Distributed power load management system based on power grid automation equipment

Abstract

The application relates to the field of automatic control of power systems, and particularly discloses a distributed power source load management and control system based on power grid automation equipment. The application mainly solves the power grid operation risk problem caused by the low data collection accuracy, poor prediction model adaptability and insufficient source-load control cooperation of the existing distributed power source management and control system. The application realizes microsecond-level data collection by deploying a multi-modal high-precision sensing network and a hardware-level time synchronization mechanism; adopts an LSTM and XGBoost hybrid prediction model based on an attention mechanism to improve the load and new energy output prediction accuracy; constructs a source-grid-load-storage multi-objective optimization decision model to generate a cooperative scheduling instruction; and relies on double-channel redundant communication and secure encrypted transmission to realize accurate linkage control of the load side and the power source side. The application effectively improves the power grid's consumption capacity for distributed energy and ensures the stability and reliability of system operation.

Description

Technical Field

[0001] This invention relates to the field of power system automation control, and specifically discloses a distributed power load management system based on power grid automation equipment. Background Technology

[0002] With the large-scale grid connection of renewable energy, the penetration rate of distributed generation in the power system continues to rise, posing a severe challenge to the traditional power grid operation and management model. Existing technologies have significant shortcomings in addressing the randomness and volatility of distributed generation. For example, the grid connection control method proposed in patent CN112421633A mainly adopts a fixed-parameter adjustment strategy, which is difficult to adaptively track the drastic power fluctuations of distributed generation such as photovoltaic and wind power, easily leading to frequent voltage exceedances at the grid connection point and system frequency instability.

[0003] At the condition monitoring level, existing systems mostly rely on centralized data acquisition architectures such as SCADA, which have long data update cycles, heavy communication burdens, and blind spots in sensing the operating status of distributed power sources at the end of the distribution network. For example, the patent system with publication number CN113346515A has a monitoring data delay of minutes, which cannot provide accurate data support for real-time control. Especially during fault transients, this delay may lead to delayed control commands or even erroneous actions.

[0004] In terms of load and power forecasting, existing technologies are poorly adapted to abnormal operating conditions. Patent CN112865150A discloses a load forecasting method based on traditional time series data. This method has large prediction errors in scenarios such as sudden weather changes and sharp load fluctuations, which seriously affects the accuracy of scheduling plans. This forces the control system to frequently activate standby capacity for compensation, increasing operating costs.

[0005] Existing solutions lack the ability to coordinate and optimize across multiple dimensions, including power generation, grid, load, and storage. Although the patent with publication number CN113991590A mentions coordination among multiple links, the control commands between subsystems often suffer from problems such as asynchronous timing and conflicting optimization objectives, failing to form an efficient and unified optimization scheduling system and hindering the grid's ability to maximize the absorption of distributed energy resources.

[0006] To address the aforementioned technical challenges, this invention proposes an intelligent management and control system for distributed power generation loads based on power grid automation equipment. By innovatively integrating technologies such as grid-connected control, multi-source monitoring, intelligent prediction, and collaborative optimization, a complete closed-loop management and control chain is constructed to enhance the power grid's capacity to accommodate high proportions of distributed power generation and improve its operational management level. Summary of the Invention

[0007] This invention utilizes a data acquisition module to collect real-time operating parameters of distributed power sources and loads using sensor technology, and constructs a power grid topology model to monitor key nodes. A data analysis and decision-making module applies intelligent algorithms to analyze the collected data, constructs a supply and demand forecasting model, identifies power grid risk points, and generates a comprehensive scheduling decision including load adjustment and power regulation strategies. A load scheduling and allocation module analyzes the scheduling decisions, dynamically executes the switching, power limiting, or time-sharing operations of controllable load nodes, and coordinates with the distributed power source coordination and control module to issue power setting commands to target power sources, jointly maintaining source-load balance and power grid stability. A communication and remote control module aggregates and transmits scheduling commands and monitoring data through an encrypted protocol, providing a remote operation interface. A security and fault tolerance module employs redundant deployment and real-time fault detection mechanisms to ensure the safety of critical equipment and data streams, triggering fault switching to maintain continuous system operation. These modules form a closed loop of "data acquisition - intelligent decision-making - dynamic regulation - collaborative control," achieving optimized configuration of distributed power sources and loads and safe and stable management of the power grid.

[0008] The objective of this invention can be achieved through the following technical solutions: A distributed power source load management system based on power grid automation equipment includes a data acquisition module, a data analysis and decision-making module, a load scheduling and allocation module, a distributed power source coordination and control module, a communication and remote control module, and a security and fault tolerance module, wherein: The data acquisition module collects real-time power system parameters from power grid equipment. The data analysis and decision-making module uses real-time power system parameters to conduct intelligent algorithm analysis on the power grid operation status and load demand, and forms a dispatching decision scheme. The load dispatching and allocation module, in conjunction with the dispatching decision scheme, performs dynamic allocation operations on the power grid load to achieve optimal allocation of power resources and load balance; The distributed power generation coordination and control module, relying on the optimal allocation of power resources and load balancing, carries out control strategy adjustment of the output power of distributed power sources to build a stable and coordinated state of the power grid. The communication and remote control module, combined with a stable and coordinated state, performs remote transmission of control commands and monitoring data via communication protocols, generating efficient monitoring and operation of system equipment; The security and fault tolerance module relies on efficient monitoring of system equipment to implement redundant design and fault detection mechanisms to protect critical equipment and data flow, ensuring system reliability and data security.

[0009] Preferably, when the data acquisition module collects real-time data on the operation of distributed power sources and loads within the power grid using the sensing technology of power grid automation equipment, it includes: Establish a power grid topology diagram based on the IEC 61970 CIM standard, defining distributed photovoltaic as PV nodes, energy storage units as balancing nodes, and adjustable loads as time-varying PQ nodes; Deploy 0.2S-class voltage transformers and wideband Rogowski coil current sensors at key nodes, configure smart meters according to the DL / T 645-2007 protocol, and add silicon photovoltaic irradiance meters and PT1000 temperature sensors to form a multimodal sensing network. A GPS / BeiDou dual-mode timing module is used to output a 1PPS pulse to the sample-and-hold circuit to ensure that the sampling time deviation across nodes is controlled within 1 microsecond. Collect real-time operating parameters and environmental parameters of key nodes; The collected parameters are preprocessed and a three-level cleaning process is performed: physical rule filtering, statistical model filtering, and state association verification.

[0010] Preferably, when the data acquisition module transmits the acquired power system parameters and equipment status data to the central processing unit via wireless communication technology, it includes: The data is encapsulated into a compact JSON format standardized data packet, with the size of a single data packet compressed to less than 256 bytes; Transmitted via LoRaWAN wireless communication link, Class C terminal mode enabled, adaptive frequency hopping mechanism configured, using SF9 spreading factor and 125kHz bandwidth combination; The edge gateway uploads data through the 5G private network, enables URLLC network slicing, configures a dedicated QoS flow identifier QFI82, and ensures end-to-end transmission latency ≤20ms. Verify the integrity and timeliness of transmitted data by enabling CRC-32 checksum and timing continuity check.

[0011] Preferably, when the data analysis and decision-making module analyzes the collected power grid data using intelligent algorithms, it includes: Receives real-time data streams transmitted by the central processing unit; A load demand forecasting model based on LSTM and a photovoltaic output forecasting model based on XGBoost were constructed, with a three-layer 128-unit hidden layer and 1000 decision trees configured. Predictive models are used to analyze the power grid's supply and demand balance in future periods, and a real-time correction mechanism is introduced to handle scenarios of sudden heavy rain and rapid cloud movement. The forward-backward substitution method is used to calculate branch power flow, identify potential grid congestion areas, voltage over-limit risks, and optimization scheduling opportunities, and consider the dynamic adjustment threshold of equipment aging coefficient.

[0012] Preferably, when the data analysis and decision-making module generates a scheduling decision based on the analysis results, it includes: Based on the results of supply and demand balance analysis, risk identification results, and preset optimization objectives; The NSGA-II multi-objective optimization algorithm was used, with a population size of 200 and 500 iterations, to generate a Pareto optimal solution set. Generate load adjustment strategy schemes and distributed power generation output adjustment strategy schemes, and formulate mitigation measures for grid congestion areas and voltage over-limit risks; The output contains comprehensive scheduling decision instructions that include adjustment strategies and mitigation measures, and is encapsulated using Protocol Buffers serialization.

[0013] Preferably, when the load dispatching and allocation module dynamically adjusts the power grid load according to the dispatching scheme, it includes: Analyze the load adjustment strategy schemes in the integrated dispatch decision instructions; Calculate the adjustment amount and priority of each controllable load node, and calculate the optimal sequence of compensation costs by referring to the electricity market contract database; It executes commands for switching, power limiting, or adjusting the operating time of controllable load nodes, and transmits control commands using the IEC 60870-5-104 protocol; Monitor the power grid operation status feedback after load adjustment and activate the feeder current change verification mechanism.

[0014] Preferably, when the load dispatching and allocation module ensures optimal allocation of power resources and load balance, it includes: Evaluate the degree to which the load adjustment performance meets the preset optimization goals; Establish a dynamic priority correction model to adjust node credit ratings and adjustable capacity coefficients based on actual response performance; By sending energy storage scheduling requests to the distributed power coordination module through the cross-module communication interface, a load-energy storage joint regulation system is formed. Model predictive control algorithm is used to adjust the tap changer of on-load tap changing transformer, and voltage sensitivity analysis is introduced to switch capacitor banks.

[0015] Preferably, when the distributed power supply coordination and control module adjusts the power output power through a control strategy, it includes: Analysis of distributed power generation output adjustment strategies in integrated dispatch decision instructions; Calculate the power setpoint or adjustment command for each target distributed power source and adopt an adaptive droop control algorithm; Power setpoints or adjustment commands are sent to the corresponding distributed power source local controller, using the IEC 61850-8-1 MMS protocol; Verify the actual output of the distributed power source and its ability to follow commands, and enable a real-time power tracking error detection mechanism.

[0016] Preferably, when the communication and remote control module transmits control commands and monitoring data via a communication protocol and network platform, it includes: It aggregates load dispatching commands, power control commands, and monitoring data; Block encryption is performed using the national standard SM4 algorithm, digital signature is performed using the SM2 algorithm, and the key is automatically rotated every 24 hours. The service VLAN ID is set to 2023 and the priority is marked as 6, and the transmission is carried out through dual channels of EPON fiber optic network and 5G wireless private network. It provides a 3D visualization remote monitoring interface based on HTML5 and WebGL technologies, supporting the simultaneous display of 2000 dynamic elements.

[0017] Preferably, when the security and fault-tolerant module protects critical system equipment and data flow through redundant design and fault detection mechanisms, it includes: Deploy a dual-active architecture central processing unit cluster and achieve failover within 50 milliseconds using Keepalived; Construct a dual transmission path consisting of an optical fiber communication ring network and a wireless mesh network, and configure OSPF and HWMP routing protocols. Run a multi-level heartbeat detection process and implement full-stack monitoring, including IPMI interface data collection and HTTP health check endpoints; Trigger automated emergency response to ensure business migration and zero data loss within 300 milliseconds; Implement end-to-end security protection, adopt AES-256 encryption algorithm and RBAC access control model, and deploy a deep packet inspection firewall.

[0018] In the above-mentioned distributed power load management system based on power grid automation equipment: The data acquisition module constructs a power grid topology model based on sensing technology, deploys a sensor network at key nodes, and collects real-time data on voltage, current, power, and equipment status parameters of distributed power sources and loads. It also performs preprocessing on the raw data, including cleaning, format conversion, and timestamp alignment. The preprocessed data is then encapsulated into standardized data packets using wireless communication technology and transmitted to the central processing unit for verification of integrity and timeliness, forming the data foundation for system decision-making.

[0019] The data analysis and decision-making module receives real-time data streams and constructs load demand forecasting models and distributed power generation output forecasting models through intelligent algorithms. It analyzes the power grid supply and demand balance in future periods, identifying congested areas, voltage exceedance risks, and optimal dispatch points. Based on the analysis results, this module combines preset optimization objectives to generate load adjustment strategies, power generation adjustment strategies, and risk mitigation measures, ultimately outputting a comprehensive dispatch decision containing multi-dimensional control commands.

[0020] The load dispatching and allocation module analyzes the comprehensive dispatching decisions, calculates the adjustment amount and priority of each controllable load node, and performs dynamic adjustments such as switching, power limiting, or operating periods. Simultaneously, the distributed power supply coordination and control module analyzes the power regulation strategy, calculates the power setpoint of the target power source, issues commands to the local controller, and verifies output responsiveness. The two modules work collaboratively: the load dispatching module evaluates the adjustment effect and dynamically optimizes the allocation strategy, while the power supply control module adjusts the output in real time, jointly maintaining voltage / frequency stability at key nodes and achieving source-load coordination and optimized allocation of power resources.

[0021] The communication and remote control module aggregates load dispatching commands, power control commands, and monitoring data, encapsulates them using an encrypted protocol, and transmits them via wired / wireless networks. It also provides a remote monitoring interface to support manual operation.

[0022] The security and fault tolerance module implements triple protection: deploying a dual-machine hot standby cluster of critical servers to establish redundant backup of communication paths; running a real-time fault detection process to monitor hardware / software / communication status; and triggering an automatic fault isolation and switching mechanism to ensure the continuous operation of core functions, while ensuring the security of encrypted data transmission, storage, and access control, forming a closed-loop reliable management and control system.

[0023] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention dynamically analyzes decision instructions and executes controllable load node switching power limitation operations through a load dispatching module. Simultaneously, the power control module sends power setpoints to distributed power sources and verifies their tracking performance, forming real-time linkage control between load and power source. Combined with the analysis and decision module's proactive identification and mitigation of grid congestion and voltage over-limit risks, it effectively suppresses supply and demand fluctuations and significantly improves grid frequency / voltage stability and anti-disturbance capability.

[0024] 2. The security module of this invention deploys a dual-machine hot standby cluster of critical servers and redundant communication links to ensure high hardware availability; it runs a real-time fault detection process to monitor anomalies and trigger automatic isolation and switching to ensure the continuous operation of core functions under single point of failure; at the same time, the communication module uses an end-to-end encrypted protocol to transmit scheduling instructions and monitoring data, implements data storage and access control, and builds a defense-in-depth architecture from hardware redundancy to data security.

[0025] 3. The analysis and decision-making module of this invention integrates multiple objectives such as economy, stability and environmental protection to generate a joint optimization scheme for load adjustment and power output, maximizing the consumption of clean energy; the load scheduling module continuously evaluates the execution effect and adaptively adjusts the allocation strategy, responding to environmental parameters and real-time load changes, improving the system's resource utilization efficiency and dynamic elasticity, and realizing an upgrade from passive response to proactive optimization in the management and control mode. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the structure of a distributed power load management system based on power grid automation equipment according to the present invention. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. Example

[0028] like Figure 1 As shown, this embodiment provides a distributed power load management system based on power grid automation equipment, which includes the following modules: The data acquisition module collects real-time power system parameters from power grid equipment. The data analysis and decision-making module uses real-time power system parameters to conduct intelligent algorithm analysis on the power grid operation status and load demand, and forms scheduling decisions and optimization schemes. The load dispatching and allocation module, in conjunction with dispatching decisions and optimization schemes, performs dynamic allocation operations on the power grid load to achieve optimal power resource allocation and load balance; The distributed power generation coordination and control module, relying on the optimal allocation of power resources and load balancing, carries out control strategy adjustment of the output power of distributed power sources to build a stable and coordinated state of the power grid. The communication and remote control module, in conjunction with the stable coordinated state of the power grid and the stable coordinated state of the power supply, performs remote transmission of control commands and monitoring data through communication protocols, generating efficient monitoring and operation of system equipment; The security and fault tolerance module relies on efficient monitoring of system equipment to implement redundant design and fault detection mechanisms to protect critical equipment and data flow, ensuring system reliability and data security.

[0029] In a preferred embodiment of the present invention, the specific implementation of the data acquisition module includes: This module first constructs a digital topology model of the power grid based on the IEC 61970 CIM standard, accurately identifying all distributed power generation nodes, such as photovoltaic inverters, wind power converters, energy storage converter PCS, load nodes (interruptible loads for industrial and commercial use, movable loads for residential use), and network connections. The key innovation lies in the deployment of a multimodal, high-precision sensor network: at the distributed photovoltaic grid connection point, we deploy 0.2S-level voltage transformers and wideband Rogowski coil current sensors, with a ratio error better than ±0.2% and an angle error less than ±10′, capable of accurately capturing harmonics and transient processes; on both the AC and DC sides of the energy storage unit, Hall effect bidirectional current sensors are installed with an accuracy of ±0.5%, and a PT1000 platinum resistance temperature sensor is integrated, with a range of -40℃ to +85℃; for load nodes, smart meters supporting the DL / T 645-2007 protocol are configured, possessing Class 1 accuracy and second-level data freeze functionality. In addition, each photovoltaic array is equipped with a secondary standard silicon photovoltaic cell irradiance meter with a range of 0-1500W / m² and an error of ±3%, as well as a backplane temperature sensor, forming a comprehensive perception of "power supply-grid-load-environment".

[0030] To address the industry challenge of inconsistent time scales in multi-source data, this system implements hardware-level precise time synchronization. Each sensor node incorporates a GPS / BeiDou dual-mode timing module, outputting a 1PPS (pulse-per-second) signal with an accuracy of ±30ns to a high-speed sample-and-hold circuit. At the rising edge of the pulse, the instantaneous voltage and current values ​​of all nodes are synchronously locked, and A / D conversion is initiated, ensuring that the sampling time deviation across a wide area of ​​nodes is strictly controlled within 1 microsecond. Voltage / current sensors acquire waveform data at a sampling rate of 4kHz, and the built-in DSP chip performs a Fast Fourier Transform to calculate the fundamental RMS value, harmonic distortion rate, and active / reactive power with an integration period of 80ms in real time. Smart meters read power scalar values ​​and circuit breaker open / close position status at 200ms intervals; meteorological sensors output cosine-corrected light intensity and temperature data at a frequency of 1Hz.

[0031] The collected raw data streams undergo a rigorous three-level cleaning process to ensure data quality. The first level is physical rule filtering: data points with voltage values ​​exceeding ±20% of the rated value are automatically discarded; invalid reactive power data where the current phase angle deviates from the voltage phase angle by more than ±90° is marked; and alarms are triggered for abnormal data where photovoltaic power output exceeds 0.5% of the rated capacity at night. The second level is statistical model filtering: a 60-second sliding window is used to calculate the mean μ and standard deviation σ of each data stream. If the current sampled value satisfies |x-μ|>3σ, it is determined to be transient interference and discarded; when 10 consecutive sampled points deviate from the historical average of ±15%, the expert system review mechanism is activated. The third level is state correlation verification: when a circuit breaker is in the open state but power >0.1kW is detected, it is marked as logically conflicting data; if the DC current direction and AC power sign do not match during the charging process of the energy storage unit, a device fault alarm is generated. This process ensures data accuracy and logical consistency.

[0032] After cleaning, the data enters the standardization phase. The system strictly adheres to the IEC 61850 standard, mapping heterogeneous data to a unified semantic model: RMS voltage / current values ​​are mapped to the PhV / PhA attributes of the MMXU logical node; active / reactive power is mapped to the MMXU.Pwr object; and circuit breaker status is mapped to the Pos.stVal attribute of the XCBR logical node. Timestamps are uniformly converted to UTC time in ISO 8601 format, and all scalar values ​​are appended with unit identifiers and data quality flags. Environmental parameters are encapsulated as MMET logical device instances. The hierarchical structure of logical devices and logical nodes is precisely defined through the system configuration language, ultimately generating a CID configuration file conforming to the IEC 61850-7-4 specification, achieving unified semantic expression and interoperability of multi-source heterogeneous data.

[0033] To address micro-scale deviations in data transmission, the system initiates high-precision timestamp synchronization processing. The deviation between the UTC timestamp of each data frame and the local clock of the acquisition terminal is detected. When the deviation exceeds 10ms, cubic spline interpolation is used to recalculate the precise time points of current, voltage, and other time-series data. Subsequently, all data is aligned using a fixed 1-second time window, and all data within the window is uniformly marked as the window's start time. Missing data windows are filled by copying data from the previous valid window. If consecutive missing data exceeds 5 seconds, an advanced data integrity alarm is triggered. The final standardized monitoring dataset is transmitted via a LoRaWAN wireless network. The data is encapsulated in compact JSON format packets, using Class C terminal mode and an adaptive frequency hopping mechanism: primary frequency 868MHz, backup 923.6MHz, configured with an SF9 spreading factor and 125kHz bandwidth. Each data packet is appended with AES-128 encryption and a 4-byte message integrity code to ensure that the packet loss rate is stably controlled below 5% even in complex electromagnetic environments.

[0034] In a preferred embodiment of the present invention, the specific implementation of the data analysis and decision-making module includes: The data analysis and decision-making module receives real-time normalized data streams from the data acquisition module via an Apache Kafka message queue. The data stream updates at 1-second granularity, with dimensions of [time step × number of nodes × parameter channels]. The parameter channels include 12 feature quantities such as voltage, current, active power, reactive power, illumination, and temperature. The module first performs streaming preprocessing: extracting valid data packets within the sliding time window, using Lagrange interpolation to spatially compensate for missing node data; and normalizing all electrical parameters to a per-unit system, with the reference voltage being the rated voltage and the reference power being the total system capacity, eliminating the influence of dimensions and providing high-quality, standardized input tensors for subsequent prediction and analysis algorithms.

[0035] The core innovation of this module is the adoption of a hybrid intelligent forecasting architecture. For short-term load forecasting, a bidirectional long short-term memory network (Bi-LSTM with Attention) model is constructed. The model input is designed with three channels: historical load curves, active power sequences of 96 points from the past 24 hours, weather forecast data, temperature, humidity, and wind speed forecasts for the next 2 hours, calendar features, weekday type, holiday markers, and intraday time segmentation. The LSTM core network is configured with three layers of 128 hidden layers, a time step of 24, and a dropout rate of 0.2 to prevent overfitting. For photovoltaic output forecasting, an XGBoost gradient boosting tree model is used, whose input incorporates physical parameters such as real-time irradiance, module temperature, cloud cover, and moving vectors. Key derived variables such as the product of GHI and temperature, and the cloud cover moving average decay coefficient are constructed in feature engineering. The model is trained on three years of historical data, using Huber Loss as the loss function, AdamW as the optimizer, and an initial learning rate of 0.001 with cosine annealing for decay. The MAPE error of the final load forecasting model on the test set remained stable within 4.5%, and the photovoltaic forecasting error remained within 7.2%.

[0036] After the forecast is completed, the system initiates multi-period rolling optimization calculations. Triggered every 5 minutes, it continuously forecasts the power grid supply and demand status for 16 time periods, ranging from 15 minutes to 2 hours. The supply-demand balance deviation index is dynamically calculated using the formula: Deviation Index = (Total Load Forecast - Total Photovoltaic Output Forecast - Baseload Power Capacity) / System Spinning Reserve Capacity. A real-time correction mechanism is incorporated: when radar detects a sudden downpour, a humidity correction factor of 1.15 is added to the load forecast; when the all-sky imager identifies cloud movement speeds exceeding 8 m / s, a dynamic attenuation algorithm is activated for the photovoltaic output forecast. Simultaneously, based on a forward-backward power flow algorithm, all branches of the distribution network are scanned to identify congested lines where the current exceeds 85% of the long-term allowable current carrying capacity for 3 consecutive minutes, and over-limit nodes where the voltage deviates from the rated value by ±10% for more than 2 minutes. Risk identification also considers equipment aging factors; for lines that have been in operation for over 10 years, their current carrying capacity threshold is automatically lowered to 80%.

[0037] Based on the prediction and risk identification results, the system constructs a mathematical optimization model with the objectives of minimizing network loss, maximizing renewable energy absorption, and achieving the most stable voltage. The optimization algorithm employs an improved fast non-dominated sorting genetic algorithm with an elitist strategy. The population size is set to 200, the number of iterations to 500, the crossover probability to 0.85, and the mutation probability to 0.015. Each individual is encoded as a 48-bit binary code: the first 16 bits control the start-up and shutdown of load groups, the middle 16 bits define the energy storage charging and discharging power, and the last 16 bits encode the capacitor bank switching status. After the algorithm runs, it outputs a Pareto optimal solution set. Finally, a weighted method is used, with a network loss weight of 40%, a absorption weight of 35%, and a voltage stability weight of 25%, to select the optimal scheduling scheme from the solution set. The scheme can be specified as follows: commercial air conditioning groups are delayed in batches (priority 1 delayed by 15 minutes, priority 2 delayed by 30 minutes), industrial loads are reduced according to contract tiers, and the energy storage system discharges at a 0.8C rate for 2 hours at midday.

[0038] The generated optimization strategy is encapsulated into structured, comprehensive scheduling decision instructions. These instructions are serialized using Protocol Buffers format and define a complex message structure containing fields such as LoadControl, EssSchedule, and CongestionMitigation. For example, the LoadControl message body includes the load group ID, operation type, DELAY_START or SHEDDING, parameters, delay minutes or reduction percentage, and effective time window. A digital signature and security checksum based on the SHA-256 algorithm are appended to the end of the instruction to ensure its integrity and immutability during transmission. Finally, this instruction set is pushed to the load scheduling and power control module in real time via a highly available gRPC streaming interface.

[0039] In a preferred embodiment of the present invention, the specific implementation of the load scheduling and allocation module includes: Upon receiving the integrated dispatch decision command, the load dispatch module first decrypts and verifies the digital signature to ensure the command's legitimacy. Then, the parsing engine extracts all elements from the `load_control` array within the command. Each element contains the load group ID, operation type code, parameter set, and a UTC effective time window accurate to milliseconds. For industrial loads requiring reduction, the system links to the electricity market contract database to query the interruptible load compensation price for each user. The priority calculation engine generates a queue based on compensation cost: users with the lowest compensation price and high historical performance credit are marked as priority 1 and given priority for reduction; users with high compensation prices or low credit have lower priority. This algorithm ensures the lowest possible compensation cost across the entire network while meeting the total reduction requirements.

[0040] Based on the calculated priority sequence and specific control parameters, the system issues control commands to heterogeneous load terminals through various communication protocols. For industrial high-voltage load nodes in the distribution network, tripping or power limit commands are directly sent to their smart circuit breakers via the IEC 60870-5-104 protocol. These commands include the device ID, target value, and execution timestamp. The circuit breaker executes at the specified time and provides a status signal. For large commercial buildings, the system interacts with the building management system via the BACnet / IP protocol, sending commands to increase the air conditioning temperature setpoint by 1°C or reduce the lighting circuit power to 65%. For distributed loads on the residential side, time control commands are broadcast to smart meters via the DL / T 645-2007 protocol. The meter's built-in clock chip synchronizes with the BeiDou time signal, automatically performing delayed start or early shutdown operations at preset times. All command issuance employs a closed-loop mechanism of "request-confirmation-execution-feedback" to prevent misoperation.

[0041] After the command is issued, the system initiates a high-frequency monitoring and verification process. During industrial load reduction periods, the current values ​​uploaded by the RTUs on the corresponding feeders are scanned to calculate the deviation between the actual reduction and the expected target. If the deviation persists for more than ±15% for one minute, it is determined to be an execution anomaly, and the standby load switching process is immediately triggered. For commercial buildings, total power changes are monitored at one-minute intervals. If power exceedances are detected, secondary adjustment commands are automatically issued, such as shutting down some fresh air units or advertising lighting. The response of residential loads is compared and analyzed using the daily freeze curves of smart meters to identify load groups that have not deviated according to the commands, downgrade their credit ratings, and reduce their priority in subsequent scheduling.

[0042] The system does not statically execute preset plans but possesses online adaptive optimization capabilities. The module calculates the degree of agreement between the actual load control effect and the expected target across the entire network every hour. If the agreement falls below 85%, a dynamic adjustment strategy is initiated: backup users are activated in the industrial load priority queue for compensation; the offset coefficient of residential load is corrected based on the historical response rate of the area. Simultaneously, root cause analysis is performed on abnormal events occurring during execution, and the fault knowledge base is updated. This knowledge is used to optimize subsequent scheduling strategies, such as avoiding frequently failing nodes or reserving longer command transmission time for areas with poor communication quality.

[0043] This module does not operate in isolation, but rather collaborates deeply with the distributed power supply coordination and control module through a cross-module communication interface. During peak load periods, this module sends energy storage dispatch request messages to the power supply module, requesting the target compensation power value, duration, and efficiency requirements. When a distribution network frequency fluctuation exceeds 0.05Hz, it requests the energy storage system to switch to virtual synchronous generator mode to participate in primary frequency regulation. Similarly, when performing load transfer to alleviate line congestion, it communicates with the power supply module in real time to ensure that the output changes of the distributed power supply match the direction of load transfer, avoiding new power flow anomalies and jointly building a stable control system for source-load coordination.

[0044] In a preferred embodiment of the present invention, the specific implementation of the distributed power coordination and control module includes: After receiving instructions from the analysis and decision-making module, the distributed power coordination and control module first parses them. and Fields. The parsed instructions may include: active power setpoint, reactive power setpoint, power change rate limit, and control mode switching commands for frequency regulation. Subsequently, the control algorithm calculates the precise instructions to be sent to the local controller of each distributed power source based on the real-time grid status. For energy storage systems, an adaptive droop control algorithm is used to dynamically calculate the active and reactive power to be generated based on the voltage and frequency deviation at the point of common coupling, using the following formula: ,in This is an adjustable droop factor. The calculation process strictly follows the equipment constraints to ensure that the command value does not exceed the inverter's overload capacity or violate the SOC safety boundary of the energy storage battery.

[0045] After calculating the power setpoint, the system sends it to the local controllers of each distributed power source via a highly reliable communication protocol. For large-scale photovoltaic power plants and energy storage power plants, the IEC 61850-8-1 MMS protocol is used for communication, mapping control commands to the corresponding logical nodes. For small and medium-sized distributed power sources, the DNP3 or IEC 60870-5-104 protocol is used. The command message contains key information such as device ID, control mode, target value, execution time, and timeout. All commands are simulated and verified before being sent to ensure that they will not cause equipment overload or system oscillation, and are transmitted through a secure encrypted tunnel to prevent commands from being stolen or tampered with.

[0046] After the command is issued, the system enters a rigorous closed-loop verification phase. After the local controller executes the command, its operating status and output value are transmitted back in real time via the data acquisition module. The control module compares the actual output value with the issued command value to calculate the power tracking error. If the error ΔP exceeds 5% of the rated output for more than 2 seconds, it is considered a poor tracking performance, and the system immediately triggers an alarm and initiates a compensation mechanism. The compensation mechanism includes: reissuing commands, activating the backup power unit, or notifying the load dispatch module for coordinated adjustment. This closed-loop verification mechanism ensures the effective execution of control commands, avoiding a situation where commands are issued without regard to the outcome.

[0047] For qualified energy storage systems, the core function of this module is to switch its control mode from conventional PQ control to virtual synchronous generator control, enabling it to simulate the inertial response and primary frequency regulation function of a synchronous generator. When the system detects a frequency deviation exceeding 0.05Hz, it automatically issues a mode switching command. In VSG mode, the energy storage inverter simulates the rotor motion equation and excitation regulation through algorithms, providing virtual inertia (J) and damping (D). Its active power-frequency droop characteristic curve is as follows: This not only effectively suppresses frequency fluctuations but also provides crucial instantaneous reactive power support for the power grid, enhancing its voltage stability and disturbance rejection capabilities—something traditional control strategies cannot achieve.

[0048] This module maintains real-time communication with the load scheduling module, enabling source-load linkage. When the load module performs large-scale load input, this module receives an advance warning message that "load is about to be input" and calculates the required power output increment in advance, preparing for increased power generation to avoid system power shortages caused by sudden load input. Conversely, when the load drops sharply, the power module also receives an early warning, thus reducing output in advance to prevent power surplus. This event-triggered collaborative strategy surpasses traditional periodic regulation, greatly improving the system's response speed and stability.

[0049] In a preferred embodiment of the present invention, the specific implementation of the communication and remote control module includes: The communication module, acting as a data hub, first establishes multiple data aggregation channels. It receives in real-time JSON-formatted control commands from the load scheduling module, energy storage charging and discharging commands from the power control module, and high-precision synchronization phasor data (50 frames per second) from the PMU. These data use various protocols. The module's built-in protocol conversion engine, based on the Apache Camel framework, uniformly converts this heterogeneous data into the system's internal standard Google Protobuf format. During the conversion process, key data is precisely mapped and time alignment is performed. A sliding time window algorithm with a window size of 200ms is used to eliminate timescale discrepancies between sources, ensuring spatiotemporal consistency in subsequent processing.

[0050] All control commands and sensitive monitoring data are encrypted with high strength before transmission. The system uses the national standard SM4 algorithm for block encryption, with a key length of 128 bits, operating in GCM mode to provide confidentiality and integrity protection. The initialization vector is dynamically generated by a true random number generator in the hardware security module. The encrypted data packets are then signed using a digital signature based on the SM2 elliptic curve algorithm, with the signing process using the private key certificate from the scheduling center. Non-sensitive data is protected using CRC-32 checksums. The encryption module is deployed on a dedicated cryptographic machine, and the key strictly adheres to the "one machine, one key" principle, automatically rotating every 24 hours to completely eliminate the risk of key leakage.

[0051] To ensure absolute communication reliability, the system constructs a dual-channel redundant communication network using both fiber optic and wireless channels. The primary channel utilizes an EPON fiber optic network, with OLT equipment deployed at the provincial dispatch center and ONU terminals installed in substations. Data packets are encapsulated in EPON frame format, with the service VLAN ID set to 2023 and a priority marker of 6, ensuring priority forwarding of control commands. The backup channel uses a 5G wireless private network connected to the power slicing network and configured with a dedicated QoS stream to ensure end-to-end transmission latency ≤15ms. The dual channels operate in hot standby mode, using the BFD protocol for millisecond-level link detection; after a 300ms interruption of the primary channel, the service flow automatically and seamlessly switches to the backup channel.

[0052] On the dispatch center side, a web-based SCADA system is deployed as a remote monitoring interface. The front-end is developed using HTML5, WebGL, and the Three.js engine to construct a 3D visualization model of the power grid facilities. System status is dynamically rendered: overloaded lines are displayed as a red gradient band, with transparency increasing linearly from 85% to 100% load rate; voltage exceeding limits are marked with a purple pulse icon, with flashing frequency positively correlated with the degree of deviation. Dispatchers can import Excel templates and configure control parameters for up to 500 loads or power sources at once using the "Batch Command Issuance" function on the interface. All manual operations are recorded by the log module, supporting multi-dimensional auditing by operator, time, equipment type, etc. Critical operations require secondary authentication and automatically trigger screen recording for retention.

[0053] The module possesses robust anomaly detection and self-healing capabilities. It monitors the EPON optical power value in real time, triggering an alert when it falls below -27dBm; it continuously monitors the RSRP of the 5G signal, determining signal degradation if it remains below -110dBm for 3 consecutive seconds. If encrypted data packets fail verification at the receiving end, the receiver initiates a selective retransmission request based on the packet sequence number, with the retransmission interval using an exponential backoff algorithm, increasing from 100ms to 800ms. The human-machine interface implements operation error prevention logic; for example, if an attempt is made to issue a command exceeding 120% of the device's rated capacity, the system will automatically intercept it and display a secondary confirmation dialog box. When the same device experiences three consecutive control failures or communication interruptions exceeding 5 minutes, the system automatically generates a diagnostic report and pushes it to the maintenance personnel's mobile terminal.

[0054] In a preferred embodiment of the present invention, the specific implementation of the security and fault tolerance module includes: At the physical layer, the system employs a dual-active data center architecture to deploy the central processing unit cluster. The primary and backup data centers are more than 50 kilometers apart, providing disaster recovery capabilities. Each center is configured with identical server nodes, equipped with Intel Xeon Gold processors and 256GB of memory, interconnected via dual 10 Gigabit fiber optic network cards. The operating system is a customized CentOS, with deeply optimized kernel parameters such as tcp_keepalive_time=300s and tcp_retries2=5. A virtual IP address is configured as the cluster's external service address via the Keepalived service, with primary and backup nodes having priorities of 100 and 90 respectively. Combined with the VRRP protocol, fault detection and failover within 50 milliseconds are achieved. Shared storage uses a dual-controller fiber optic SAN, configured with a multi-path I / O strategy to ensure automatic failover in the event of a single FC link failure.

[0055] At the network layer, the system constructs dual redundant paths: an optical fiber communication ring network and a wireless mesh network. The optical fiber ring network is built based on the OSPF dynamic routing protocol, with dual-fiber bidirectional self-healing rings deployed between nodes, using G.652.D single-mode fiber, and the ring network self-healing time is <50ms. The wireless mesh network adopts the 802.11ax standard, configured with the HWMP routing protocol, and path selection comprehensively considers RSSI, channel utilization, and hop count. The system operates an intelligent traffic allocation strategy: real-time control commands are preferentially transmitted through the highly reliable optical fiber ring network; massive monitoring data is diverted to the high-bandwidth wireless mesh network. When the packet loss rate of any link exceeds 0.1%, the service flow automatically switches to the other channel.

[0056] The system operates a multi-layered, fine-grained real-time fault detection process. At the hardware level, the CPU temperature, fan speed, and power status of server nodes are collected every 5 seconds via the IPMI interface; an alert is triggered if the temperature exceeds 85°C for 30 seconds. At the operating system level, a custom daemon is deployed to monitor the memory usage and response latency of critical services such as kubelet and etcd. At the application level, all microservices must provide an HTTP health check endpoint and return a JSON response containing version number and load status within 300ms. Communication links between nodes are checked every 200ms by exchanging 128-byte heartbeat packets; a communication interruption is determined by the loss of three consecutive frames.

[0057] Upon detection of any fault, the system triggers a pre-defined automated emergency response process. The Keepalived service immediately revokes the VIP binding of the faulty node and guides traffic switching via broadcast ARP update messages. The Kubernetes control plane marks the faulty node as NotReady and begins migrating its running Pods to healthy nodes according to priority: the real-time data analytics service is migrated first, and the migration process ensures TCP session persistence through preStop hooks and ready probes. The storage system automatically isolates the faulty disk and triggers RAID reconstruction. Simultaneously, the notification module alerts users via SMS, audio-visual alerts, and other methods, with alert information including the fault location code, impact scope, and a link to the handling manual. The entire failover process ensures RPO (Recovery Point Objective) = 0 and RTO (Recovery Time Objective) < 300ms.

[0058] The system implements an end-to-end defense-in-depth security strategy. Data transmission uses AES-256 encryption, operates in GCM mode, and the initialization vector is updated with each packet. Key management is based on the PKCS#11 standard, handled by the hardware security module, and keys are automatically rotated every 24 hours. Access control uses the RBAC model, defining four roles: system administrator, scheduler, operations engineer, and auditor, with over 200 detailed operation permissions. Sensitive operations require dual review, and operation logs are recorded in a blockchain-based immutable evidence storage system. A next-generation firewall is deployed at the network boundary, enabling deep packet inspection to intercept abnormal protocol formats or high-frequency access behaviors in real time, with a processing capacity of up to 100,000 concurrent policy checks per second. All security events are correlated and analyzed; when brute-force attempts are detected, GeoIP filtering is automatically activated to block IP access originating from high-risk areas.

[0059] To maintain high reliability, the system regularly conducts full-link disaster recovery drills. Monthly simulations include scenarios such as main data center power outages, fiber optic cable breaks, and core server crashes, testing virtual IP switching time, service recovery completeness, and data consistency. Based on the drill results, system parameters are continuously optimized; for example, the heartbeat timeout threshold is adjusted from 3 to 5 to reduce false positives during switching. At the hardware level, preventative maintenance is implemented, with capacitors and other vulnerable components replaced prematurely on servers older than 3 years. Software versions are rolled out using a blue-green deployment strategy. After the new version has been stably running on the backup node for 72 hours, a seamless switchover is completed by updating the VIP pointer, achieving zero-disruption service upgrades.

[0060] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0061] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A distributed power source load management system based on power grid automation equipment, characterized in that, The system includes a data acquisition module, a data analysis and decision-making module, a load scheduling and allocation module, a distributed power supply coordination and control module, a communication and remote control module, and a security and fault tolerance module, among which: The data acquisition module is used to collect real-time power system parameters from power grid equipment. The data analysis and decision-making module is used to conduct intelligent algorithm analysis on the grid operation status and load demand based on real-time power system parameters, and to form a dispatching decision scheme. The load dispatching and allocation module is used to dynamically allocate power grid loads in conjunction with dispatching decision schemes to achieve optimal power resource allocation and load balance. The distributed power source coordination and control module is used to adjust the output power of distributed power sources by relying on the optimal allocation of power resources and load balancing, so as to build a stable and coordinated state of the power grid. The communication and remote control module is used to remotely transmit control commands and monitoring data in conjunction with a stable and coordinated state, thereby generating efficient monitoring and operation of system equipment. The security and fault tolerance module is used to implement redundant design and fault detection mechanisms to protect critical equipment and data streams by relying on efficient monitoring of system equipment.

2. The distributed power supply load management system based on power grid automation equipment according to claim 1, characterized in that, Collect real-time power system parameters from power grid equipment, including: Establish a power grid topology diagram based on the IEC 61970 CIM standard, defining distributed photovoltaic as PV nodes, energy storage units as balancing nodes, and adjustable loads as time-varying PQ nodes; At key nodes, voltage transformers and wideband Rogowski coil current sensors are deployed, smart meters conforming to the DL / T 645-2007 protocol are configured, and silicon photovoltaic irradiance meters and PT1000 temperature sensors are added to form a multimodal sensing network. A GPS / BeiDou dual-mode timing module is used to output a 1PPS pulse to the sample-and-hold circuit to ensure that the sampling time deviation across nodes is controlled within 1 microsecond. Collect real-time operating parameters and environmental parameters of key nodes; The collected parameters are preprocessed and a three-level cleaning process is performed: physical rule filtering, statistical model filtering, and state association verification.

3. The distributed power supply load management system based on power grid automation equipment according to claim 2, characterized in that, When the data acquisition module transmits the acquired power system parameters and equipment status data to the central processing unit via wireless communication technology, it includes: The data is encapsulated into a compact JSON format standardized data packet, with the size of a single data packet compressed to less than 256 bytes; Transmitted via LoRaWAN wireless communication link, Class C terminal mode enabled, adaptive frequency hopping mechanism configured, using SF9 spreading factor and 125kHz bandwidth combination; The edge gateway uploads data through the 5G private network, enables URLLC network slicing, configures a dedicated QoS flow identifier QFI 82, and ensures end-to-end transmission latency ≤20ms; Verify the integrity and timeliness of transmitted data by enabling CRC-32 checksum and timing continuity check.

4. The distributed power load management system based on power grid automation equipment for effective management according to claim 1, characterized in that, When the data analysis and decision-making module analyzes the collected power grid data using intelligent algorithms, it includes: Receives real-time data streams transmitted by the central processing unit; A load demand forecasting model based on LSTM and a photovoltaic output forecasting model based on XGBoost were constructed, with a three-layer 128-unit hidden layer and 1000 decision trees configured. Predictive models are used to analyze the power grid's supply and demand balance in future periods, and a real-time correction mechanism is introduced to handle scenarios of sudden heavy rain and rapid cloud movement. The forward-backward substitution method is used to calculate branch power flow, identify potential grid congestion areas, voltage over-limit risks, and optimization scheduling opportunities, and consider the dynamic adjustment threshold of equipment aging coefficient.

5. The distributed power supply load management system based on power grid automation equipment according to claim 4, characterized in that, When the data analysis and decision-making module generates scheduling decisions based on the analysis results, it includes: Based on the results of supply and demand balance analysis, risk identification results, and preset optimization objectives; The NSGA-II multi-objective optimization algorithm was used, with a population size of 200 and 500 iterations, to generate a Pareto optimal solution set. Generate load adjustment strategy schemes and distributed power generation output adjustment strategy schemes, and formulate mitigation measures for grid congestion areas and voltage over-limit risks; The output contains comprehensive scheduling decision instructions that include adjustment strategies and mitigation measures, and is encapsulated using Protocol Buffers serialization.

6. The distributed power supply load management system based on power grid automation equipment according to claim 1, characterized in that, When the load dispatching and allocation module dynamically adjusts the power grid load according to the dispatching scheme, it includes: Analyze the load adjustment strategy schemes in the integrated dispatch decision instructions; Calculate the adjustment amount and priority of each controllable load node, and calculate the optimal sequence of compensation costs by referring to the electricity market contract database; It executes commands for switching, power limiting, or adjusting the operating time of controllable load nodes, and transmits control commands using the IEC 60870-5-104 protocol; Monitor the power grid operation status feedback after load adjustment and activate the feeder current change verification mechanism.

7. The distributed power supply load management system based on power grid automation equipment according to claim 6, characterized in that, When ensuring optimal allocation of power resources and load balance, the load dispatching and allocation module includes: Evaluate the degree to which the load adjustment performance meets the preset optimization goals; Establish a dynamic priority correction model to adjust node credit ratings and adjustable capacity coefficients based on actual response performance; By sending energy storage scheduling requests to the distributed power coordination module through the cross-module communication interface, a load-energy storage joint regulation system is formed. Model predictive control algorithm is used to adjust the tap changer of on-load tap changing transformer, and voltage sensitivity analysis is introduced to switch capacitor banks.

8. The distributed power supply load management system based on power grid automation equipment according to claim 1, characterized in that, When the distributed power supply coordination and control module adjusts the power output power through control strategies, it includes: Analysis of distributed power generation output adjustment strategies in integrated dispatch decision instructions; Calculate the power setpoint or adjustment command for each target distributed power source and adopt an adaptive droop control algorithm; Power setpoints or adjustment commands are sent to the corresponding distributed power source local controller, using the IEC 61850-8-1MMS protocol; Verify the actual output of the distributed power source and its ability to follow commands, and enable a real-time power tracking error detection mechanism.

9. The distributed power supply load management system based on power grid automation equipment according to claim 1, characterized in that, When the communication and remote control module relies on communication protocols and network platforms to transmit control commands and monitoring data, it includes: It aggregates load dispatching commands, power control commands, and monitoring data; Block encryption is performed using the national standard SM4 algorithm, digital signature is performed using the SM2 algorithm, and the key is automatically rotated every 24 hours. The service VLAN ID is set to 2023 and the priority is marked as 6, and the transmission is carried out through dual channels of EPON fiber optic network and 5G wireless private network. It provides a 3D visualization remote monitoring interface based on HTML5 and WebGL technologies, supporting the simultaneous display of 2000 dynamic elements.

10. The distributed power supply load management system based on power grid automation equipment according to claim 1, characterized in that, When security and fault-tolerant modules protect critical system devices and data flows through redundant design and fault detection mechanisms, they include: Deploy a dual-active architecture central processing unit cluster and achieve failover within 50 milliseconds using Keepalived; Construct a dual transmission path consisting of an optical fiber communication ring network and a wireless mesh network, and configure OSPF and HWMP routing protocols. Run a multi-level heartbeat detection process and implement full-stack monitoring, including IPMI interface data collection and HTTP health check endpoints; Trigger automated emergency response to ensure business migration and zero data loss within 300 milliseconds; Implement end-to-end security protection, adopt AES-256 encryption algorithm and RBAC access control model, and deploy a deep packet inspection firewall.

Images