and reverse current. The recovery control unit makes continuous judgments and only disconnects the anti-backflow isolation unit and stops outputting recovered power to the DC bus when any abnormal condition continues to meet the preset confirmation time. When the abnormal condition disappears but does not continue to meet the preset confirmation time, the recovery control unit maintains the current state and does not switch.

5. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 4, characterized in that, Pre-charging of the recycling control unit before grid connection includes: The recovery control unit keeps the anti-backflow isolation unit disconnected, controls the recovery power conversion unit to precharge the grid-connected terminal in a current-limiting manner, and compares the difference between the grid-connected terminal voltage and the DC bus voltage in real time within a preset precharge duration; when the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold and reaches a preset stabilization time, the recovery control unit allows recovery to begin; when the preset precharge duration ends and the difference continuously meets the condition of being less than the preset grid-connected voltage difference threshold but has not reached the preset stabilization time, the recovery control unit outputs a disconnection isolation control signal and shuts down the recovery power conversion unit.

6. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 5, characterized in that, When the recycling control unit determines the redundancy margin, it includes: The recovery control unit determines the load peak value based on the load within a preset statistical period; determines the redundant power supply modules participating in power supply according to the redundant working mode, and counts the available power supply capacity corresponding to the redundant power supply modules in alarm state as zero; uses the margin of the available power supply capacity of the redundant power supply modules participating in power supply relative to the load peak value as the redundancy margin, and only allows the output of target recovery power when the redundancy margin is not less than a preset redundancy margin threshold.

7. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 6, characterized in that, When the recovery control unit determines the target recovery power, it includes: The recycling control unit generates multiple candidate recycling powers within a preset decision period; it acquires the recycled electrical energy corresponding to each candidate recycling power and determines the increased cooling energy consumption to meet the power supply temperature constraint; it obtains the net benefit by subtracting the cooling energy consumption from the recycled electrical energy, and selects the candidate recycling power with the largest net benefit as the target recycling power from the candidate recycling powers.

8. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 7, characterized in that, The power supply temperature constraint includes a power supply temperature limit and a temperature rise rate limit. The recovery control unit acquires the power supply temperature and temperature rise rate during recovery, and when the power supply temperature reaches the power supply temperature limit or the temperature rise rate reaches the temperature rise rate limit, it limits the target recovery power to no more than a preset recovery upper limit or sets the target recovery power to zero.

9. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 8, characterized in that, When the recovery control unit acquires the voltage and current response of the waste heat power generation unit with constrained disturbances, estimates the open-circuit voltage and equivalent internal resistance, and adjusts the optimization step size and preset reverse current limit, it includes: The recovery control unit applies a preset disturbance to the recovery power conversion unit within a preset disturbance period and collects the responses of the waste heat power generation terminal voltage and current. When the DC bus voltage fluctuation does not exceed the preset DC bus voltage fluctuation limit, the open circuit voltage and equivalent internal resistance are estimated based on the response, and the optimization step size and preset reverse current limit are adjusted based on the open circuit voltage and equivalent internal resistance. When the DC bus voltage fluctuation exceeds the preset DC bus voltage fluctuation limit, the restricted disturbance is stopped and the optimization step size and preset reverse current limit are adjusted to preset conservative values.

10. The waste heat and electrical energy recovery system on the power supply side of a large-scale high-computing-power server according to claim 9, characterized in that, When the recycling control unit adjusts the optimization step size and the preset reverse current limit based on the open-circuit voltage and equivalent internal resistance, it further includes: The recovery control unit obtains open-circuit voltage estimation results and equivalent internal resistance estimation results in two adjacent disturbance cycles, respectively; it determines the change in the open-circuit voltage estimation results and equivalent internal resistance estimation results to obtain a stability flag; when the stability flag meets the preset stability conditions, it selects the corresponding optimization step size and preset reverse current limit from the preset level table according to the preset internal resistance range to which the equivalent internal resistance estimation result belongs, and updates the optimization step size and preset reverse current limit according to the level; when the stability flag does not meet the preset stability conditions, it gradually lowers the optimization step size and preset reverse current limit to the preset conservative level, wherein the gradual lowering is only allowed to be lowered by one level per disturbance cycle.

Images