Low-delay control method for ai-based optical storage system and related products

Abstract

The application provides a low-delay control method of an AI-based optical storage system and related products, and relates to the technical field of energy management and control. The method comprises the following steps: collecting equipment data and environmental data of multiple types of equipment in the optical storage system; transmitting the data to an edge computing unit according to a communication protocol adaptation rule; generating a decision instruction according to the data by using an AI hybrid decision model in the edge computing unit; predicting the communication delay between the edge computing unit and each equipment, and compensating the sending time of the decision instruction according to the prediction result; and sending the decision instruction to each equipment through the corresponding communication protocol according to the compensated sending time, and adjusting the running state of each equipment according to the decision instruction. According to the application, the sending time of the decision instruction generated by the edge computing unit is compensated according to the communication delay, the instruction sending lag problem caused by the communication delay is solved, and the low-delay collaborative scheduling of the optical storage system on multiple types of equipment is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy management, and in particular, to a low-delay control method for an AI-based optical storage system and related products. BACKGROUND

[0002] With the transformation of energy structure and the popularization of intelligent technology, the optical storage system composed of photovoltaic equipment, energy storage equipment, charging equipment, heat pump equipment and other electrical equipment is increasingly widely used in household and commercial scenarios. The core requirement of the optical storage system is to improve energy utilization efficiency and power supply reliability through multi-device collaborative control. Currently, the control scheme for the energy management system mainly accesses heterogeneous devices through a multi-protocol communication module and generates energy optimization control instructions by running traditional control algorithms on edge computing nodes to achieve optimal scheduling.

[0003] However, the current control scheme does not involve a compensation mechanism for communication delay after generating energy optimization control instructions, making it difficult to achieve collaborative scheduling for each device and limiting the further improvement of the overall control accuracy of the system. In addition, the control algorithm used in the current control scheme only focuses on a single target and does not consider the multi-objective optimization requirements of the optical storage system, making it difficult to cope with the dynamic and changing multi-objective optimization scenarios in the optical storage system. SUMMARY

[0004] In view of the above problems, a low-delay control method for an AI-based optical storage system and related products are proposed to overcome the above problems or at least partially solve the above problems.

[0005] One object of the present application is to achieve low-delay collaborative scheduling of the optical storage system for multiple types of devices.

[0006] A further object of the present application is to improve the accuracy and stability of communication delay prediction, thereby further improving the accuracy of compensation for decision instruction sending timing.

[0007] Another further object of the present application is to improve the adaptability and optimization capability of the optical storage system for multiple operating conditions.

[0008] In particular, the present application provides a low-delay control method for an AI-based optical storage system. The low-delay control method for the AI-based optical storage system comprises:

[0009] Collecting device data and environmental data of multiple types of devices in the optical storage system;

[0010] Transmitting each item of data to an edge computing unit of the optical storage system according to a preset communication protocol adaptation rule;

[0011] The AI hybrid decision model carried in the edge computing unit generates a decision instruction for each device according to the data;

[0012] The communication delay between the edge computing unit and each device is predicted, and the sending time of the decision instruction is compensated according to the prediction result;

[0013] The decision instruction is sent to each device through the corresponding communication protocol according to the compensated sending time, and each device adjusts the running state according to the decision instruction to realize low-delay control of the optical storage system.

[0014] Optionally, each item of data is accompanied by a time stamp when communicating; and

[0015] The step of predicting the communication delay between the edge computing unit and each device comprises:

[0016] The edge computing unit calculates the communication round trip time with each device according to the time stamp;

[0017] The communication round trip time corresponding to each device is smoothed by using a preset sliding window algorithm to obtain a prediction result corresponding to each device.

[0018] Optionally, the step of compensating the sending time of the decision instruction according to the prediction result comprises:

[0019] The planned sending time of the decision instruction is obtained based on the control cycle of the optical storage system;

[0020] The time point before the planned sending time and spaced from the planned sending time by the prediction result is taken as the target sending time to obtain the compensated sending time.

[0021] Optionally, the AI hybrid decision model comprises a lightweight prediction model for short-term load prediction, a deep reinforcement learning model for long-term scheduling optimization, and a response model for emergency response; and

[0022] The step of generating a decision instruction for each device according to the data by using the AI hybrid decision model carried in the edge computing unit comprises:

[0023] It is judged whether the device data and the environment data exceed a preset safety threshold;

[0024] If yes, the response model is triggered to directly generate an emergency handling instruction;

[0025] If not, the future power load of the optical storage system in a preset time period is predicted by using the lightweight prediction model according to the device data, the environment data, and the historical operation data of the optical storage system;

[0026] The decision instructions for each device are calculated according to the power load under multi-target constraints by using a deep reinforcement learning model.

[0027] Optionally, after the step of controlling each device to adjust the operating state according to the decision instructions, the low-delay control method of the AI-based optical storage system further comprises:

[0028] uploading the data to a cloud platform connected to the optical storage system at a first preset period;

[0029] training the model parameters of the deep reinforcement learning model according to the received data within a second preset period, wherein the second preset period is greater than the first preset period;

[0030] downloading the trained model parameters to the edge computing unit to update the model parameters of the deep reinforcement learning model.

[0031] Optionally, the step of transmitting each item of data to the edge computing unit of the optical storage system according to the preset communication protocol adaptation rule comprises:

[0032] allocating an adapted communication protocol to each device according to the preset communication protocol adaptation rule;

[0033] transmitting the data corresponding to each device to the edge computing unit through the communication protocol adapted for the device.

[0034] Optionally, the multiple types of devices include at least one of a photovoltaic device, an energy storage device, a charging device, and a heat pump device; and

[0035] The step of controlling the operating state of each device according to the decision instructions comprises:

[0036] converting the decision instructions into control signals matched with the communication protocol corresponding to each device;

[0037] executing the control signals through the control interface of each device to adjust the operating state of the device.

[0038] According to another aspect of the present application, there is also provided a computer program product comprising a computer program which, when executed by a processor, implements the steps of any of the above low-delay control methods of the AI-based optical storage system.

[0039] According to yet another aspect of the present application, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of any of the above low-delay control methods of the AI-based optical storage system.

[0040] According to a further aspect of the present application, there is also provided a computer device comprising a memory, a processor, and a computer program stored on the memory, the processor executing the computer program to implement the steps of any of the above AI-based low-latency control methods for a photovoltaic storage system.

[0041] The AI-based low-latency control method for a photovoltaic storage system of the present application ensures comprehensive perception of the system operating state and external environment state and compatible transmission by collecting device data and environment data of multiple types of devices and transmitting them to the edge computing unit according to preset communication protocol adaptation rules, and realizes low-latency generation of each decision instruction by generating decision instructions using the AI hybrid decision model carried in the edge computing unit. In addition, the AI-based low-latency control method for a photovoltaic storage system of the present application also predicts the communication delay between the edge computing unit and each device, compensates the transmission timing of the decision instruction according to the prediction result, and issues the decision instruction according to the compensated transmission timing to control the device to adjust the state, effectively overcoming the transmission delay uncertainty of the decision instruction for multiple types of devices. Thus, the AI-based low-latency control method for a photovoltaic storage system of the present application transmits the collected data according to the communication protocol adaptation rules, and compensates the transmission timing of the decision instruction generated by the edge computing unit according to the communication delay, solves the instruction issuance lag problem caused by communication delay in the prior art, ensures the timing accuracy of the decision instruction for each device, and thus ensures the coordinated control of the photovoltaic storage system for multiple types of devices, and further realizes low-latency coordinated scheduling of the photovoltaic storage system.

[0042] Further, the AI-based low-latency control method for a photovoltaic storage system of the present application provides a reliable time reference for accurate measurement of delay by attaching a time stamp to each item of data communication, and realizes accurate quantification of communication delay by calculating the round-trip time according to the time stamp by the edge computing unit. In addition, the AI-based low-latency control method for a photovoltaic storage system of the present application also effectively filters out noise interference caused by instantaneous jitter by using a preset sliding window algorithm to smooth the round-trip time. Thus, the AI-based low-latency control method for a photovoltaic storage system of the present application improves the accuracy and stability of communication delay prediction by predicting the communication delay between the edge computing unit and each device according to the time stamp attached to the data and the preset sliding window algorithm, provides a precise data basis for subsequent compensation operations, and thus improves the accuracy of coordinated scheduling of the photovoltaic storage system.

[0043] Further, the low-delay control method of the AI-based optical storage system of the present application realizes the targeted adaptive processing of different scenes by configuring the AI hybrid decision model as a combination of a lightweight prediction model, a deep reinforcement learning model and a response model, and can quickly identify the system operating condition and trigger the corresponding model by judging whether the data exceeds the preset safety threshold. In the case where the data exceeds the safety threshold, the response model is triggered to directly generate emergency event disposal instructions, realizing the rapid response under emergency conditions and avoiding disposal delay caused by time-consuming complex calculation. In the case where the data does not exceed the safety threshold, the lightweight prediction model is first used to predict the future short-term power load according to the collected data and historical operating data, and then the deep reinforcement learning model is used to calculate the decision instructions for the operation of each device according to the prediction results under multi-objective constraints, thereby realizing the accurate generation of decision instructions under normal conditions. Thus, the low-delay control method of the AI-based optical storage system of the present application realizes the intelligent identification and hierarchical decision of the system operating state by constructing an AI hybrid decision model containing multiple types of models and a dual protection mechanism of safety response and optimized scheduling, realizes the rapid disposal under emergency conditions and the multi-objective optimization under normal conditions, and thus improves the adaptability and optimization capability of the optical storage system to various operating conditions.

[0044] The above and other objects, advantages and features of the present application will become more apparent from the following detailed description of some embodiments thereof, taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0045] Some embodiments of the present application will now be described in detail with reference to the drawings, which are provided by way of example and are not limiting of the present application. Like references denote like elements or portions in the figures. It will be appreciated that these figures are not necessarily drawn to scale. In the figures:

[0046] Figure 1 is a control block diagram of the low-delay control method of the AI-based optical storage system according to an embodiment of the present application;

[0047] Figure 2 is a flowchart of a control example of the low-delay control method of the AI-based optical storage system according to an embodiment of the present application;

[0048] Figure 3 is a schematic diagram of a computer program product according to an embodiment of the present application;

[0049] Figure 4 is a schematic diagram of a computer-readable storage medium according to an embodiment of the present application;

[0050] Figure 5 is a schematic block diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0051] Those skilled in the art should understand that the embodiments described below are only a part of the embodiments of the present application, and are not the whole embodiments of the present application, and the part of the embodiments are intended to explain the technical principles of the present application, and are not intended to limit the protection scope of the present application. Based on the embodiments provided by the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should fall within the protection scope of the present application.

[0052] Figure 1 is a control block diagram of a low-delay control method of an AI-based optical storage system according to an embodiment of the present application. As shown in the figure, the low-delay control method of the AI-based optical storage system of the present embodiment can generally include: Figure 1

[0053] Step S102, collecting device data and environment data of multiple types of devices in the optical storage system. It should be noted that the multiple types of devices in the present embodiment include at least one of photovoltaic devices, energy storage devices, charging devices and heat pump devices. Specifically, the charging device can be a charging pile. The device data in the present embodiment can include device state data and device operation data of each device, for sensing the system operation state. In addition, the environment data in the present embodiment can include environment state parameters collected in real time by sensors at a plurality of monitoring positions in the optical storage system, for sensing the external environment state, so as to ensure comprehensive sensing of the system operation state and the external environment state of the optical storage system.

[0054] Step S104, transmitting each item of data to an edge computing unit of the optical storage system according to a preset communication protocol adaptation rule. It should be noted that the edge computing unit in the present embodiment can be a core unit of the optical storage system that undertakes local AI decision, multi-device communication adaptation, real-time data processing and collaborative control, and is loaded with an AI hybrid decision model, so as to realize the functions of local real-time decision and device collaborative control. In addition, the communication protocol adaptation rule in the present embodiment can be used to formulate differentiated adaptation strategies for the communication characteristics of different devices such as photovoltaic devices, energy storage devices, charging devices and heat pump devices, so as to ensure the compatibility and real-time performance of data transmission.

[0055] ​In step S106, the AI hybrid decision model carried in the edge computing unit is used to generate decision instructions for the operation of each device according to the data. It should be noted that the decision instructions in the present embodiment can be control instructions for adjusting the operating state of the target device. For example, the decision instructions for the energy storage device can include adjusting the charging power or discharging power of the energy storage device, and the decision instructions for the heat pump device can include adjusting the operating frequency of the heat pump device. Thus, the low-delay control method of the AI-based optical storage system in the present embodiment realizes low-delay generation of each decision instruction by using the AI hybrid decision model carried in the edge computing unit to generate the decision instructions.

[0056] In step S108, the communication delay between the edge computing unit and each device is predicted, and the transmission timing of the decision instructions is compensated according to the prediction result. It should be noted that in the present embodiment, the communication characteristics and the adapted communication protocols of the multiple types of devices such as the photovoltaic device, the energy storage device, the charging pile and the heat pump device can be different, and thus the communication delay between the edge computing unit and the multiple types of devices is also different. Thus, the low-delay control method of the AI-based optical storage system in the present embodiment effectively overcomes the problem of lag in the transmission of control instructions for multiple types of devices caused by communication delay by predicting the communication delay of each path and compensating the transmission timing of each decision instruction according to the prediction result.

[0057] In step S110, the decision instructions are transmitted to each device through the corresponding communication protocol according to the compensated transmission timing, and each device adjusts the operating state according to the decision instructions, so as to realize low-delay control of the optical storage system. Thus, the low-delay control method of the AI-based optical storage system in the present embodiment effectively overcomes the transmission delay uncertainty of the control instructions for multiple types of devices by transmitting the decision instructions according to the compensated transmission timing to control the devices to adjust the state.

[0058] Thus, the low-delay control method of the AI-based optical storage system in the present embodiment transmits the collected data through the communication protocol adaptation rule, and compensates the transmission timing of the decision instructions generated by the edge computing unit, solves the problem of lag in the transmission of instructions caused by communication delay in the prior art, ensures the timing accuracy of the decision instructions for each device, and thus ensures the cooperative control of the optical storage system for multiple types of devices, and further realizes low-delay cooperative scheduling of the optical storage system.

[0059] In some embodiments, the hardware architecture of the optical storage system of the present application can include a perception layer, an edge computing unit, and an execution layer. Specifically, the edge computing unit is equipped with an AI chip accelerated by a neural network processing unit and a multi-protocol communication module. The AI chip is used to load and run an AI hybrid decision model, and the multi-protocol communication module is used to communicate with various devices in an adaptive communication protocol. In addition, the perception layer can include a plurality of power sensors arranged at a plurality of devices and a plurality of environmental sensors deployed at a plurality of preset environmental monitoring locations in the optical storage system.

[0060] Based on this, in the above step S102, the device state data can include the module temperature of the photovoltaic module monitored by each power sensor, switch device state data, protection and safety signals, and energy storage state of charge, etc. The device operation data can include power, voltage, and grid interconnection and scheduling parameters monitored by the power sensor, etc. In addition, the environmental data can include parameters such as wind speed, temperature, irradiance, and human position collected in real time by each environmental sensor.

[0061] Therefore, the low-delay control method of the AI-based optical storage system of the embodiment of the present application realizes comprehensive perception of the system operation state and the external environment state through distributed collection of multi-source heterogeneous data at multiple types of devices and multiple environmental monitoring locations, thereby laying a solid data foundation for subsequent decision instruction generation and collaborative scheduling.

[0062] In some embodiments, the above step S104 can include the following steps: assigning an adaptive communication protocol to each device according to a preset communication protocol adaptation rule; and transmitting data corresponding to the device to the edge computing unit through the communication protocol adapted for each device. Specifically, the communication protocol in the present embodiment can include Modbus protocol, PLC protocol, Zigbee protocol, and Wi-Fi protocol, etc.

[0063] For example, the preset communication protocol adaptation rule can include the following rules: the Modbus protocol is preferentially assigned to the inverter of the photovoltaic device for transmitting data such as photovoltaic output, voltage, current, and maximum power point tracking state; the PLC protocol is assigned to the energy storage device for transmitting data such as energy storage state of charge, battery health state, charging and discharging current / voltage, and fault code; the Zigbee protocol is assigned to the charging pile for transmitting data such as charging power of the charging pile, vehicle battery state of charge, and start / stop state; in addition, the Modbus protocol or Zigbee protocol is assigned to the heat pump device for transmitting data such as heat pump outlet water temperature, compressor frequency, and performance coefficient.

[0064] In addition, the preset communication protocol adaptation rule can also set a dual-path redundant communication for the key device. Specifically, a main communication protocol and a backup communication protocol are set for the key device, and when the main path using the main communication protocol is interrupted or the communication delay exceeds a preset delay threshold, the backup path using the backup communication protocol is switched to for communication, thereby ensuring the continuity and low delay of data transmission. It should be noted that the preset delay threshold can be set according to the communication requirements of the optical storage system, and can be any value selected from 20ms to 30ms. For example, the preset delay threshold can be pre-set to 25ms.

[0065] For example, the preset communication protocol adaptation rule can also include the following rules: the energy storage device is allocated a PLC protocol as the main communication protocol, and a Modbus protocol is allocated as the backup communication protocol; the charging pile is allocated a Zigbee protocol as the main communication protocol, and a Wi-Fi protocol is allocated as the backup communication protocol.

[0066] Therefore, the low-delay control method of the AI-based optical storage system according to the embodiments of the present application can realize automatic identification and optimal connection of different communication standard devices by allocating adaptive communication protocols to each device according to the preset communication protocol adaptation rule, and can ensure the reliability and efficiency of data transmission by transmitting data through the communication protocol adapted to each device, thereby reducing communication failure or delay caused by protocol mismatch and providing a basic guarantee for low-delay control of the system from the communication level.

[0067] In some embodiments, the AI hybrid decision-making model of the present application can include a lightweight prediction model for short-term load prediction, a deep reinforcement learning model for long-term scheduling optimization, and a response model for emergency response. Specifically, the lightweight prediction model can use a lightweight LSTM (Long Short-Term Memory) model. The deep reinforcement learning model can use a PPO (Proximal Policy Optimization) algorithm to train the scheduling strategy of each device. In addition, the response model can use a rule base based on a finite state machine, in which a plurality of emergency event handling instructions corresponding to a plurality of emergency events can be pre-set to quickly match and respond to emergency events.

[0068] Based on this, the step S106 can include the following steps: judging whether the equipment data and the environment data exceed the preset safety threshold; if yes, triggering the response model to directly generate the emergency event handling instruction; if no, using the lightweight prediction model to predict the electricity load of the optical storage system in a future preset time period according to the equipment data, the environment data and the historical operation data of the optical storage system; and using the deep reinforcement learning model to calculate the decision instruction for the operation of each equipment according to the electricity load under multi-target constraints.

[0069] It should be noted that the preset safety threshold can be set according to the safety control needs of the optical storage system. For example, the preset safety threshold can include a preset gradient threshold. The early warning is triggered in the case that the temperature gradient is greater than or equal to the preset gradient threshold. It should be noted that the preset gradient threshold can be determined based on pre-experiment and stored in the edge computing unit of the optical storage system. Specifically, the preset gradient threshold can be selected from any value of 5-15℃ / min. For example, the preset gradient threshold can be pre-set to 10℃ / min. For another example, the preset safety threshold can also include a preset voltage threshold. The early warning is triggered in the case that the grid voltage is less than or equal to the preset voltage threshold. It should be noted that the preset voltage threshold can be determined based on pre-experiment and stored in the edge computing unit of the optical storage system. Specifically, the preset voltage threshold can be set to be lower than the normal power supply voltage.

[0070] In the case that the data exceeds the safety threshold, the response model is triggered to directly generate the emergency event handling instruction, realizing the rapid response under the emergency working condition and avoiding the handling delay caused by time-consuming complex calculation. In the case that the data does not exceed the safety threshold, the lightweight prediction model is first used to predict the future short-term electricity load according to the collected data and the historical operation data, and then the deep reinforcement learning model is used to calculate the decision instruction for the operation of each equipment according to the prediction result under multi-target constraints, thereby realizing the accurate generation of the decision instruction under the normal working condition.

[0071] In addition, the multi-target constraints can include multi-dimensional constraints such as anti-backflow constraint, energy storage state of charge constraint, temperature fluctuation constraint and economic target constraint. Specifically, the constraint conditions of the multi-target constraints can include but are not limited to the following constraint conditions: the grid reverse power is less than or equal to a preset power threshold; the room temperature fluctuation is less than or equal to a preset temperature fluctuation threshold; and the energy storage state of charge is maintained in a preset interval. Specifically, the preset power threshold can be selected from any value of 1-10kW. For example, the preset power threshold can be pre-set to 3kW. The preset temperature fluctuation threshold can be selected from any value of 0.8-15℃. For example, the preset temperature fluctuation threshold can be pre-set to 5℃ / min. The preset interval can be selected from any value interval of 10%-90%. For example, the preset interval can be pre-set to 20%-80%.

[0072] Therefore, the low-delay control method of the AI-based optical storage system according to the embodiments of the present application realizes intelligent identification and hierarchical decision of the system operation state by constructing an AI hybrid decision model containing multiple types of models and a dual protection mechanism of safety response and optimized scheduling, realizes rapid disposal under emergency working conditions and multi-objective optimization under normal working conditions, and thus improves the adaptability and optimization capability of the optical storage system to various operation conditions.

[0073] In some embodiments, each piece of data of the present application is accompanied by a time stamp when communicated, providing a reliable time reference for accurate measurement of delay. Based on this, the step of predicting the communication delay between the edge computing unit and each device in the above step S108 can include the following steps: the edge computing unit calculates the communication round trip time between the edge computing unit and each device according to the time stamp as the prediction result corresponding to each device. It should be noted that the communication round trip time can be used to characterize and predict the communication delay characteristics between the edge computing unit and each device.

[0074] Specifically, the step of calculating the communication round trip time between the edge computing unit and each device according to the time stamp by the edge computing unit can be specifically implemented as: the edge computing unit periodically sends a test instruction with a time stamp to each device; records the sending time of the test instruction, the receiving time of each device receiving the test instruction, the feedback time of each device feeding back the received instruction, and the receiving feedback time of the edge computing unit; calculates the first difference between the receiving time and the sending time and the second difference between the feedback time and the receiving feedback time; calculates the sum of the first difference and the second difference as the single communication round trip time between the edge computing unit and the device, to accurately quantify the communication delay.

[0075] In addition, the step of predicting the communication delay between the edge computing unit and each device in the above step S108 can also include the following steps: the edge computing unit calculates the communication round trip time between the edge computing unit and each device according to the time stamp; and uses a preset sliding window algorithm to smooth the communication round trip time corresponding to each device to obtain the prediction result corresponding to each device.

[0076] Specifically, the preset sliding window algorithm can pre-set a window width and collect the average value of any type of data within the pre-set window width. For example, the moving average of the wind speed within the pre-set window width, so as to eliminate the fluctuation interference caused by the instantaneous wind speed. In the present embodiment, the pre-set window width can be selected from any value of 0.1-5 minutes. In one specific embodiment, the window width can be pre-set to 5 minutes to filter out the noise interference caused by instantaneous jitter.

[0077] Therefore, the low-delay control method of the AI-based optical storage system improves the accuracy and stability of the communication delay prediction by predicting the communication delay between the edge computing unit and each device according to the time stamp attached to the data and the preset sliding window algorithm, provides accurate data basis for subsequent compensation operations, and thus improves the accuracy of the collaborative scheduling of the optical storage system.

[0078] In some embodiments, the step of compensating for the transmission timing of the decision instruction according to the prediction result in the above step S108 can include the following steps: obtaining a planned transmission time of the decision instruction based on the control cycle of the optical storage system; and taking a time point before the planned transmission time and spaced from the planned transmission time by the prediction result as a target transmission time to obtain the compensated transmission timing. It should be noted that the control cycle of the optical storage system can be set according to the real-time demand and resource consumption balance of the optical storage system.

[0079] Therefore, the low-delay control method of the AI-based optical storage system ensures that the instruction transmission is synchronized with the system control demand by obtaining the planned transmission time of the decision instruction based on the control cycle of the optical storage system, takes a time point before the planned transmission time and spaced from the planned transmission time by the prediction result as a target transmission time, realizes accurate offset of the communication delay, so that the decision instruction can arrive at the target device exactly at the planned time, avoids instruction execution lag or advance due to delay, and guarantees the time synchronization of the collaborative control of each device, thereby further improving the precision of the multi-device collaborative scheduling of the optical storage system.

[0080] In some embodiments, in the hardware architecture of the optical storage system, the execution layer can include a control interface arranged on each device. The control interface of each device can be used to receive and execute a control signal transmitted by a communication protocol corresponding to the device to adjust the operating state of the device. In addition, the execution layer can also include a solid-state relay array to realize fast execution of part of emergency handling instructions.

[0081] Based on this, the step of controlling the operating state of each device according to the decision instruction in the above step S110 can include the following steps: converting the decision instruction into a control signal matched with a communication protocol corresponding to each device; and executing the control signal through the control interface of each device to adjust the operating state of the device. It should be noted that adjusting the operating state of the device can include adjusting power, frequency or switching state, etc., for example, adjusting the power generation power of the photovoltaic device to adapt to the power load and adjusting the charging frequency of the energy storage device to protect the battery life.

[0082] Therefore, the low-delay control method of the AI-based optical storage system according to the embodiments of the present application ensures that each device can accurately identify and execute the instruction by converting the decision instruction into a control signal matching the communication protocol of each device, and executes the control signal through the control interface of the device to achieve precise control, thereby improving the pertinence and effectiveness of device control, and further ensuring the stable operation and collaborative scheduling of multiple types of devices in the optical storage system.

[0083] In some embodiments, after the step S110 described above, the low-delay control method of the AI-based optical storage system according to the present application can further include the following steps: uploading data to a cloud platform connected with the optical storage system according to a first preset period; training model parameters of the deep reinforcement learning model by the cloud platform according to the received data within a second preset period; and issuing the trained model parameters to the edge computing unit to update the model parameters of the deep reinforcement learning model. It should be noted that the second preset period is greater than the first preset period to ensure that the training operation of the cloud platform does not interfere with the real-time low-delay control of the edge computing unit. In addition, the first preset period can be set as an integer multiple of the control period to ensure that the cloud platform can continuously obtain the operation data of the optical storage system while avoiding high-frequency uploading of the resources of the edge computing unit.

[0084] Specifically, the step of training the model parameters of the deep reinforcement learning model by the cloud platform according to the received data within the second preset period can be specifically performed as the following steps: constructing a training model with the same structure as the deep reinforcement learning model in the edge computing unit on the cloud platform; integrating the data received within the second preset period with the historical operation data stored by the cloud platform to construct a training sample set for model training; iteratively optimizing the model parameters of the training model based on the training sample set until the maximum iteration number or optimization convergence is reached; and taking the model parameters obtained by the last iteration optimization as the trained model parameters.

[0085] Therefore, the low-delay control method of the AI-based optical storage system according to the embodiments of the present application trains the parameters of the deep reinforcement learning model by the cloud platform according to the received data within the second preset period, utilizes the full historical data reserve and computing power advantage of the cloud platform to globally optimize the model parameters, and updates the trained model parameters to the edge computing unit to realize the long-term and continuous optimization of the AI hybrid decision model, avoids the adaptability decline caused by the solidification of the model parameters, thereby realizing the collaborative control of the edge computing unit and the cloud, and further ensuring the continuous low-delay control of the optical storage system.

[0086] In one specific embodiment, taking the control scheme under the power grid outage condition as an example, the specific control process of the low-delay control method of the AI-based optical storage system according to the present application is as follows:

[0087] Step one, collect the equipment data and environmental data of multiple types of equipment in the light storage system, the equipment data including the grid voltage;

[0088] Step two, transmit each item of data to the edge computing unit of the light storage system according to the preset communication protocol adaptation rule;

[0089] Step three, determine whether the data exceeds the preset safety threshold;

[0090] Step four, in the case where the grid voltage is less than or equal to the preset voltage threshold, trigger the response model to directly generate emergency event handling instructions under the power grid outage condition;

[0091] Step five, according to the emergency event handling instructions under the power grid outage condition, switch to the off-grid mode, increase the discharge power of the energy storage equipment and reduce the operating frequency of the heat pump equipment.

[0092] Therefore, the low-delay control method of the AI-based light storage system of the embodiment of the application effectively reduces the overall control delay by triggering the response model when the power grid is determined to be out of power, directly generating emergency event handling instructions under the power grid outage condition, and guarantees the key load power supply rate, which helps to realize the stable operation of the light storage system.

[0093] In another specific embodiment, taking the control scheme under the photovoltaic surplus condition as an example, the specific control process of the low-delay control method of the AI-based light storage system of the application is as follows:

[0094] Step one, collect the equipment data and environmental data of multiple types of equipment in the light storage system, the equipment data including the photovoltaic output;

[0095] Step two, transmit each item of data to the edge computing unit of the light storage system according to the preset communication protocol adaptation rule;

[0096] Step three, determine whether the data exceeds the preset safety threshold;

[0097] Step four, in the case where the data does not exceed the preset safety threshold, use a lightweight prediction model to predict the electricity load of the light storage system in a future preset time period according to the equipment data, the environmental data and the historical operation data of the light storage system;

[0098] Step five, in the case where the photovoltaic output exceeds the electricity load, use a deep reinforcement learning model to calculate the decision instruction under the photovoltaic surplus condition according to the electricity load under multi-objective constraints;

[0099] Step six, according to the decision instruction under the photovoltaic surplus condition, preferentially increase the charging power of the energy storage equipment, increase the operating frequency of the heat pump equipment and increase the charging frequency of the vehicle battery.

[0100] Therefore, the low-delay control method of the AI-based optical storage system of the embodiment of the present application realizes hierarchical consumption of excess electric energy by preferentially increasing the charging power of the energy storage device, increasing the operation frequency of the heat pump device, and increasing the charging frequency of the vehicle battery in the case where it is determined that the photovoltaic output exceeds the power consumption load, thereby reducing the photovoltaic light abandonment rate and helping to realize reasonable allocation of resources in the optical storage system.

[0101] Figure 2 is a flowchart of a control example of the low-delay control method of the AI-based optical storage system according to an embodiment of the present application. As shown in Figure 2 the flow of the low-delay control method of the AI-based optical storage system of the embodiment includes the following steps:

[0102] Step S202, collect device data and environmental data of multiple types of devices in the optical storage system.

[0103] Step S204, transmit each item of data to the edge computing unit of the optical storage system according to a preset communication protocol adaptation rule.

[0104] Step S206, determine whether the data exceeds a preset safety threshold. It should be noted that the data in this step includes device data and environmental data. In the case where the data exceeds the preset safety threshold, step S208 is performed. In the case where the data does not exceed the preset safety threshold, step S210 is performed.

[0105] Step S208, trigger a response model to directly generate an emergency event handling instruction. After this step, step S214 is continued.

[0106] Step S210, use a lightweight prediction model to predict the power consumption load of the optical storage system in a future preset time period according to the device data, the environmental data, and the historical operation data of the optical storage system.

[0107] Step S212, use a deep reinforcement learning model to calculate a decision instruction for the operation of each device under multi-objective constraints according to the power consumption load. After this step, step S214 is continued.

[0108] Step S214, calculate the communication round trip time between the edge computing unit and each device according to the time stamp.

[0109] Step S216, use a preset sliding window algorithm to smooth the communication round trip time corresponding to each device to obtain a prediction result corresponding to each device.

[0110] Step S218, obtain the planned sending time of the decision instruction based on the control period of the optical storage system.

[0111] Step S220, taking the time point before the scheduled sending time and interval the predicted result as the target sending time, to obtain the compensated sending opportunity.

[0112] Step S222, according to the compensated sending opportunity, the decision instruction is issued to each device through the corresponding communication protocol.

[0113] Step S224, according to the decision instruction, the running state of each device is adjusted to realize the low delay control of the optical storage system. At this point, this process ends. Of course, after step S224, it can also return to the above step S202 to restart this process.

[0114] Therefore, the low delay control method of the AI-based optical storage system according to the embodiment of the application transmits the collected data through the communication protocol adaptation rule, and compensates the sending opportunity of the decision instruction generated by the edge computing unit according to the communication delay, solves the instruction issuing lag problem caused by the communication delay in the prior art, ensures the timing accuracy of the decision instruction for each device, and thus ensures the collaborative control of the optical storage system for multiple types of devices, and further realizes the low delay collaborative scheduling of the optical storage system.

[0115] Further, the low delay control method of the AI-based optical storage system according to the embodiment of the application predicts the communication delay between the edge computing unit and each device according to the time stamp attached to the data and the preset sliding window algorithm, improves the accuracy and stability of the communication delay prediction, provides accurate data basis for the subsequent compensation operation, and thus improves the accuracy of the collaborative scheduling of the optical storage system.

[0116] Further, the low delay control method of the AI-based optical storage system according to the embodiment of the application, by constructing an AI hybrid decision model containing multiple types of models and a dual protection mechanism of safe response and optimized scheduling, realizes intelligent identification and hierarchical decision of the system running state, realizes rapid disposal under emergency working condition and multi-objective optimization under normal working condition, and thus improves the adaptability and optimization ability of the optical storage system to multiple running conditions.

[0117] The embodiment also provides a computer program product 10, a computer readable storage medium 20 and a computer device 30. Figure 3 is a schematic diagram of a computer program product according to an embodiment of the application. Figure 4 is a schematic diagram of a computer readable storage medium according to an embodiment of the application. Figure 5 is a schematic block diagram of a computer device according to an embodiment of the application.

[0118] The computer program product 10 comprises a computer program 11 which, when executed by the processor 32, implements the steps of any of the above-described AI-based low-latency control methods for optical storage systems. The computer-readable storage medium 20 has stored thereon the above-described computer program 11 which, when executed by the processor 32, implements the steps of any of the above-described AI-based low-latency control methods for optical storage systems. The computer device 30 can comprise a memory 31, a processor 32 and the computer program 11 stored on the memory 31 and running on the processor 32.

[0119] The computer program 11 for carrying out the operations of the present application can be in an assembly language, an Instruction Set Architecture (ISA) language, machine language, a machine dependent instruction, a microcode, a firmware instruction, a state setting data, a configuration data of an integrated circuit, or a source code or an object code written in any combination of one or more programming languages and process programming languages.

[0120] The computer program 11 can be executed in whole or in part on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present application.

[0121] For the description of the present embodiment, the computer program product 10 is a relevant product containing the computer program 11.

[0122] For the purposes of this description, the computer readable storage medium 20 is a tangible device that can retain and store computer- programmable instructions 11. The computer readable storage medium 20 can be any available medium or device that can be accessed by a general purpose or special purpose computer system. By way of example, and not limitation, computer readable storage media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium or device from which a computer can read. For the purposes of this description, the computer readable storage medium 20 can be any tangible device that can store the computer program 11 for use by or in connection with the instruction execution system, apparatus, or device.

[0123] To thus far, those skilled in the art will appreciate that although exemplary embodiments of the application have been shown and described herein, various modifications and changes can be made without departing from the spirit and scope of the application. Accordingly, all such modifications and changes are intended to be included within the scope of the application.

Claims

1. A low-latency control method for an AI-based photovoltaic energy storage system, characterized in that... The method comprises the following steps: Collecting device data and environmental data of multiple types of devices in the optical storage system; According to the preset communication protocol adaptation rule, each item of data is transmitted to the edge computing unit of the optical storage system, wherein the edge computing unit is equipped with a neural network processing unit accelerated AI chip, the AI chip is used to carry and run an AI hybrid decision model, the AI hybrid decision model includes a lightweight prediction model for short-term load prediction, a deep reinforcement learning model for long-term scheduling optimization, and a response model for emergency response; Determine whether the device data and the environmental data exceed the preset safety threshold; If yes, trigger the response model to directly generate emergency handling instructions; If not, use the lightweight prediction model to predict the electricity load of the optical storage system in a future preset time period according to the device data, the environmental data and the historical operation data of the optical storage system; Using the deep reinforcement learning model, the decision instructions for each device are calculated under multi-objective constraints according to the electricity load; Predict the communication delay between the edge computing unit and each device, and compensate the sending time of the decision instructions according to the prediction result; According to the compensated sending time, the decision instructions are sent to each device through the corresponding communication protocol, and the running state of each device is adjusted according to the decision instructions to realize the low delay control of the optical storage system; According to the first preset period, the data is uploaded to the cloud platform connected with the optical storage system; The cloud platform trains the model parameters of the deep reinforcement learning model according to the received data in the second preset period; The model parameters obtained by training are sent to the edge computing unit to update the model parameters of the deep reinforcement learning model. 2.The low-latency control method of an AI-based optical storage system according to claim 1, wherein, Each item of data is attached with a time stamp when communicating; and The step of predicting the communication delay between the edge computing unit and each device comprises: The edge computing unit calculates the communication round trip time between each device according to the time stamp; Using a preset sliding window algorithm, the communication round trip time corresponding to each device is smoothed to obtain the prediction result corresponding to each device.

3. The low delay control method of the AI-based optical storage system according to claim 2, wherein The step of compensating the sending time of the decision instructions according to the prediction result comprises: Based on the control cycle of the optical storage system, the planned sending time of the decision instructions is obtained; The time point before the planned sending time and interval from the planned sending time by the prediction result is taken as the target sending time to obtain the compensated sending time.

4. The low delay control method of the AI-based optical storage system according to claim 1, wherein The step of transmitting each item of data to the edge computing unit of the optical storage system according to the preset communication protocol adaptation rule comprises: According to the preset communication protocol adaptation rule, an adapted communication protocol is assigned to each of the devices; The data corresponding to each of the devices is transmitted to the edge computing unit through the adapted communication protocol of each of the devices. 5.The low-latency control method of an AI-based optical storage system according to claim 4, wherein, The multiple types of devices include at least one of a photovoltaic device, an energy storage device, a charging device, and a heat pump device; and The step of controlling the operation state of each of the devices according to the decision instruction comprises: The decision instruction is converted into a control signal matched with the communication protocol corresponding to each of the devices; The control signal is executed through the control interface of each of the devices to adjust the operation state of the device.

6. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the low-delay control method of the AI-based optical storage system according to any one of claims 1 to 5.

7. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the steps of the low-delay control method of the AI-based optical storage system according to any one of claims 1 to 5.

8. A computer device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1-7. The processor executes the computer program to implement the steps of the low-delay control method of the AI-based optical storage system according to any one of claims 1 to 5.

Images