Wind and light storage multi-source AGC control and AI servitization integration system and method based on 4diac

Abstract

The invention discloses a 4diac-based wind and light storage multi-source AGC control and AI servitization integration system and method. The system comprises a 4diac control layer, a unified interface layer, an AI service layer, an external EMS / SCADA platform and a multi-source field device. The method comprises the following steps: constructing a control link configuration service interface layer consisting of a plurality of event-triggered functional units in a 4diac framework, submitting a data request containing operation state information to an external service through an asynchronous mechanism in a control period, receiving prediction or optimization feedback, injecting a feedback result into a functional unit link, and transmitting the feedback result to the external service. And the power distribution module is used for correcting deviation or control strategies, completing a power distribution process based on event transfer among the functional units and outputting a multi-source adjustment instruction. According to the invention, the service interface is connected with the external artificial intelligence model, the field state data is sent to the AI service in the control period, and the prediction or optimization result is injected into the adjusting link, so that the feed-forward correction of the power deviation is realized.

Description

TECHNICAL FIELD

[0001] The application relates to the field of multi-source coordinated regulation, and in particular to a wind-solar-storage multi-source AGC control and AI service integration system and method based on 4diac. BACKGROUND

[0002] With the continuous expansion of renewable energy such as wind power and photovoltaic power, the operation state of the station presents the characteristics of large power fluctuation, diversified equipment combination and complex operation strategy. The grid connection point often needs to adjust the power in real time according to the dispatching instruction, which makes the AGC regulation capacity a key link to maintain the stable operation of the system. However, in the traditional control system architecture, the control logic usually relies on PLC, DCS or EMS for centralized implementation, and the control structure is fixed and the expansion capacity is limited, which is not conducive to the coordinated regulation of different types of equipment, and it is also difficult to quickly embed new algorithms or prediction models into the control link.

[0003] In the wind-solar-storage collaborative operation scene, the types of equipment are various, and the running conditions are obviously different. The control system needs to handle various factors such as equipment restrictions, dynamic ranges and response speed differences. This makes the AGC regulation process involve multi-level data, complex state judgment and multi-device joint optimization. Although AI technology has obvious advantages in prediction and optimization tasks, existing engineering solutions often lack a standardized mechanism to decouple and integrate AI services with on-site control systems, so that the prediction or optimization results can directly and effectively affect the real-time regulation process.

[0004] The IEC 61499 standard uses an event-driven and distributed function block model to organize the control process, which can decompose the control logic into multiple independent functional units and implement combined execution through an event chain structure. 4diac, as an open source implementation of the standard, provides a reusable and distributed configuration control environment. However, there is currently a lack of engineering practice solutions that combine 4diac and AI services for wind-solar-storage system AGC regulation, so a method is needed that can realize the modularization of the control link, be expandable and support the access of external intelligent algorithms. SUMMARY

[0005] The purpose of the present application is to provide a wind-solar-storage multi-source AGC control and AI service integration system and method based on 4diac; for splitting the AGC control logic into multiple independent units, completing the series regulation process through an event-driven mechanism, and accessing external AI models to the control link in a service manner through a functional interface, so that the prediction value can directly affect the regulation strategy.

[0006] Technical solution: The wind and light storage multi-source AGC control and AI service integration system based on 4diac provided by the application comprises a 4diac control layer, a unified interface layer, an AI service layer, and an external EMS / SCADA platform and multi-source field equipment; the 4diac control layer realizes AGC control logic through a functional block network, and the AGC control functional block network comprises a deviation calculation functional block Deviation_Calc, an AI feedforward fusion functional block AI_Fusion, a power constraint functional block Limit_Handler, an adjustable margin calculation functional block Avail_Calc, a multi-source power distribution functional block Dispatch_Calc and a control instruction output functional block Cmd_Output which are sequentially coupled through events.

[0007] Further, the deviation calculation functional block Deviation_Calc receives the grid point AGC target power Grid_Target and the grid point actual measured power Grid_Actual as data inputs, and outputs an AGC deviation ΔP; the AI feedforward fusion functional block AI_Fusion receives the AGC deviation ΔP and a prediction result AI_Predicted from AI service, and fuses the two to obtain a feedforward compensated deviation ΔP_FF.

[0008] Further, the power constraint functional block Limit_Handler receives the rated power, the current power and the operation constraint of each photovoltaic unit, wind turbine and energy storage unit, and outputs the effective power adjustment range of each device; the adjustable margin calculation functional block Avail_Calc receives relevant parameters of multiple devices in an array form, determines the actual device quantity through a Count input, performs traversal calculation on the array, and obtains an up and down power margin array of each device.

[0009] Further, the multi-source power distribution functional block Dispatch_Calc receives ΔP_FF and the adjustable margin array, distributes the total adjustment power among photovoltaic, wind power and energy storage according to a preset power distribution strategy or a collaborative optimization result returned by AI service, generates active power instructions of each device, and the control instruction output functional block Cmd_Output converts the instructions into an external communication protocol format and sends the instructions to a field equipment controller through Modbus, IEC 60870-5-104 or other interfaces supported by 4diac.

[0010] Further, the AI service layer is deployed on an edge server or a cloud platform, and internally integrates a photovoltaic output prediction model, a wind power output prediction model and an energy storage optimization scheduling model; after receiving a request, the AI service calls the corresponding model for inference, generates a short-term predicted power or an optimized allocation scheme, and returns the result as a JSON response to the unified interface layer.

[0011] Further, the AGC control function block network is coupled by an event chain, the event output of the deviation calculation function block triggers the event input of the AI feedforward fusion function block and the power constraint function block, the event output of the power constraint function block and the adjustable margin calculation function block triggers the event input of the multi-source power distribution function block, and the event output of the multi-source power distribution function block triggers the event input of the control instruction output function block.

[0012] The wind-solar-storage multi-source AGC control and AI service integration method based on 4diac provided by the application comprises the following steps:

[0013] S1: a control link composed of a plurality of event-triggered function units is constructed in the 4diac framework, and the function units include a deviation analysis unit, a limiting check unit, an adjustable interval calculation unit and a power coordination distribution unit in sequence;

[0014] S2: a service interface layer is configured, and the interface layer exchanges data with an external algorithm service or an energy management platform through a service interface function block;

[0015] S3: in a control cycle, a data request containing running state information is submitted to the external service through an asynchronous mechanism, and a prediction or optimization feedback is received;

[0016] S4: the feedback result is injected into the function unit link for correcting the deviation or the control strategy;

[0017] S5: based on the completion of the power distribution process through the event transmission between the function units, a multi-source regulation instruction is output.

[0018] Further, the service interface layer in step S2 adopts an asynchronous non-blocking thread model, does not block the 4diac main event execution thread after initiating an AI service call, and triggers a preset event after receiving the response of the AI algorithm service, and writes the parsed prediction result or optimization result into the data input port of the corresponding function block.

[0019] Further, the result provided by the AI algorithm service includes:

[0020] a short-term photovoltaic output power prediction result;

[0021] a short-term wind power output prediction result;

[0022] Optimal adjustment strategy under energy storage charging and discharging power and state constraints

[0023] Power distribution optimization result of multi-device coordination under AGC instruction constraints.

[0024] Further, the step S4 functional unit network adopts an event chain mode to trigger, so that the deviation calculation, constraint processing and power distribution modules are sequentially executed through event coupling.

[0025] Advantages: Compared with the prior art, the present application has the following significant advantages: the control link is constructed in a functional block composite structure, which is convenient for splitting, reuse and migration; the AI prediction module is accessed through a service interface, without the need to change the main control logic to complete the extension; the asynchronous triggering mechanism avoids blocking the main control thread, ensuring the stability of the AGC adjustment period; the array processing mode enables the system to process different sizes of device clusters; the prediction information can directly participate in the adjustment amount calculation, improving the power tracking ability and device coordination level; and the system is compatible with the EMS or dispatching system, and can be used as an execution unit of the upper layer control strategy. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The figure is a schematic diagram of the overall architecture of the AGC control and AI service integration system of the present application.

[0027] Figure 2 The figure is a schematic diagram of the AGC control functional block network structure of the present application.

[0028] Figure 3 The figure is a schematic diagram of the communication process between the unified interface layer and the AI service of the present application.

[0029] Figure 4 The figure is a schematic diagram of the AGC control process based on AI feedforward compensation of the present application.

[0030] Figure 5 The figure is a schematic diagram of the array and batch processing functional block structure of the present application. DETAILED DESCRIPTION

[0031] The technical solutions of the present application will be further described below with reference to the accompanying drawings.

[0032] To solve the problems of high structural coupling, weak scalability and lack of prediction assistance in multi-source AGC adjustment, the present application proposes a coordinated adjustment architecture with 4diac as the core. The architecture consists of three parts:

[0033] (1) AGC control link based on event-driven mechanism

[0034] Several independent functional modules are designed in the 4diac platform, including deviation analysis module, prediction fusion module, power boundary identification module, adjustable interval calculation module and power coordination module. The modules are executed through event connection, and data is transmitted through ports, so that the control link can be triggered as needed according to the running state, reducing unnecessary computational burden.

[0035] 4diac control structure design

[0036] In the 4diac control layer, each module is implemented in the form of IEC 61499 function block, and the logic of each module is as follows:

[0037] Deviation Block: reads the real-time power of the grid point and the command power, and calculates the basic adjustment deviation ΔP.

[0038] Prediction Fusion Block: combines ΔP with the predicted power or optimization quantity returned by AI to obtain ΔP′ after feedforward compensation.

[0039] Limit Block: determines the adjustable range based on the rated power, real-time output value, SOC, and ramping constraints of the device.

[0040] Margin Block: calculates the up and down adjustment intervals of each device through array traversal.

[0041] Dispatch Block: allocates the adjustment power of each device based on ΔP′ and the adjustable interval.

[0042] These modules are executed sequentially through event chains, making the operation within the AGC adjustment period more explicit and allowing for segmented debugging.

[0043] (2) Service interface layer for AI model

[0044] The service interface module is deployed outside the control layer, which sends real-time operation data to external AI inference services through an asynchronous mechanism, including photovoltaic / wind power output, environmental parameters, grid point load, etc. The module uses JSON data format and HTTP communication method, which can maintain loose coupling with AI services deployed on the edge or cloud.

[0045] The service interface module has the following functions:

[0046] 1) Pack the key parameters of the current control period into JSON data;

[0047] 2) sent to the AI server through an asynchronous HTTP client;

[0048] 3) wait for the return without blocking the main control thread;

[0049] 4) write the predicted value, optimized value, etc. to the corresponding function block input.

[0050] 5) This way allows the AI model to be upgraded or replaced online without the need to modify the main control link.

[0051] AI service logic

[0052] The AI model generally consists of a photovoltaic / wind power prediction model and an energy storage strategy optimization model, and runs on an edge server or in the cloud to provide inference services through a REST API. The model can be based on historical power data, environmental variables (irradiance, temperature, wind speed, etc.), scheduling instruction trends, and device remaining capacity or SOC;

[0053] The AI output results are used to correct the next cycle deviation in advance, provide device adjustment priorities, provide energy storage charging and discharging strategies, and provide collaborative optimization allocation schemes.

[0054] (3) Unified adjustment mechanism for multiple source devices

[0055] In view of the characteristics of a large number of wind-solar-storage devices and large differences in operating characteristics, the application adopts an arrayed data processing method. Real-time output, power limit range, adjustable margin, and other parameters of each device are input in array form, and batch processing and adjustment are completed by a loop logic. This structure is suitable for different scales of stations and does not need to change the control logic due to the addition or reduction of devices.

[0056] (4) AI feedforward compensation and adjustment optimization

[0057] After receiving the data, the external AI service performs photovoltaic / wind power short-term prediction or energy storage adjustment strategy optimization, and returns the predicted value or optimization suggestion to the 4diac control link. The AGC deviation is corrected by the predicted value, which can effectively reduce the error caused by random fluctuations and make the power distribution result more consistent with the device operation trend.

[0058] Through the above design, the application constructs a modular, high real-time, and scalable wind-solar-storage multi-source AGC adjustment system architecture, which can adapt to different device combinations and operating conditions and support the rapid introduction of new algorithm models.

[0059] For example Figure 1As shown, the system overall architecture of the present application includes: a 4diac control layer, a unified interface layer, an AI service layer, and an external EMS / SCADA platform and multi-source field devices. Among them, the multi-source field devices include photovoltaic arrays, wind turbine generators and energy storage systems.

[0060] The 4diac control layer is based on the IEC 61499 standard and realizes the AGC control logic through a function block network. The AGC control function block network includes a deviation calculation function block Deviation_Calc, an AI feedforward fusion function block AI_Fusion, a power constraint function block Limit_Handler, an adjustable margin calculation function block Avail_Calc, a multi-source power distribution function block Dispatch_Calc and a control instruction output function block Cmd_Output coupled by events in turn, as shown in FIG. 2.

[0061] The data input of the deviation calculation function block Deviation_Calc includes the grid point AGC target power Grid_Target and the grid point actual measured power Grid_Actual, and the data output is the AGC deviation ΔP. The AI feedforward fusion function block AI_Fusion receives the AGC deviation ΔP and the prediction result AI_Predicted from the AI service, and fuses the two to obtain the deviation ΔP_FF after feedforward compensation.

[0062] The power constraint function block Limit_Handler receives the rated power, current power and operating constraints (such as upper and lower limits of SOC, climb rate limit) of each photovoltaic unit, wind turbine generator and energy storage unit, and outputs the effective power adjustment range of each device. The adjustable margin calculation function block Avail_Calc receives the relevant parameters of multiple devices in array form, determines the actual number of devices through the Count input, and iteratively calculates the array to obtain the up and down power margin array of each device.

[0063] The multi-source power distribution function block Dispatch_Calc receives ΔP_FF and the adjustable margin array, and distributes the total adjustment power among photovoltaic, wind power and energy storage according to the preset power distribution strategy or the collaborative optimization result returned by the AI service, to generate the active power instruction of each device. The control instruction output function block Cmd_Output converts the instruction into an external communication protocol format, and sends it to the field device controller through the Modbus, IEC 60870-5-104 or other interfaces supported by 4diac.

[0064] The unified interface layer is located between the 4diac control layer and the AI service layer, and realizes data interaction with external AI services and EMS platforms through a service interface function block (SIFB), as shown in FIG. 3. The SIFB internally includes a JSON serialization / deserialization module, an HTTP request module, and an asynchronous thread scheduling module, which are used to construct requests, send requests, and receive responses.

[0065] During the AGC control cycle, when the function block network executes to a certain preset event, the SIFB is triggered to initiate an AI service call. The SIFB creates an asynchronous thread without blocking the 4diac main thread, packages the photovoltaic irradiance, environmental temperature, current active power, wind speed, SOC, and other operating data into a JSON request, and sends it to the AI service side REST interface through the HTTP POST method.

[0066] The AI service layer is deployed on an edge server or a cloud platform, and internally integrates photovoltaic output prediction models, wind power output prediction models, and energy storage optimization scheduling models. After receiving the request, the AI service calls the corresponding model for inference, generates short-term predicted power or an optimized allocation scheme, and returns the result to the unified interface layer as a JSON response.

[0067] After receiving the response in the asynchronous thread, the SIFB parses the JSON data, writes the photovoltaic predicted power PV_Pred, the wind power predicted power WT_Pred, and the energy storage charge and discharge recommended power ES_Cmd into the data input end of the corresponding function block, and triggers the subsequent AI_Fusion function block for execution through event output.

[0068] The AGC control flow is shown in FIG. 4, including real-time acquisition of grid-connected point power and device status → calculation of AGC deviation ΔP → calling of AI services to obtain prediction results → prediction feedforward compensation to obtain ΔP_FF → power constraint and adjustable margin calculation → multi-source power distribution → output of device control instructions → return to the control loop.

[0069] Application scenario examples with specific data

[0070] In a zero-carbon park example, the rated capacity of the grid-connected point is 30 MW, including 10 MW of photovoltaic power, 15 MW of wind power, and 5 MW / 10 MWh of energy storage systems. The AGC control cycle is 2 seconds.

[0071] At a certain moment, the dispatching issues a grid point target power Grid_Target of 20 MW, and the measured grid power Grid_Actual is 17 MW. The AGC deviation ΔP calculated by the deviation calculation function block Deviation_Calc is 3 MW.

[0072] The current total PV output is 8 MW, the total wind power output is 9 MW, the energy storage is at 60% SOC, and the charge and discharge power limit is ±4 MW. The SIFB collects grid point power, PV / wind power output, irradiance, wind speed, temperature, and SOC data, and sends them to the AI service in JSON format:

[0073] {

[0074] "grid_target": 20.0,

[0075] "grid_actual": 17.0,

[0076] "pv_power": 8.0,

[0077] "wt_power": 9.0,

[0078] "es_soc": 60.0,

[0079] "irr": 750.0,

[0080] "wind_speed": 7.5,

[0081] "temp": 30.0

[0082] }

[0083] The AI service internally has a PV prediction model and a wind power prediction model trained based on historical data. The models predict that the PV can be increased by 1 MW and the wind power can be increased by 0.5 MW in the next AGC cycle, and according to the SOC and ramping constraints, the energy storage is suggested to discharge 1.5 MW, thereby giving the following prediction and optimization results:

[0084] {

[0085] "pv_pred": 9.0,

[0086] "wt_pred": 9.5,

[0087] "es_suggest": 1.5

[0088] }

[0089] The AI_Fusion function block feeds forward the AGC deviation ΔP = 3 MW and the prediction result to obtain a corrected deviation ΔP_FF. For example, it is predicted that the natural rise is 1.5 MW according to the prediction value, and only 1.5 MW needs to be additionally adjusted. The Avail_Calc function block calculates the up-regulation margin array of 10 photovoltaic units, such as [0.3, 0.4, …] MW, and the up-regulation margin array of several wind turbines according to the current output and the upper limit of each photovoltaic subarray. The available discharge margin of the energy storage is 3 MW according to the SOC.

[0090] The Dispatch_Calc multi-source power distribution function block adopts the following strategy: 50% of the adjustment amount is borne by the energy storage, and the rest is distributed according to the adjustable margin proportion of photovoltaic and wind power. For ΔP_FF = 1.5 MW, the energy storage bears 0.75 MW discharge, and the remaining 0.75 MW is distributed to the corresponding equipment according to the adjustable margin proportion of photovoltaic and wind power. Finally, the active power instruction array of each photovoltaic unit, wind turbine unit and energy storage unit is formed, and the Cmd_Output function block sends it to the field control device through the IEC 60870-5-104 protocol.

[0091] As can be seen from the above, the method of the present application can introduce AI prediction feedforward and collaborative optimization results in AGC control, improve the adjustment accuracy and renewable energy utilization rate under the premise of meeting the device constraints.

[0092] 3. Implementation of array and batch processing function block

[0093] As shown in Figure 5 , the present application designs an array and batch processing function block in 4diac to support unified processing of multiple devices. The array input interface receives PV_Power[i] and the corresponding upper and lower limit arrays, and PV_Count is used to indicate the number of devices participating in adjustment. The batch processing unit internally traverses i = 0…Count−1 through a loop structure, performs multiplication, subtraction or upper / lower limit clipping operations on each element, and writes the result to the output array PV_Adj_Val[i].

[0094] In order to ensure the runtime expansion capability, the upper limit of the array length in the present application is configurable, for example, supporting a maximum of 50 or 200 devices. In actual operation, the number of effective devices is controlled through Count, without the need to modify the control logic to expand or reduce the device size.

[0095] Those skilled in the art can understand that equivalent transformations of the AGC function block network structure, the AI service calling manner and the multi-source power distribution strategy without departing from the concept of the present application all fall within the protection scope of the present application.

Claims

1. A 4Diac-based multi-source AGC control and AI service integration system for wind, solar, and energy storage, characterized in that, It includes a 4diac control layer, a unified interface layer, an AI service layer, and external EMS / SCADA platforms and multi-source field devices. The 4diac control layer implements AGC control logic through a function block network. The AGC control function block network includes, in sequence, an event-coupled deviation calculation function block Deviation_Calc, an AI feedforward fusion function block AI_Fusion, a power constraint function block Limit_Handler, an adjustable margin calculation function block Avail_Calc, a multi-source power allocation function block Dispatch_Calc, and a control command output function block Cmd_Output.

2. The integrated system for multi-source AGC control and AI service based on 4diac wind, solar, and energy storage according to claim 1, characterized in that, The Deviation_Calc function block takes the grid connection point AGC target power Grid_Target and the grid connection point measured power Grid_Actual as its data inputs and outputs the AGC deviation ΔP. The AI_Fusion function block receives the AGC deviation ΔP and the prediction result AI_Predicted from the AI ​​service and fuses the two to obtain the feedforward compensated deviation ΔP_FF.

3. The integrated system for multi-source AGC control and AI service based on 4diac wind, solar, and energy storage according to claim 1, characterized in that, The power constraint function block Limit_Handler receives the rated power, current power, and operating constraints of each photovoltaic unit, wind turbine unit, and energy storage unit, and outputs the effective power adjustment range of each device. The adjustable margin calculation function block Avail_Calc receives the relevant parameters of multiple devices in array form, determines the actual number of devices by inputting Count, and iterates through the array to calculate the upper and lower power margin arrays of each device.

4. The integrated system for multi-source AGC control and AI service based on 4diac wind, solar, and energy storage according to claim 1, characterized in that, The multi-source power allocation function block Dispatch_Calc receives ΔP_FF and the adjustable margin array, and allocates the total regulated power among photovoltaic, wind power and energy storage according to the preset power allocation strategy or the collaborative optimization results returned by the AI ​​service. It generates active power commands for each device, and the control command output function block Cmd_Output converts the commands into an external communication protocol format and sends them to the field device controller through Modbus, IEC 60870-5-104 or other interfaces supported by 4diac.

5. The integrated system for multi-source AGC control and AI service based on 4diac wind, solar, and energy storage according to claim 1, characterized in that, The AI ​​service layer is deployed on an edge server or cloud platform and integrates photovoltaic output prediction model, wind power output prediction model and energy storage optimization scheduling model. After receiving a request, the AI ​​service calls the corresponding model to perform inference, generates short-term predicted power or optimized allocation scheme, and encapsulates the result as a JSON response to return to the unified interface layer.

6. The integrated system for multi-source AGC control and AI service based on 4diac wind, solar, and energy storage according to claim 1, characterized in that, The AGC control function block network is coupled using an event chain. The event output of the deviation calculation function block triggers the event input of the AI ​​feedforward fusion function block and the power constraint function block. The event output of the power constraint function block and the adjustable margin calculation function block triggers the event input of the multi-source power allocation function block. The event output of the multi-source power allocation function block triggers the event input of the control command output function block.

7. A method for integrating multi-source AGC control and AI service-based management of wind, solar, and energy storage based on 4diac, characterized in that, Includes the following steps: S1: Construct a control link within the 4diac framework, consisting of multiple event-triggered functional units, which in turn include a deviation analysis unit, a limit verification unit, an adjustable range calculation unit, and a power collaborative allocation unit. S2: Configure the service interface layer, which exchanges data with external algorithm services or energy management platforms through service interface function blocks; S3: During the control cycle, submit a data request containing running status information to the external service through an asynchronous mechanism, and receive prediction or optimization feedback; S4: Inject the feedback result into the functional unit link to correct deviations or control strategies; S5: Completes the power allocation process based on the event transmission between functional units and outputs multi-source regulation commands.

8. The method for integrating wind, solar, and energy storage multi-source AGC control and AI service based on 4diac according to claim 7, characterized in that, The service interface layer in step S2 adopts an asynchronous non-blocking thread model. After initiating the AI ​​service call, it does not block the 4diac main event execution thread. After receiving the response of the AI ​​algorithm service, it triggers a preset event and writes the parsed prediction result or optimization result into the data input port of the corresponding function block.

9. The method for integrating wind, solar, and energy storage multi-source AGC control and AI service based on 4diac according to claim 8, characterized in that, The results provided by the AI ​​algorithm service include: Short-term forecast results for photovoltaic output power; Short-term forecast results for wind power output; Optimization adjustment strategy for energy storage charging and discharging power and state constraints; The results of power allocation optimization for multi-device collaboration under AGC command constraints.

10. The method for integrating wind, solar, and energy storage multi-source AGC control and AI service based on 4diac according to claim 7, characterized in that, The functional unit network in step S4 is triggered by an event chain, so that the deviation calculation, constraint processing and power allocation modules are executed sequentially through event coupling.

Images