Optical storage and charging scene data stream processing and collaborative decision method

Abstract

The present application relates to a kind of light storage fills scene data stream processing and collaborative decision-making method, it is related to energy management technical field, including: light storage fills multi-source data acquisition and pre-processing;Cross-level data interaction mechanism is established, and Modbus unified protocol is used to realize the transmission of linkage instruction between equipment, it is connected to EMS and power grid dispatching platform upwards, connect user side system downwards, and formulate integrated IEC 104 standard data interaction specification;Knowledge graph construction includes defining equipment, index, scene and other core entities;"perception-interaction-decision" closed-loop collaboration, system data acquisition provides the basis for knowledge graph, data interaction guarantees scene information integrity, and the rule generated by knowledge graph reversely guides operation control instruction.The present application can effectively improve the operation efficiency of light storage fills system, reduce degree of electricity cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy management, in particular to a photovoltaic- energy storage- charging pile scene data stream processing and collaborative decision-making method. BACKGROUND

[0002] The current photovoltaic- energy storage- charging pile system data collaborative decision-making development faces three technical bottlenecks. First, protocol heterogeneity causes data fragmentation. Photovoltaic arrays, energy storage systems and charging pile facilities often use different communication protocols (such as Modbus, CAN bus, IEC104, etc.), forming a vertically closed data system. This protocol difference leads to the need for complex protocol conversion gateways for cross-system interaction, not only significantly increasing system integration costs, but also causing cascading delays in command transmission, severely restricting the real-time collaboration capabilities of multiple devices.

[0003] Second, decision lag limits dynamic response efficiency. Traditional SCADA (Supervisory Control And Data Acquisition) systems mainly undertake data acquisition and monitoring functions, and control commands are highly dependent on manual decision-making. In the face of dynamic scenarios such as power grid price fluctuations and load mutations, the system lacks the ability to autonomously generate optimized strategies, making it difficult to capture fleeting energy arbitrage windows and respond to device overload risks in a timely manner, affecting the economy and safety of system operation.

[0004] Finally, static rules weaken the collaborative potential. Existing systems generally use fixed threshold control strategies, lacking the ability to adaptively adjust to real-time operating conditions. There is a lack of effective linkage mechanism between photovoltaic output, energy storage charging and discharging, and charging demand, resulting in low efficiency of energy space-time matching; fault diagnosis relies on rigid threshold determination, which is prone to false triggering for transient fluctuations, increasing unnecessary downtime risks. These fragmented control modes ultimately limit the overall operational efficiency and economy of the system, restricting the high proportion of clean energy consumption. SUMMARY

[0005] To solve the above technical problems, the present application provides a photovoltaic- energy storage- charging pile scene data stream processing and collaborative decision-making method, which effectively improves the operational efficiency of the photovoltaic- energy storage- charging pile system and reduces the cost of electricity.

[0006] To achieve the above purpose, the present application adopts the following technical solution:

[0007] A photovoltaic- energy storage- charging pile scene data stream processing and collaborative decision-making method, comprising the following steps:

[0008] S1, uniformly deploy sensors and intelligent instruments at the device layer of the photovoltaic- energy storage- charging pile, collect the original operating data of the photovoltaic inverter, energy storage system and intelligent charging pile in parallel, and complete filtering, abnormality rejection and timestamp injection to form a standardized data stream; photovoltaic- energy storage- charging pile represents photovoltaic, energy storage and charging pile;

[0009] S2, with the standardized data stream as input, two link encapsulations are completed in the protocol conversion gateway in parallel: the upward link is to map the data to IEC104 frames and insert high priority queues, real-time interface energy management system, power grid dispatching platform, and the downward link is to encapsulate the same data into JSON format and open to user side system through RESTful API, realizing cross-level information synchronization;

[0010] S3, the standardized data stream is continuously written into the graph database, the "device-index-scene" triplets are dynamically established, and the product weight is assigned according to the sampling frequency and business importance, thereby generating an online updateable knowledge graph;

[0011] S4, the real-time data stream is injected into the knowledge graph to trigger rule reasoning, output control instructions, and the running results are fed back to the knowledge graph and the rule confidence is updated after the execution, thereby completing the closed-loop control of perception-interaction-decision.

[0012] Advantages:

[0013] 1. The application improves the system decision response capability: through the deep integration of knowledge graph and collection and monitoring, the transformation from passive monitoring to active decision is realized. Based on real-time data, control instructions are dynamically generated, which significantly shortens the system response time. The cross-level protocol conversion architecture (such as Modbus to IEC104) effectively reduces the instruction transmission delay and improves the photovoltaic consumption efficiency.

[0014] 2. The application optimizes the energy economic management: combined with the peak-valley arbitrage rule and the multi-objective optimization model, the cost is reduced under the premise of ensuring energy demand. The dynamic weight adjustment mechanism can flexibly adapt to different operation scenarios, prolonging the service life of the equipment. The introduction of equipment health state evaluation provides data support for preventive maintenance.

[0015] 3. The application enhances the reliability of equipment operation: the multi-dimensional verification mechanism (such as 3σ principle and equipment health model) is integrated to improve the accuracy of abnormal diagnosis. The historical fault rule library constructed based on historical data supports early warning of potential risks and reduces unplanned downtime events.

[0016] 4. The application promotes multi-level collaborative interaction: the standardized API interface opens the data access channel of the user side, improving the information interaction efficiency. It is upward compatible with the power grid dispatching demand and downward linked with the terminal energy consumption equipment, thereby constructing a source-grid-load-storage collaborative framework and providing a technical basis for system expansion. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the photovoltaic storage charging scene data stream processing and collaborative decision method provided in the embodiment of the application.

[0018] Figure 2 A communication protocol and standard diagram provided for an embodiment of the present application shows a schematic diagram of the interaction mechanism of the Modbus protocol and the IEC 104 standard in the optical storage and charging. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0020] As Figure 1 shown, the optical storage and charging scene data flow processing and collaborative decision-making method of the present application adopts a "perception-interaction-decision" closed-loop architecture, including the following steps:

[0021] S1, multi-source data acquisition and layered processing, including constructing an acquisition framework, acquiring data in real time through sensors and intelligent instruments, forming a standardized data set after preprocessing, and finally performing feature extraction to provide data support for the construction of a knowledge graph.

[0022] S2, cross-level data interaction mechanism establishment, including using the Modbus unified protocol to realize the transmission of linkage instructions between devices, formulating integrated IEC104 standard data interaction specifications, and connecting to the EMS (energy management system) / power grid dispatching platform upward and connecting to the user side system downward.

[0023] S3, knowledge graph construction, including defining core entities such as devices, indicators, and scenes.

[0024] S4, "perception-interaction-decision" closed-loop collaboration, including SCADA data supporting knowledge graph construction, interaction mechanism guaranteeing information integrity, and graph rules guiding SCADA instructions in reverse.

[0025] Specifically, in S1, the preprocessing includes noise filtering and outlier detection. Data cleaning techniques are used to remove damaged or irrelevant data points that may affect subsequent analysis, including filtering out erroneous readings and handling missing values. Noise suppression algorithms are applied to eliminate high-frequency noise that may occur during data acquisition. Moving average method or more advanced signal processing techniques can be used to smooth the data and improve its accuracy. High-frequency noise in the collected data is filtered out after filtering, and the filtering process is as follows:

[0026] ;

[0027] wherein, is the filtered voltage data, is the impulse response of the filter, t is the time variable, is the integral variable used to iterate over all possible time points to calculate the integral.

[0028] Outliers in voltage data are detected and removed by the mean and standard deviation method:

[0029] ;

[0030] ;

[0031] where N is the total number of data points, i represents the i-th data point, is the standard deviation of voltage. is the mean of voltage data, is the standard deviation of voltage data.

[0032] If the mean of the voltage data of the data point satisfies , it is determined to be an outlier.

[0033] After cleaning and denoising, the data is standardized to ensure consistency across data from different sources and types, which is crucial for efficient data parsing and integration. The unified format is as follows:

[0034] ;

[0035] where Pecord represents the data in a unified format, is a function of voltage over time, is a function of current over time, is a function of temperature over time.

[0036] Each data record is attached with a timestamp and relevant metadata, including source identifier and security zone classification, to maintain data traceability and facilitate secure data processing.

[0037] Feature extraction is a fundamental step in the data parsing process, aiming to derive meaningful information from the original high-frequency data collected by the optical storage and charging system. The first task of feature extraction involves calculating various statistical quantities that provide a general description of the trends and dispersion in the data set. Commonly used statistical quantities include mean (which gives the average value of data points) and standard deviation (which measures the variation or dispersion of data), whose calculation formulas are as follows:

[0038] ;

[0039] ;

[0040] where M() represents the mean calculation, represents the i-th data point, N is the total number of data points, is the mean, is the standard deviation (as a normalized variance), SD() represents the sample standard deviation calculation.

[0041] In addition to these basic statistics, the system also calculates more advanced statistical features, such as skewness and kurtosis, where skewness measures the asymmetry of the data distribution, and kurtosis indicates the sharpness of the peak of the data distribution. The formula is as follows:

[0042] ;

[0043] ;

[0044] where S is the sample skewness, n is the sample size, K is the sample kurtosis value, and a is the sample index value, represents the a-th sample.

[0045] Preferably, before filtering, each sensor is synchronized with the edge node hardware clock, so that all sampling points in the subsequent standardized data stream have a consistent time reference, avoiding cross-device timing misalignment.

[0046] Specifically, the S2 includes:

[0047] S2-1 constructs a device layer instruction coordination system (Modbus protocol domain):

[0048] In the bottom layer of the light storage and charging system, the application constructs a device coordination network based on the Modbus protocol. Photovoltaic inverters, energy storage systems (BMS), intelligent charging piles and other terminal devices realize interconnection through a unified Modbus RTU / TCP interface. This protocol unification design completely solves the "data island" problem caused by different device protocol heterogeneity in traditional systems.

[0049] S2-2 builds a power grid level data channel (IEC 104 protocol domain):

[0050] In order to realize seamless connection with the power grid dispatching platform, the application deploys an intelligent protocol conversion gateway at the edge layer of the system, continuously performing three key conversions: first, in the data preprocessing stage, the original Modbus data first enters the filtering channel, and then performs 3σ abnormal filtering. Then enter the protocol conversion stage, establish the register address mapping relationship. Finally, after the above data conversion, the data format conversion adopts linear normalization, and the calculation formula of the first normalization is as follows:

[0051] ;

[0052] where, is the normalized value, For actual value, For range.

[0053] A one-to-one mapping table of Modbus register address and IEC104 information object address is established in the protocol conversion gateway, the mapping table is loaded once at the start of the protocol conversion gateway, and the upward conversion does not need to be searched twice, so that the encapsulation time is shortened.

[0054] S2-3 constructs a user side interface:

[0055] For a user system, the application creates an OPC UA service channel, the OPC UA service channel is an object-oriented data modeling framework conforming to the IEC 62541 standard, constructs a device digital twin through an address space, and provides a safe and reliable real-time data pipeline. The subscription push mode of the OPC UA service channel is an event-driven real-time data distribution mechanism, a persistent subscription connection is established on the server side, data node changes (such as voltage and SOC (state of charge)) are monitored with millisecond-level precision, and only when the data breaks through the preset dead zone threshold, the client is actively pushed and updated. The subscription push mode of the OPC UA service channel breaks through the passive limitation of traditional polling, guarantees the integrity of transient data through dynamic queue management, combines AWS-256 encryption transmission and sequence number confirmation mechanism, and realizes ≤100ms level response of key indicators (such as battery temperature and power grid frequency) of the light storage and charging system.

[0056] Specifically, the S3 includes:

[0057] S3-1 entity definition and relationship modeling:

[0058] The association relationship between entities can be represented as a triple structure: (entity, relationship, entity); and the relationship weight is defined according to the data interaction frequency (such as sampling rate) and business importance, wherein the relationship weight The calculation formula is as follows:

[0059] ;

[0060] Wherein, is the data interaction frequency (sampling rate), is the business importance coefficient, is the relationship weight.

[0061] For the relationship of entity time, the Granger causality test is used to analyze the lag correlation between variables, and the calculation formula of Granger causality F is as follows:

[0062] ;

[0063] Wherein, is the residual sum of squares of the non-variable model, is the result of the model with variables; p represents the lag order, i.e., the number of lagged terms included in the model; n represents the sample size, i.e., the number of effective observations used to estimate the model.

[0064] S3-2 knowledge extraction and weight optimization:

[0065] Through dynamic attribute extraction and subsequent correlation analysis, noise is eliminated and lag correlation is determined, and finally the optimized triple (entity A, relationship, entity B) and its relationship weight are converted into Neo4j graph database nodes and relationships, wherein the node attributes include entity type, real-time numerical value (such as voltage), and timestamp; the relationship attributes include relationship weight and lag order.

[0066] S3-3 knowledge reasoning and quality assurance:

[0067] Define the domain rules and embed the graph, and the rule format is as follows:

[0068] If (grid price > 1.0 yuan / kWh) and (SOC > 80%), start discharging;

[0069] Set the confidence threshold θ = 85%, start the quality evaluation mechanism, eliminate low-confidence relationships, and set the conditions as follows: or ; wherein, is the relationship weight, is the confidence threshold, is the frequency or frequency of the relationship, is the critical frequency threshold. SOC represents the state of charge.

[0070] Specifically, the S4 comprises:

[0071] The closed-loop cooperation of "perception- interaction-decision" is mainly realized through a three-layer cooperation architecture, including a perception layer module, an interaction layer module and a decision layer module connected in turn.

[0072] The data mapping unit of the perception layer module specifically performs the following process: mapping the collected photovoltaic current real-time data to the photovoltaic array entity of the knowledge graph; mapping the battery temperature data to the temperature attribute node of the energy storage battery entity; when the temperature attribute exceeds the threshold value, triggering the "reduce charging power" associated rule update.

[0073] The forward driving unit of the interaction layer module specifically comprises: a grid price peak monitoring sub-module for real-time detection of grid price signals; a peak-valley arbitrage rule activation sub-module for activating the charging and discharging plan generation rule in the knowledge graph when the grid price peak event occurs. The reverse driving unit of the interaction layer module specifically comprises:

[0074] A historical fault rule library is configured to store device fault response rules; and an instruction issuing and executing module is configured to control the data acquisition and monitoring system to adjust inverter power parameters when the knowledge graph generates an instruction based on the rule of "inverter over-temperature → reduced power operation".

[0075] The closed-loop execution unit of the decision layer module comprises a strategy execution verification module configured to collect actual charging and discharging efficiency data after the data acquisition and monitoring system executes the "photovoltaic excess power → energy storage charging" strategy; and a confidence level updating module configured to quantize the effectiveness of the strategy as a weight increment, and increase the rule confidence level by a preset step size when the strategy is effective.

[0076] Embodiments

[0077] The simulation system of this embodiment comprises a photovoltaic system with a peak power of 50kW; an energy storage system with a capacity of 100kWh and an initial SOC of 50%; charging piles with a maximum power of 50kW*3; a protocol conversion with a master-backup link switching time of less than or equal to 50ms; and a knowledge graph with a confidence level updating step size of ±0.05.

[0078] The photovoltaic energy storage and charging scene data stream processing and collaborative decision-making method provided in the embodiments of the present application comprises the following steps:

[0079] S1, for real-time applications, smooth the data if necessary to identify important trends and remove noise. Weighted moving average (WMA) is an effective technique to achieve this purpose, and its calculation method is to assign different weights to data points , usually giving higher weights to recent observations:

[0080] ;

[0081] where W represents the weighted moving average value, represents the weight assigned to the i-th data point .

[0082] Perform frequency and time domain calculations:

[0083] Sliding window aggregation is a technique that moves a fixed-size window over a data stream and calculates statistics within that window to capture trends and patterns over time. This method is very suitable for real-time monitoring and trend analysis, as it can provide a continuous view of data over time.

[0084] For a sliding window of size q, the moving average is calculated as follows:

[0085] ;

[0086] where, ​is the moving average of time t, representing the data points within the window, q is the size of the sliding window.

[0087] Frequency domain analysis is used to analyze the periodicity and frequency components of time series data. This method converts data from the time domain to the frequency domain, revealing potential frequency characteristics that are not easily discernible in the time domain. The Fast Fourier Transform (FFT) converts time domain data into frequency components, whose formula is:

[0088] ;

[0089] where, is the frequency domain representation of the signal, represents the time domain representation of the signal, is the frequency, is the imaginary unit. M represents the total number of time points.

[0090] S2, as Figure 2 shown, adopts the Modbus unified protocol to realize the transmission of inter-device linkage instructions, interfaces with EMS and power grid dispatching platform upwards, and connects user-side systems downwards, formulating integrated IEC 104 standard data interaction specifications:

[0091] S2-1 constructs a device layer instruction coordination system (Modbus protocol domain) (i.e. Figure 2 the device layer) including photovoltaic inverters, energy storage systems, and intelligent charging piles, which performs the following operations in the specific operation process:

[0092] Real-time data acquisition: the data acquisition and monitoring system polls the device registers at a cycle of 500ms to obtain 23 types of core parameters such as photovoltaic power generation rate (function code 16), energy storage SOC state (function code 06), and charging pile load;

[0093] Device linkage control: when it is detected that the photovoltaic power generation exceeds the charging demand (such as during the noon on a sunny day), the system automatically sends Modbus function code 16 instructions to convert the excess power into the energy storage system; when the grid price enters the peak period, the function code 06 instruction is activated to replace the grid power supply by activating the energy storage discharging mode;

[0094] Abnormal coordination processing: when the charging pile is overloaded (current exceeds the set threshold), the system links photovoltaic load reduction and energy storage compensation within 200ms to form a closed-loop protection.

[0095] S2-2 constructs a power grid level data channel (IEC 104 protocol domain):

[0096] To realize seamless docking with the grid dispatching layer (including EMS / gird dispatching platform), the application deploys a protocol conversion gateway at the edge layer of the system, continuously performing three-layer key conversion. First, in the data preprocessing stage, the original Modbus data is first filtered by a sliding window, and then the following is performed Abnormal filtering. Then, in the protocol conversion stage, the address mapping of the register is established, and the mapping relationship is as follows:

[0097] Modbus 40001 (photovoltaic voltage) -> IEC 104 IOA 1;

[0098] Modbus 40105 (energy storage SOC) -> IEC 104 IOA 105;

[0099] Third, after the above data conversion, the data format conversion needs to be converted by linear normalization (i.e. Figure 2 Data normalization).

[0100] The converted IEC 104 ASDU frame is uploaded through a dedicated channel. To ensure real-time performance, the communication protocol IEC104 protocol includes the following features: setting a high priority queue, using frame compression and implementing dual-link redundancy backup.

[0101] S2-3 builds a user-side interface, which utilizes the characteristics of the OPC UA service channel: event-driven subscription push, combined with AWS-256 encrypted transmission and sequence number confirmation mechanism, to realize ≤100ms level response of key indicators (such as battery temperature, grid frequency) of the photovoltaic storage charging system.

[0102] S3, knowledge graph construction, including defining core entities such as devices, indicators, and scenarios:

[0103] S3-1 entity definition and relationship modeling:

[0104] Core entity definition: define device entities (photovoltaic inverter, energy storage system BMS, charging pile), indicator entities (voltage, current, temperature, SOC), and scenario entities (peak valley arbitrage, fault diagnosis);

[0105] Relationship modeling: adopt a triple structure to represent entity association: triple = (entity A, relationship, entity B);

[0106] Wherein the relationship types include physical connection (such as "photovoltaic inverter-connection-charging pile"), data dependency (such as "SOC-affect-discharge instruction"), and scenario trigger (such as "grid price>1.0 yuan / kWh-trigger-peak valley arbitrage");

[0107] The calculation method of weight assignment is: ;

[0108] Wherein, the data interaction frequency (Sampling rate) and the service importance coefficient Calculate the relationship weight .

[0109] S3-2 knowledge extraction and weight optimization;

[0110] S3-3 knowledge reasoning and quality assurance.

[0111] S4, "perception-interaction-decision" closed-loop cooperation: collecting system data provides the basis for the knowledge graph, data interaction ensures the integrity of the scene information, and the rules generated by the knowledge graph guide the operation control instructions in reverse.

[0112] The perception layer module maps the real-time data stream collected by the data acquisition and monitoring system, such as photovoltaic current (±0.5A accuracy) and battery temperature (±0.1℃ accuracy), to the entity attribute nodes of the knowledge graph (such as "photovoltaic array-current attribute" and "energy storage battery-temperature attribute"), and the attribute changes trigger the associated rule updates (for example, when the temperature is greater than 45℃, the "reduce charging power" rule is activated). The device health model in the knowledge graph is called to perform secondary verification on the original alarm (for example, when the temperature fluctuation is less than 30 seconds, the false alarm shutdown instruction is suppressed). The data flow is from the acquisition system sensor to the knowledge graph entity attribute, and triggers the rule chain dynamic update.

[0113] The interaction layer module activates the business rules in the knowledge graph (for example, when the "peak-valley arbitrage mode" is activated, the "22:00-6:00 the next day charging" plan is generated) using the preset events monitored by the data acquisition and monitoring system (such as grid price > 0.8 yuan / kWh). The knowledge graph calls the historical fault rule library and issues device parameter adjustment instructions to the data acquisition and monitoring system through the Modbus-TCP protocol. The instruction closed-loop process is: first, event monitoring is performed, then rule activation and rule reasoning are performed, and finally, bidirectional synchronous control is executed.

[0114] The decision layer module constructs the target entity association in the knowledge graph: , generates a function , and dynamically adjusts the weight. , , The weights of the cost entity, energy efficiency entity, and life entity are respectively 0.5, 0.3, and 0.2. The data acquisition and monitoring system executes the strategy (such as "when the photovoltaic power is abundant, charge the energy storage"), then collects the actual performance data (charging efficiency ≥ 92% is effective), and feeds back to the knowledge graph to update the rule confidence (effective strategy weight + 0.05).

[0115] For example Figure 2A communication protocol and standard diagram provided in an embodiment of the application is shown, which shows a schematic diagram of the interaction mechanism of the Modbus protocol and the IEC 104 standard in the optical storage and charging, Figure 2 The interaction mechanism of the Modbus protocol and the IEC 104 standard in the optical storage and charging is shown in FIG. 1, which is divided into three layers, including a device layer (a Modbus protocol domain), a protocol conversion gateway (an edge layer), and a power grid scheduling layer (an IEC 104 protocol domain).

[0116] The device layer uses a unified Modbus RTU / TCP protocol to realize seamless interconnection among three core devices, i.e., a photovoltaic inverter, an energy storage system (BMS), and an intelligent charging pile. Main functions and applications include:

[0117] (1) Centralized collection of key parameters: Based on the Modbus protocol, key operation data of the device are periodically polled and collected, such as a photovoltaic power generation rate (function code 16) of a photovoltaic system and an energy storage SOC state (function code 06) of an energy storage system (BMS).

[0118] (2) Issuance of energy storage instructions: A control instruction is sent to the energy storage system (BMS) by using the function code 16, so as to realize dynamic adjustment of the charge-discharge behavior thereof.

[0119] (3) State monitoring of the charging pile: The register information of the intelligent charging pile is efficiently read by using the function code 06, so as to accurately obtain the real-time working load state thereof.

[0120] The Modbus RTU / TCP protocol is characterized in that a strict 500 ms polling cycle is set, so as to ensure that basic operation data (such as power, state, and load) have high real-time performance, thereby providing timely and reliable bottom-layer data support for an upper-layer control system.

[0121] The core task of the protocol conversion gateway (the edge layer) is to solve the multi-protocol compatibility problem of device communication in an industrial Internet of Things, so as to realize seamless data interaction between heterogeneous devices and an upper-layer system (such as a cloud platform or a monitoring center). The protocol conversion gateway includes the following core links:

[0122] (1) Data preprocessing, including: sliding window filtering, through data aggregation (such as calculation of a 5-minute average) within a time window, to eliminate jitter noise of a sensor signal and improve data stability; abnormality filtering, based on a threshold or a statistical model, to intercept invalid data (such as an instantaneous current peak) in real time, so as to prevent error data from interfering with upper-layer decision-making.

[0123] (2) Address mapping: dynamically converting a device layer register address (such as 40001 of the Modbus) into a standardized communication protocol (such as IOA 1 of the IEC 104), establishing a unified addressing system, and eliminating the difference between different device address systems.

[0124] (3) Data normalization: Convert raw device values into a standard communication format (e.g., JSON structured data), ensuring consistency in format, units, and dimensions across different protocols, and supporting direct parsing across systems.

[0125] In addition, the edge layer dynamically adapts to unlabeled protocols through protocol stack adaptation technology (accuracy ≥ 92%) and uses lightweight computing engines to complete data processing locally, significantly reducing transmission delay and cloud load, providing core support for industrial real-time control and energy efficiency optimization.

[0126] The core goal of the power grid dispatch layer (communication protocol: IEC 104 protocol) is to meet the stringent requirements of power grid control for high reliability and real-time performance, ensuring stable execution of critical instructions in complex network environments, including:

[0127] (1) High-priority queue mechanism: Through the control domain identifier of the IEC 104 protocol (such as I / S / U format frames), important instructions (such as emergency shutdown, fault isolation) are assigned the highest transmission priority. Dispatch instructions can be handled in queue, ensuring that critical operations are completed within 500ms, avoiding response delays due to network congestion.

[0128] (2) Frame compression optimization technology: Using APDU (Application Protocol Data Unit) dynamic compression algorithm, by simplifying the redundant fields (such as fixed address, repeated timestamp) in ASDU (Application Service Data Unit), the message length is compressed to 30%-50% of the original size, significantly reducing bandwidth occupancy and improving transmission efficiency.

[0129] (3) Dual-link redundancy backup: Deploying primary and backup dual TCP / IP communication channels, through real-time heartbeat detection (interval ≤ 20ms) to monitor link status. When the main link fails, the standby link automatically switches within 50ms, combined with sequence number retransmission mechanism to ensure zero data loss, meeting the availability requirements of power dispatch.

[0130] The implementation goal of the user-side interface (OPC UA service) is to provide a safe and efficient data interaction channel, supporting real-time monitoring and active intervention on the user side. Key features include:

[0131] (1) Event-driven subscription and push mechanism: Abandoning the traditional polling mode, adopting a subscription-published model. When device data changes (such as voltage limit, switch position change), the OPC UA service channel actively pushes event messages to the user end, reducing invalid data transmission and network load, while supporting ≤100ms end-to-end response.

[0132] (2) End-to-end security reinforcement: integrate AWS-256 encryption algorithm for data transmission encryption, combined with digital certificate two-way authentication (X.509 standard) and message signature mechanism to prevent man-in-the-middle attacks and data tampering. The security policy library supports dynamic updates and can resist quantum computing brute force cracking threats.

[0133] (3) Real-time intervention channel: through the OPC UA service channel call wake-up (Method Call) service, users can directly issue control instructions (such as load adjustment, start and stop the unit). The instruction transmission delay is ≤100ms, and supports transaction rollback mechanism to ensure that misoperations can be revoked immediately, improving system operation safety.

[0134] Through the protocol conversion gateway, IEC 104 and OPC UA are seamlessly coordinated to build a complete "control-user" bidirectional closed loop: IEC 104 focuses on the deterministic transmission of high-voltage instructions (such as emergency shutdown, power regulation) at the dispatching end, ensuring the absolute reliability of the power grid side operation; at the same time, the OPC UA service channel provides a flexible interaction channel for the user side, supporting millisecond-level data active push driven by events. When the main dispatching link fails, the local redundant data pool immediately pushes the key operating parameters to the user terminal through the OPC UA service channel, triggering the off-grid decision mechanism - users can quickly execute pre-set emergency strategies (such as switching to island mode, starting emergency support of energy storage) based on real-time load status and local data such as energy storage SOC, maintaining the continuity of core power supply while completely avoiding the risk of cascading collapse of the power grid, achieving a leap in system-level disaster tolerance resilience.

[0135] The light storage and charging system breaks through the coordination of economy and reliability to achieve a double leap in energy utilization efficiency and safety protection: in terms of economy, the system fully utilizes the peak-valley electricity price difference (typical price difference 0.6 yuan / kWh or more) to realize daily arbitrage income of 43.56 yuan in the mode of charging 85.3kWh with valley electricity and discharging 72.6kWh with peak electricity, while reducing the conversion loss of energy storage by 8% through the direct supply of charging piles with surplus photovoltaic power during lunchtime, and improving the self-consumption rate of photovoltaic power to more than 92% through intelligent dispatching algorithms, significantly reducing the cost of electricity (0.36 yuan / kWh vs. traditional 0.48 yuan / kWh); in terms of reliability, the system adopts a millisecond-level response mechanism, when the charging pile load suddenly increases to 120% overload, the whole process of abnormal detection, instruction generation, photovoltaic load reduction, and energy storage compensation is completed within 182ms, and the response to the power grid dispatching instruction is realized within 200ms through the protocol conversion gateway (Modbus→IEC 104).

[0136] Those skilled in the art will readily understand that the above description is only a preferred embodiment of the present application and is not intended to limit the present application, and any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for light storage charging scene data stream processing and collaborative decision making, characterized in that, Comprise the following steps: S1, deploy sensors and smart meters in the device layer of the light storage and charging system, collect raw operation data of photovoltaic inverters, energy storage systems, and smart charging piles in parallel, complete filtering, abnormality rejection, and timestamp injection, and form standardized data streams; Light storage and charging refers to photovoltaic, energy storage, and charging piles; S2, take the standardized data streams as input, complete two link encapsulations in the protocol conversion gateway in parallel: the upward link is to map the data into IEC 104 frames and insert them into a high-priority queue, real-time interface with energy management systems and power grid dispatching platforms, and the downward link is to encapsulate the same data into JSON format and open it to user-side systems through RESTful API, realizing cross-level information synchronization; S3, continuously write the standardized data streams into a graph database, dynamically establish "device-index-scenario" triples, and generate an online-updatable knowledge graph by multiplying the sampling frequency and business importance; before establishing the "device-index-scenario" triples, calculate the mean of each index in a sliding window manner, discard the sampling value if the difference between the current sampling value and the mean exceeds the preset threshold, and ensure that the data entering the light storage and charging graph is valid data; if the relationship weight in the product weight of a certain relationship is lower than the confidence threshold, automatically delete the relationship; S4, inject real-time data streams into the knowledge graph to trigger rule reasoning, output control instructions, reverse them to the corresponding devices through the channel of the original protocol conversion gateway, execute them, feed back the running results to the knowledge graph and update the rule confidence, and complete the closed-loop control of perception, interaction, and decision-making; Before rule reasoning, call the device health model to do secondary verification on the original alarm, and only when the knowledge graph reasoning and the device health model are triggered at the same time can the final control instruction be formed.

2. The method of claim 1, wherein, In S1, synchronize the hardware clocks of each sensor and edge node before filtering, so that all sampling points in the subsequent standardized data streams have a consistent time reference, avoiding cross-device time sequence misplacement.

3. The method of claim 1, wherein, In S2, establish a one-to-one mapping table of Modbus register addresses and IEC 104 information object addresses in the protocol conversion gateway, load the mapping table once when the protocol conversion gateway starts, ensure that upward conversion does not need secondary table lookup, and shorten the encapsulation time.

4. The method of claim 1, wherein, In S2, linearly normalize the data before JSON format encapsulation, and convert running parameters of different dimensions to the same numerical interval.

5. The method of claim 1, wherein, In S4, the control instruction delivery path and the data upload path share the same physical channel, so that the instruction stream and the data stream complete convergence in the protocol conversion gateway, reducing additional wiring.

6. The method of claim 1, wherein, In S2, assign the highest transmission priority through the control domain identifier of the IEC 104 protocol.

7. The method of claim 1, wherein, In S4, the knowledge graph calls the historical fault rule library and issues control instructions through the Modbus-TCP protocol.

Images