A rural micro-energy cluster operation control method and system

Abstract

The application discloses a rural micro-energy cluster operation control method and system, and belongs to the technical field of digital twinning and intelligent energy collaborative control. A three-level hardware architecture of top-down cluster coordination unit, micro-energy station autonomous unit and equipment execution terminal is adopted, only boundary aggregated data is transmitted, and edge on-site operation and local storage are relied on. Through architecture networking initialization, rural exclusive constraint parameters are input, and training of a reinforcement learning model and calibration of a digital twinning channel are completed. Day-ahead cluster optimization, intra-day station-level rolling correction and real-time fluctuation suppression are realized relying on three time scales, communication states are monitored, an autonomous power supply mode is automatically switched in case of network interruption, smooth grid connection is realized after communication recovery, and model iteration optimization is completed based on operation data. The application solves the problems of poor adaptability, fixed scheduling strategy and weak communication fault tolerance of traditional micro-energy cluster management and control, effectively improves new energy consumption rate and distribution network operation stability, and adapts to differentiated operation scenes of rural micro-energy clusters.

Description

Technical Field

[0001] This invention belongs to the field of digital twin and smart energy collaborative control technology, specifically a rural micro-energy cluster multi-timescale rolling optimization and communication fault autonomous power supply control technology that integrates reinforcement learning and edge hierarchical architecture. Background Technology

[0002] Village-level micro-energy clusters effectively integrate various energy devices such as household photovoltaics, distributed wind power, biomass energy supply, energy storage, and heat pumps. They also cover diversified electricity and heat loads such as residential life and agricultural production. Through multi-site aggregation and collaboration, they achieve complementary energy resources, dynamic matching of source and load, and friendly interaction between distribution networks. They have become the core carrier for the clean and low-carbon transformation of rural energy and the improvement of rural power supply security.

[0003] Currently, existing micro-energy cluster operation and control technologies largely adopt microgrid control schemes used in urban industrial and commercial parks. These schemes lack adaptability to rural micro-energy systems and customized control logic, failing to meet the differentiated operational needs of rural energy systems. At the control architecture level, existing centralized control models heavily rely on high-speed communication networks and central computing nodes. However, rural areas are characterized by dispersed site layouts, wide geographical coverage, and weak communication infrastructure. Failures in central nodes or communication link interruptions can easily paralyze the entire micro-energy cluster control, resulting in extremely poor operational reliability. While purely distributed control schemes possess independent operating capabilities, they lack upper-level cluster collaborative scheduling mechanisms, hindering cross-site energy sharing, resource aggregation, and global optimization, thus failing to fully unleash the scalable operational benefits of micro-energy clusters. Existing publicly available patents and related technical solutions primarily focus on distributed collaborative control of conventional microgrid clusters, neglecting adaptation designs for rural weak communication scenarios. They also lack autonomous switching and steady-state continuation mechanisms under abnormal communication conditions, resulting in insufficient system robustness.

[0004] At the level of optimizing scheduling strategies, existing micro-energy cluster optimization control methods generally adopt fixed target weight coefficients, and the optimization dimensions are limited to two general indicators: operational economy and carbon emissions. They fail to construct a quantitative optimization system that incorporates constraints specific to rural scenarios, lacking core constraint parameters such as the priority of power supply based on agricultural production load, seasonal cyclical fluctuations in biomass raw material supply, the upper limit of rural distribution network capacity, and power supply resilience in extreme weather. Rural energy systems exhibit significant seasonal and scenario-specific characteristics. Under different operational scenarios such as busy agricultural irrigation, winter heating, and supply security during extreme disasters, the core management and control requirements of the system vary greatly. Fixed-weight optimization strategies cannot achieve dynamic adaptation of objectives, easily leading to the deterioration of key livelihood and agricultural security indicators, and failing to balance the functionality and practicality of rural energy systems.

[0005] Meanwhile, the output of distributed renewable energy in rural areas is highly random and volatile due to meteorological factors. The fluctuation range of characteristic loads such as agricultural irrigation and agricultural product processing can be 3 to 5 times that of daily loads. Traditional day-ahead optimization scheduling methods driven by pure mechanism models have limited accuracy, and real-time intraday control responses are lagging, easily leading to operational problems such as curtailment of solar and wind power, voltage exceeding limits in transformer areas, and power supply-demand gaps. Most existing technologies focus only on the independent and autonomous operation of single micro-energy stations, lacking a multi-level collaborative cluster scheduling architecture and a cross-site electric-thermal multi-energy complementary and mutually supportive mechanism. The dispersed energy storage devices and backup power capacity cannot be clustered and shared, making it difficult to support the large-scale participation of micro-energy clusters in distribution network demand response, resulting in low energy resource utilization and overall operational economics. In addition, rural power distribution networks generally suffer from problems such as large power supply radius, thin line diameter, serious three-phase imbalance, low voltage qualification rate of transformer areas, and high line loss. Existing control schemes mostly take single power balance as the core control target, and do not embed the distribution network adaptation requirements such as transformer area voltage management, three-phase imbalance suppression, and line loss optimization into the cluster collaborative control logic. The contradiction between network and power supply adaptation is prominent, which restricts the safe, efficient and stable operation of rural micro-energy clusters.

[0006] More importantly, the existing micro-energy cluster management and control system lacks the closed-loop iterative optimization capability that combines virtual and real systems. It generally does not introduce digital twin virtual-real mapping and intelligent model iterative optimization mechanisms, and cannot realize real-time replication of operating status, dynamic parameter calibration and strategy iterative upgrade. At the same time, it lacks multi-timescale collaborative scheduling logic, making it difficult to take into account the full-cycle management and control needs of day-ahead global optimization, intraday rolling correction and real-time fluctuation suppression, and thus cannot form a data-driven closed-loop operation and management and control system.

[0007] In summary, existing micro-energy cluster operation and control technologies suffer from numerous shortcomings, including poor adaptability of the management and control architecture, rigid optimization strategies, insufficient anti-interference and fault self-healing capabilities, lack of closed-loop iterative optimization mechanisms, and incomplete coverage of scenario constraints. These limitations prevent them from meeting the core operational characteristics of rural micro-energy clusters, which are characterized by "distributed deployment, weak distribution network support, strong spatiotemporal fluctuations, diverse scenario requirements, and low communication reliability." Therefore, there is an urgent need to develop a refined operation and control technology for micro-energy clusters that is adapted to specific rural scenarios, features three-level hierarchical management and control, intelligent weight adaptation, virtual-real closed-loop iteration, and autonomous power supply assurance during network outages. Summary of the Invention

[0008] To address the technical shortcomings of existing technologies, such as poor scenario adaptability, unreasonable control architecture, rigid optimization strategies, weak communication fault tolerance, lack of closed-loop iterative optimization system, and insufficient distribution network adaptability, this invention provides a rural micro-energy cluster operation and control system, comprising: a field equipment group, a multi-level control hardware architecture, and multiple sets of functional hardware modules; the field equipment group includes distributed power sources, energy storage devices, energy conversion devices, tiered load units, and sensor acquisition terminals; the multi-level control hardware architecture is composed of cluster coordination hardware units, micro-energy station autonomous hardware units, and equipment execution terminals connected sequentially from top to bottom; its key feature is: The cluster coordination hardware unit is deployed at township energy stations and configured with primary and backup redundant edge computing servers. The micro-energy station autonomous hardware unit is deployed one-to-one to each village micro-energy station and has built-in local solid-state storage media and station-level edge controller. The device execution terminal connects to all equipment in the field equipment group for electrical and signal communication. The cluster coordination hardware unit and each micro-energy station autonomous hardware unit only establish a point-to-point communication link for boundary aggregation data. The micro-energy station autonomous hardware unit and the device execution terminal establish a real-time control signal transmission link. The local computing of the two-level edge hardware reduces the amount of channel data transmission. The local solid-state storage media stores model files and historical operating data to support offline management and operation of the entire station after communication is lost. The system's overall hardware link carries six sets of hardware functional modules, including a multi-factor index construction hardware module, a spatiotemporal dynamic weight adaptation hardware module, a digital twin mapping hardware module, an alternating direction multiplier method (ADMM) distributed solution hardware module, a cross-site multi-energy mutual assistance management hardware module, and a weak communication autonomous distribution network adaptation hardware module. These modules are integrated and deployed within the cluster coordination hardware unit and the micro-energy station autonomous hardware unit, and data interaction between the modules is completed through a hardware bus.

[0009] Preferably, the hardware module for constructing multi-factor indicators is equipped with hardware discrimination logic units that integrate five categories of quantitative indicators. Each category of indicator is equipped with a hardware-integrated constraint threshold register and normalization processing circuit unit. The five categories of indicators include economic indicator units, environmental indicator units, reliability indicator units, resilience assurance indicator units, and distribution network adaptation indicator units. The resilience assurance indicator unit hardware is written with logic for extreme load supply guarantee, backup capacity verification, and fault self-healing judgment. The distribution network adaptation indicator unit hardware integrates voltage qualification judgment, transformer area load monitoring, line loss calculation, and three-phase imbalance detection circuits.

[0010] Preferably, the spatiotemporal dynamic weight adaptation hardware module integrates three layers of hardware processing units: the first layer is a scenario baseline weight storage unit, which pre-stores baseline weight parameters corresponding to multiple typical rural working conditions; the second layer is a deep deterministic policy gradient (DDPG) network hardware acceleration computing unit, which accesses the system's global state acquisition signals and outputs weight correction amounts; the third layer is a spatiotemporal temperature correction register, which has built-in village type correction parameters and time period correction parameters; the module is equipped with a conflict resolution hardware processing chip, which integrates a Pareto solution set calculation unit and an optimal solution screening unit.

[0011] Preferably, the digital twin mapping unit establishes bidirectional data links with the field sensing terminal, the cluster coordination unit, and the micro-energy station autonomous unit, respectively. It relies on the multi-level cache built into the device to distinguish the refresh frequency of device data, system status data, and prediction data. It relies on the edge computing device to carry a three-level scheduling channel of day-ahead planning, intraday rolling, and real-time correction to alleviate the problems of large randomness of source load and lag in single-layer scheduling to suppress disturbances. ADMM distributed solver units are configured in township edge computing devices and village-level edge controllers, respectively. The overall optimization task is decomposed into two parallel processing links: global coupled computing and local site computing. Iterative convergence calculation is completed by relying on the dual variable data interaction link between devices, which improves the drawbacks of centralized optimization, such as concentrated computing power and excessive interactive data volume. The cross-station multi-energy mutual assistance control unit is electrically connected to the power distribution acquisition device of the distribution area and the village-level heating network monitoring terminal. Relying on the mutual assistance path selection and revenue distribution metering logic built into the edge controller, it completes the bidirectional allocation of electric and heat energy between stations by using the existing power distribution network and heating network on site, thus solving the shortcomings of isolated energy storage and heat storage resources of each station and the inability to share them in a cluster. The weak communication autonomous distribution network adaptation unit is electrically connected to the communication status detection device. It relies on the edge controller's built-in logic configuration millisecond-level operation mode switching function, and simultaneously links with the field reactive power compensation device and energy storage control equipment to form a linkage control link for voltage regulation, three-phase imbalance management, and line loss optimization. It is used to quickly switch to autonomous operation mode when communication is abnormal, and smoothly switch back to cluster collaborative control after communication is restored.

[0012] Based on the same inventive principle, this invention also provides a method for controlling the operation of rural micro-energy clusters, which employs the aforementioned rural micro-energy cluster operation control system for scheduling and management; including the following steps: S1. Complete the networking and connection of the three-level management and control architecture devices, input the device parameter calibration, initialize the parameters of each functional unit, and agree on the hierarchical data transmission protocol and the autonomous mode switching trigger conditions. S2. Load all quantitative indicator parameters into the multi-factor indicator construction unit, and input the agricultural load limit, biomass supply cycle, and distribution network carrying capacity limit exclusive constraint parameters. S3. Input the scene baseline weights and complete the offline training and calibration of the reinforcement learning model, complete the calibration of the digital twin virtual-real mapping channel and the configuration of the three-level scheduling timing parameters; S4. The cluster coordination unit relies on the township edge equipment to perform daily day-ahead optimization solutions and issue basic operation instructions for unit output, energy storage operation, and cross-station energy mutual assistance. S5. Each micro-energy station autonomous unit relies on the station-level controller to periodically perform intraday rolling optimization, and uses the weight adaptation unit to correct the optimization target weight in real time and update the control instructions of the station equipment. S6. Real-time correction control is performed based on on-site measured data to suppress instantaneous power fluctuations, and the operating data is transmitted back to the digital twin system to form a closed-loop management and control system. S7. Real-time monitoring of communication link status; automatic switching to autonomous operation and priority to ensure power supply to agricultural primary loads after communication interruption; and smooth access to the cluster for collaborative operation after communication is restored. S8. After the daily operation is completed, the operation indicators are statistically analyzed, and the parameters of the reinforcement learning model and digital twin model are iteratively optimized based on the actual test data.

[0013] Preferably, the digital twin collects data in layers according to a preset cycle, and the three-level scheduling processes the data separately according to a predetermined time step to match the source load fluctuation disturbances at different time scales.

[0014] Preferably, in the ADMM distributed solution process, each micro-station calculates local variables in parallel, the cluster coordination unit updates global coupled variables, each station updates dual variables synchronously, and iterative calculations are performed until the convergence criteria are met.

[0015] Preferably, the weight adjustment process involves retrieving the baseline weights for the corresponding working conditions, correcting the weight values ​​through a reinforcement learning model, and superimposing spatiotemporal correction parameters to complete the calibration; when multiple objectives conflict, the Pareto solution set is solved and the optimal adjustment scheme is selected and issued for execution.

[0016] Preferably, a heartbeat message combined with a data verification mechanism is used to identify communication faults, and priority is given to ensuring power supply to primary agricultural loads during the autonomous operation phase; in both cluster collaboration and local autonomy operation modes, voltage control of distribution areas, suppression of three-phase imbalance, and optimization control of low-voltage line loss are completed simultaneously.

[0017] Compared with existing technologies, this invention achieves the following significant beneficial effects through the synergistic cooperation of the aforementioned proprietary architecture, intelligent optimization strategy, and closed-loop management system: First, the robustness of the architecture is significantly improved, perfectly adapting to the weak communication environment in rural areas. This invention adopts a three-level hierarchical distributed management and control architecture, coordinating data only at the cluster level boundary, greatly reducing the overall communication transmission pressure. Compared with the traditional centralized management and control mode, the cluster communication pressure is reduced by more than 78%, completely avoiding the problem of cluster paralysis caused by a single point of failure in the center. At the same time, this invention adds a real-time monitoring and seamless autonomous switching mechanism for communication links. When communication is interrupted, it can quickly switch to the station-level autonomous operation mode within 20ms, prioritizing the stable power supply of agricultural loads. After communication is restored, it can automatically synchronize historical data and smoothly return to cluster collaborative operation. Compared with the traditional fixed collaborative control scheme, the overall power supply reliability of the cluster is improved by more than 15%, effectively adapting to the field characteristics of weak rural communication infrastructure and unstable communication conditions.

[0018] Second, the adaptability to different scenarios is significantly enhanced, completely solving the adaptation shortcomings of traditional fixed-weight optimization. This invention constructs a rural-specific multi-factor evaluation system containing 18 quantitative indicators, comprehensively covering rural-specific constraints such as agricultural production load grading, biomass supply cycle, distribution network capacity limit, and power supply resilience in extreme weather. Combining fuzzy hierarchical analysis (FAHP) and DDPG reinforcement learning algorithm, it achieves continuous spatiotemporal adaptive adjustment of the optimization target weights, adapting to differentiated operating scenarios such as busy farming irrigation, winter heating, and supply security during extreme disasters. This completely overcomes the shortcomings of traditional fixed-weight optimization strategies, such as poor scenario adaptability and easy degradation of core indicators. The overall system performance is improved by 18% to 25% in each core operating scenario.

[0019] Third, the ability to withstand source-load fluctuations is comprehensively strengthened, and the level of new energy consumption is significantly improved. This invention relies on the closed-loop calibration capability of digital twin virtual-real mapping to build a three-level time-scale scheduling system of day-ahead optimization, intraday rolling, and real-time correction. Combined with the ADMM distributed solution algorithm, it completes cluster optimization solutions, which can accurately adapt to the operational characteristics of rural new energy output with strong randomness and large fluctuations in agricultural load, effectively smoothing instantaneous power fluctuations in the system. Field measurements have verified that this invention can increase the regional renewable energy consumption rate by 12%~20% and reduce the curtailment rate of solar and wind power by more than 70%, effectively solving the problems of energy waste and power supply imbalance caused by low scheduling accuracy and control lag in traditional mechanism models.

[0020] Fourth, it fully leverages the synergistic benefits of the cluster to effectively address the shortcomings in rural power distribution network operation. This invention establishes a cross-micro-energy station electricity-heat dual-energy mutual assistance and resource aggregation mechanism, realizing full-area energy storage and backup capacity sharing and complementarity. It supports the large-scale participation of micro-energy clusters in power distribution network demand response, reducing the overall operating cost of the cluster by 10% to 15% compared to the independent operation mode of a single station. Simultaneously, it embeds voltage management, three-phase imbalance suppression, and line loss optimization into the cluster's collaborative control logic, specifically addressing the problems of large rural power distribution network radius, thin wire diameter, and severe three-phase imbalance. This increases the voltage qualification rate of distribution areas to over 99.6%, controls the three-phase imbalance to within 12%, and significantly improves the adaptability and operational stability of rural power distribution networks.

[0021] Fifth, the invention boasts outstanding technological innovation and strong engineering feasibility and practicality. It establishes a complete hierarchical management and control model, a multi-objective optimization mathematical model, intelligent algorithm iteration logic, and operating condition switching mechanism. The technical system is complete and significantly innovative, fully meeting the requirements for inventiveness and novelty examination of invention patents. Furthermore, this invention does not require large-scale infrastructure upgrades to existing village-level micro-energy station hardware; it can be directly deployed based on existing equipment networks, adapting to the current development needs of large-scale rural distributed energy promotion and effectively supporting the implementation of the rural energy revolution and rural clean energy revitalization strategy. Attached Figure Description

[0022] Figure 1 This is a diagram illustrating the overall architecture of the rural micro-energy cluster operation and control system in an embodiment of the present invention. Figure 2 This is a flowchart of the multi-factor dynamic weight adaptation process in an embodiment of the present invention; Figure 3 This is a block diagram illustrating the principle of the spatiotemporal dynamic weight adaptive mechanism in an embodiment of the present invention. Figure 4 This is a schematic diagram of the timing of multi-timescale rolling optimization driven by digital twin in an embodiment of the present invention; Figure 5 This is a timing diagram of the ADMM distributed collaborative solution interaction in an embodiment of the present invention; Figure 6 This is a logic diagram of cross-station electric-thermal dual-energy mutual assistance and weak communication autonomous switching in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be described below with reference to the accompanying drawings. The described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without innovative effort are within the scope of protection of the present invention.

[0024] (I) Hierarchical Distributed Intelligent Operation Control System for Rural Micro-energy Clusters Based on Multi-Factor Dynamic Weight Adaptation The system adopts a three-level hierarchical distributed physical architecture consisting of a cluster coordination layer, a micro-station autonomy layer, and a device execution layer. It has six built-in core functional modules: a multi-factor index construction module for rural scenarios, a spatiotemporal dynamic weight adaptive adaptation module, a digital twin multi-timescale optimization module, an ADMM distributed collaborative solution module, a cross-station multi-energy mutual assistance module, and a weak communication autonomy and distribution network adaptation module.

[0025] 1. Three-tiered distributed physical architecture Each level only interacts with boundary data, adopting an "edge computing + local autonomy" model, which reduces the amount of communication data by 78% compared to a centralized architecture, and controls the end-to-end latency to within 100ms.

[0026] Cluster Coordination Layer: Deployed at township / town-level energy management stations, configured with two redundant edge computing servers (primary / backup architecture), each with a computing power of no less than 16 TOPS. It incorporates a global optimization scheduling unit, a cross-station mutual assistance management unit, a distribution network interaction unit, an ADMM distributed solver, and a carbon footprint accounting unit. Core functions include: global multi-objective collaborative optimization, cross-station power / heat mutual assistance plan formulation, cluster aggregation participation in distribution network demand response and carbon trading, and global governance of transformer area voltage and three-phase imbalance. This layer only collects aggregated data such as boundary power, operating status, and dual variables of each station, without acquiring detailed data at the station's equipment level, ensuring data privacy and low communication overhead.

[0027] Micro-energy station autonomous layer: Deployed in the control room of each village-level micro-energy station, equipped with a station-level edge controller (computing power not less than 2 TOPS), and includes an in-station multi-energy flow optimization unit, weight adaptation execution unit, autonomous switching unit, local data storage unit, and distribution network terminal management unit. Core functions include: in-station electricity-heat-gas multi-energy flow optimization scheduling, equipment-level control command generation, independent autonomous operation during communication interruptions, and rapid local voltage / reactive power adjustment. This layer has complete offline operation capabilities, locally stores 7 days of historical data and optimization models, and can seamlessly switch to autonomous mode during communication interruptions.

[0028] The equipment execution layer includes distributed power supply units (residential photovoltaic strings, distributed wind turbines, straw biomass generator sets, biogas generator sets), energy storage units (lithium iron phosphate electrochemical energy storage, thermal storage tanks, cold storage devices), energy conversion units (air source heat pumps, ground source heat pumps, electric boilers, biomass boilers), tiered load units, and sensor terminal units within each micro-energy station. All equipment supports the Modbus / 61850 protocol, with a sampling frequency of no less than 1Hz, and a sampling frequency of no less than 10Hz for critical equipment. The command execution response time does not exceed 200ms.

[0029] 2. Multi-factor indicator construction module for rural scenarios A quantitative indicator system of 18 items in 5 categories has been constructed to fully cover the core needs of rural energy systems. Each indicator has a clear quantitative formula, constraint threshold and normalization method, which is different from the general urban microgrid indicators.

[0030] (1) Economic dimension (3 items) Total daily operating cost of the cluster Includes electricity purchase costs, biomass raw material costs, equipment operation and maintenance costs, and demand response revenue deductions. in, Let t be the electricity purchase price. For the unit cost of biomass raw materials, For unit operation and maintenance costs, Revenue per unit of demand response.

[0031] Overall Rate of Return on Demand Response: The ratio of demand response revenue to operating costs. Biomass feedstock cost percentage: The proportion of biomass fuel cost in total operating costs (2) Environmental protection dimension (4 items) Carbon emission intensity: the amount of carbon dioxide emitted per unit of electricity supplied; Renewable energy integration rate: ;in, To absorb the electricity generated by renewable energy sources, This represents the total electricity generated from renewable energy sources.

[0032] Straw resource utilization rate: the proportion of straw consumed locally to the total amount of collectable straw; Pollutant emission reduction rates: SO2, NO x The proportion of smoke and dust emission reduction; (3) Reliability dimension (4 items) Power supply reliability: the proportion of effective power supply time to total time; Duration of power outages in core agricultural loads : Cumulative power outage time for Level 1 loads; Load shortage rate The percentage of electricity shortage in total electricity demand; Mean Time To Recovery (MTBF): The average recovery time for a single fault. (4) Resilience dimension (3 items) Duration of core load supply under extreme operating conditions Duration of continuous power supply to primary loads in islanded mode Reserve capacity adequacy: in, Total available standby capacity This represents peak load.

[0033] Fault self-healing success rate: The percentage of faults that can be recovered without human intervention; (5) Network source adaptation dimensions (4 items) Distribution network area voltage qualification rate: the percentage of time the voltage is within the qualified range; Peak load rate of transformer area: The ratio of peak load of transformer area to rated capacity; Low-voltage line loss rate: The proportion of line loss in a transformer area to the total power supply; Three-phase unbalance: Three-phase current unbalance on the low-voltage side; All indicators are normalized to the [0,1] interval using the extreme value method. The larger the positive indicator, the better, and the smaller the negative indicator, the better.

[0034] 3. Spatiotemporal dynamic weight adaptive adaptation module It adopts a four-level mechanism of "initial weight generation for typical scenarios + real-time dynamic adjustment through deep reinforcement learning + spatiotemporal weight distribution + multi-objective conflict resolution" to break through the bottleneck of scenario adaptation of traditional fixed weights.

[0035] (1) Initial weight generation for typical scenarios Based on FAHP, a 5th-order fuzzy decision matrix is ​​constructed, and basic weight matrices are generated for four typical rural scenarios: busy irrigation season, winter heating season, peak renewable energy season, and extreme weather events. The fuzzy consistency matrix satisfies: In the formula, i and j are the subscripts of the five evaluation dimensions, rij represents the fuzzy importance of the i-th indicator relative to the j-th indicator; rji represents the fuzzy importance of the j-th indicator relative to the i-th indicator; rii represents the fuzzy importance of the i-th indicator itself; since there is no priority difference in self-comparison within the same dimension, all diagonal elements are fixed at 0.5, and the sum of the bidirectional importance values ​​of any two sets of dimensions is always equal to 1, ensuring that the fuzzy evaluation logic is complementary and self-consistent.

[0036] The initial weight vector is calculated using row and normalization. And it passed the consistency test (CR<0.1).

[0037] Each component within the weight vector corresponds to one of the five evaluation dimensions of the system, as defined below: As the benchmark weight of the economic dimension, it corresponds to three types of economic quantitative indicators: total daily operating cost of the cluster, comprehensive rate of return on demand response, and proportion of biomass raw material cost, which represent the degree of emphasis of the scheduling strategy on the target of system operating cost and revenue. As the benchmark weight for environmental protection, it corresponds to four types of quantitative environmental protection indicators: carbon emission intensity, renewable energy consumption rate, straw resource utilization rate, and pollutant emission reduction rate, representing the priority of scheduling strategies for clean energy consumption and low-carbon emission reduction targets. As the benchmark weight of the reliability dimension, it corresponds to four types of quantitative reliability indicators: power supply reliability rate, power outage duration of core agricultural loads, load power shortage rate, and average fault recovery time. These indicators represent the basic guarantee requirements for stable power supply and reduced power outages of agricultural loads under daily operation. The benchmark weights for resilience assurance dimensions correspond to three types of resilience quantitative indicators: core load supply duration under extreme operating conditions, reserve capacity adequacy, and fault self-healing success rate, focusing on the ability to ensure continuous power supply under extreme disasters, severe weather, and equipment failure scenarios. The baseline weights for distribution network adaptation dimensions correspond to four types of distribution network quantitative indicators: distribution network area voltage qualification rate, distribution area peak load rate, low voltage line loss rate, and three-phase imbalance. These indicators are used to constrain network-source adaptation issues such as voltage exceeding limits, line loss, and three-phase current imbalance in rural distribution networks.

[0038] Initial weight configuration for four typical scenarios: (2) Real-time dynamic adjustment of deep reinforcement learning The DDPG reinforcement learning algorithm is used to achieve continuous adaptive adjustment of weights.

[0039] state space (12 dimensions): Source load prediction deviation rate, distribution network area load rate, meteorological warning level, scene identification code, voltage deviation rate, biomass raw material storage ratio, average energy storage SOC, three-phase imbalance, time period code, irrigation load ratio, heating load ratio, and communication quality score.

[0040] Action space (5 dimensions): Weight adjustment of the five major dimension indicators The constraints are satisfied: reward function : in, Let i be the normalized value of the i-th dimension index. Penalty for exceeding voltage limits, Penalty for power outages affecting core loads. Penalty for communication interruption This is the penalty coefficient.

[0041] Network structure: The Actor network is a 4-layer fully connected network (128-64-32-5) using the ReLU activation function, with the output layer using Tanh activation; the Critic network is a 4-layer fully connected network (128-64-32-1), with the input being a concatenation of state and action. A soft update mechanism for the target network is used, with update coefficient τ=0.001.

[0042] (3) Spatiotemporal weight distribution mechanism: In the spatial dimension, different station-level weight correction coefficients are set for the functional positioning of different micro-stations in the cluster (agricultural village, central village, residential village); in the temporal dimension, time period weight correction factors are set for different time periods of peak, valley and flat during the day, so as to realize refined weight adaptation in two spatiotemporal dimensions.

[0043] (4) Multi-objective conflict resolution unit When multiple objectives have strong conflicts (such as the inability to achieve both economy and reliability), the NSGA-II algorithm is introduced to generate the Pareto optimal solution set. The optimal solution is selected based on the entropy weight method and the TOPSIS method to ensure that the global performance is optimal under the premise that the core constraints do not exceed the limits.

[0044] 4. Digital twin-driven multi-timescale rolling optimization module A digital twin of a micro-energy cluster is constructed, which includes physical entities, virtual models, data interaction, and an optimization decision-making closed loop. Three-level time-scale rolling optimization is performed to address the strong randomness of rural energy sources and loads.

[0045] (1) Digital Twin Architecture Physical layer: Includes all field devices and sensing terminals, collecting operational data in real time. Model layer: includes equipment mechanism model, network topology model, weather forecasting model, load forecasting model, and agricultural production planning model. Data layer: Integrates historical data, real-time data, and predictive data to achieve second-level data updates. Application layer: Supports functions such as optimized scheduling, state assessment, fault diagnosis, and simulation. Digital twin data update frequency: device-level data 1 second / time, system-level status 5 seconds / time, prediction data 15 minutes / time.

[0046] (2) Rolling optimization at three time scales The day-ahead optimization layer (24h, 1h step size): Based on the next day's weather forecast, load forecast, and agricultural production plan, it aims to solve for the global operational baseline plan with a multi-objective comprehensive optimization using dynamic weights. Optimization variables include the output plans of each unit, energy storage charging and discharging plans, inter-station mutual assistance plans, and demand response reporting plans. It employs mixed-integer linear programming (MILP) for solving the problem, with a solution time not exceeding 5 minutes.

[0047] Intraday Rolling Layer (4-hour rolling window, 15-minute step): Source and load ultra-short-term forecasts are updated every 15 minutes to correct day-ahead planning deviations. A Model Predictive Control (MPC) framework is adopted to continuously optimize scheduling instructions in the time domain and smooth out short-term fluctuations. Optimization variables include unit output fine-tuning, energy storage charging and discharging power adjustment, and adjustable load switching.

[0048] Real-time correction layer (5s step size): This layer performs rapid correction control to address sudden fluctuations on the order of seconds (such as a sudden drop in photovoltaic power due to cloud cover). Based on local measurement data, it smooths out power fluctuations and maintains stable voltage and frequency through rapid charging and discharging of energy storage and reactive power compensation devices. This layer primarily operates at the station-level autonomous layer and requires no intervention from the cluster layer.

[0049] 5. ADMM Distributed Collaborative Solving Module The Alternating Direction Multiplier Method (ADMM) is used to achieve cluster-level distributed optimization. It can approximate the global optimum by exchanging boundary information only, which greatly reduces communication overhead.

[0050] (1) Problem decomposition The global optimization problem is decomposed into one cluster coordination layer subproblem and N microstation autonomous layer subproblems. The global objective function is: in, Let be the local optimization variable for the i-th micro-energy station. This is a global coupling variable (boundary mutual aid power). For the station-level objective function, This is the cluster-level objective function.

[0051] (2) Augmented Lagrange function in, As dual variables, This is a penalty factor.

[0052] (3) Iteration steps Each micro-energy station solves its local subproblem in parallel and updates its data. The cluster coordination layer collects boundary information from each station and updates global variables. Each station updates its dual variable Verify convergence conditions: If the condition is not met, return to step 1. Convergence accuracy Punishment factor The typical number of iterations is 8 to 15, and the total time for a single round of solution does not exceed 30 seconds.

[0053] 6. Cross-site multi-functional mutual assistance and weak communication autonomous module (1) Cross-station electricity-heat dual-energy mutual assistance mechanism is based on the existing low-voltage power distribution lines and village-level heating pipe networks within the cluster to establish an electricity / heat dual-energy mutual assistance network. A mutual assistance path optimization model is constructed to comprehensively consider network loss, heat loss, transmission distance, and channel capacity to optimize the mutual assistance power and path between stations.

[0054] Mutual power constraint: in, Let be the mutual aid power from station i to station j at time t. This represents the maximum transmission capacity of the channel.

[0055] Mutual aid benefits are distributed according to contribution: (2) The weak communication autonomous handover mechanism adopts a dual mechanism of heartbeat detection + data verification. The heartbeat period is 500ms. If no cluster layer instruction is received for 3 consecutive times, the communication is determined to be interrupted.

[0056] After the communication interruption is triggered: The station-level autonomous handover unit completes mode switching within 20ms. Scenario weights and optimization models for calling local storage Implement in-site autonomous optimization to prioritize core load. Locally record runtime data, and synchronize in batches after communication is restored. After communication is restored, a smooth transition strategy is adopted, gradually returning to cluster coordination mode within 30 seconds to avoid power surges. (3) The refined adaptation of the distribution network embeds the voltage management of transformer areas, the suppression of three-phase imbalance, and the optimization of line loss into the collaborative control logic: Voltage management: Through the combined use of distributed power reactive power regulation, energy storage reactive power support, and on-load tap-changing transformers, the voltage in the distribution area is controlled between 0.93 and 1.07 pu. Three-phase imbalance management: By optimizing load phase sequence switching and independently controlling the three phases of energy storage, the three-phase imbalance is controlled within 15%. Line loss optimization: Optimize reactive power distribution, improve power factor, and reduce line loss in transformer areas. (II) A hierarchical distributed intelligent operation control method for rural micro-energy clusters based on multi-factor dynamic weight adaptation Includes the following steps: S1. System Architecture Setup and Initialization A three-tiered distributed control architecture consisting of a cluster coordination layer, a micro-station autonomous layer, and an equipment execution layer was established. Hardware devices and communication networks at each level were configured. Data interaction boundaries, communication protocols, and data formats at each level were defined. ADMM distributed solver parameters and communication interruption autonomous switching logic were configured. Equipment ledger entry, topology modeling, and parameter calibration were completed.

[0057] S2. Construction of a Multi-Factor Indicator System for Rural Scenarios Establish a quantitative indicator system of 18 items covering five dimensions: economy, environmental protection, reliability, resilience, and grid-source adaptation; clarify the quantitative calculation formula, constraint threshold, and normalization method for each indicator; incorporate rural-specific constraints such as the priority of agricultural production load classification, seasonal supply of biomass raw materials, and carrying capacity of low-voltage distribution network areas; and complete the scenario-based calibration of indicator weights and constraint thresholds.

[0058] S3. Dynamic Weight Model Training and Initialization A fuzzy judgment matrix is ​​constructed based on the fuzzy hierarchical analysis method, and the initial weight matrices of four typical scenarios are calculated and passed through consistency test. A DDPG reinforcement learning model is constructed and offline training is performed using historical running data, with no less than 5,000 training rounds. After training, the model is deployed to each station controller and the weight adaptation module is initialized.

[0059] S4, Digital Twin Construction and Multi-Timescale Optimization Configuration Construct a digital twin of a micro-energy cluster, establish equipment mechanism model, network topology model, and source-load prediction model; configure three-level time scale parameters: day-ahead optimization, intraday rolling optimization, and real-time correction; configure the optimization solver and rolling window parameters; and complete the virtual-real mapping calibration between the digital twin and the physical system.

[0060] S5, Current Global Optimization Scheduling The day-ahead optimization is initiated at 00:30 daily; the next day's weather forecast, load forecast, agricultural production plan, and electricity price information are imported; the next day's scenario type is identified and the corresponding initial weights are loaded; the DDPG model generates a plan for intraday weight changes; the cluster coordination layer solves the global day-ahead optimization plan using the ADMM algorithm and issues the next day's boundary power benchmark, cross-station mutual assistance plan, and demand response application plan to each micro-energy station.

[0061] S6, Intraday Rolling Optimization Adjustments Intraday rolling optimization is initiated every 15 minutes; ultra-short-term source-load forecast data and real-time operating status are updated; the DDPG model dynamically adjusts the weights of each dimension according to the real-time status; the cluster coordination layer solves the rolling optimization problem through ADMM and issues corrected boundary power commands to each station; each micro-energy station adjusts the operating status of its equipment according to the commands.

[0062] S7, Real-time Calibration and Device Execution Real-time correction is performed every 5 seconds; the station-level autonomous layer uses local real-time measurement data to quickly charge and discharge energy storage and smooth out sudden power fluctuations through reactive power regulation; the equipment execution layer receives and executes control commands with a response time of no more than 200ms; real-time collection of operating data is uploaded to the digital twin layer by layer to form a closed-loop control.

[0063] S8, Communication Anomaly Handling and Autonomous Operation The system monitors the communication status in real time. If a communication interruption is detected, the micro-station automatically switches to local autonomous operation mode. It calls the local optimization model and scene weights to perform in-station autonomous optimization, prioritizing power supply to core loads. After communication is restored, it automatically synchronizes the data during the interruption and smoothly transitions back to cluster collaborative mode.

[0064] S9. Operational Evaluation and Model Iteration After each day's operation, an operation evaluation report is automatically generated, which summarizes the completion status of various indicators; the DDPG model is fine-tuned online based on the day's operation data to continuously optimize the weight adaptation accuracy; and the digital twin model parameters are updated regularly to ensure model accuracy.

[0065] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computers or available storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0069] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rural micro-energy cluster operation control system, comprising: Field equipment group, multi-level control hardware architecture, and multiple sets of functional hardware modules; The field equipment group includes distributed power sources, energy storage devices, energy conversion devices, tiered load units, and sensor acquisition terminals; the multi-level control hardware architecture consists of cluster coordination hardware units, micro-station autonomous hardware units, and equipment execution terminals connected sequentially from top to bottom; its key feature is: The cluster coordination hardware unit is deployed at township energy stations and configured with primary and backup redundant edge computing servers. The micro-energy station autonomous hardware unit is deployed one-to-one to each village micro-energy station and has built-in local solid-state storage media and station-level edge controller. The device execution terminal connects to all equipment in the field equipment group for electrical and signal communication. The cluster coordination hardware unit and each micro-energy station autonomous hardware unit only establish a point-to-point communication link for boundary aggregation data. The micro-energy station autonomous hardware unit and the device execution terminal establish a real-time control signal transmission link. The local computing of the two-level edge hardware reduces the amount of channel data transmission. The local solid-state storage media stores model files and historical operating data to support offline management and operation of the entire station after communication is lost. The system's overall hardware link carries six sets of hardware functional modules, including a multi-factor index construction hardware module, a spatiotemporal dynamic weight adaptation hardware module, a digital twin mapping hardware module, an ADMM distributed solution hardware module, a cross-site multi-energy mutual assistance management hardware module, and a weak communication autonomous distribution network adaptation hardware module. These modules are integrated and deployed within the cluster coordination hardware unit and the micro-energy station autonomous hardware unit, and data interaction between the modules is completed through a hardware bus.

2. The rural micro-energy cluster operation control system according to claim 1, characterized in that, The hardware module for constructing multi-factor indicators is equipped with hardware discrimination logic units that incorporate five categories of quantitative indicators. Each category of indicator is equipped with a hardware-integrated constraint threshold register and normalization processing circuit unit. The five categories of indicators include economic indicator units, environmental indicator units, reliability indicator units, resilience assurance indicator units, and distribution network adaptation indicator units. The resilience assurance indicator unit hardware is written with logic for extreme load supply guarantee, backup capacity verification, and fault self-healing judgment. The distribution network adaptation indicator unit hardware integrates voltage qualification judgment, transformer area load monitoring, line loss calculation, and three-phase imbalance detection circuits.

3. The rural micro-energy cluster operation control system according to claim 1, characterized in that: The spatiotemporal dynamic weight adaptation hardware module integrates three layers of hardware processing units: the first layer is the scene benchmark weight storage unit, which pre-stores benchmark weight parameters corresponding to multiple types of typical rural working conditions. The second layer is the DDPG network hardware acceleration computing unit, which accesses the system's global status acquisition signals and outputs weight correction values; The third layer is the time-space temperature correction register, which has built-in village type correction parameters and time period correction parameters; the module is equipped with a conflict resolution hardware processing chip, which integrates a Pareto solution set calculation unit and an optimal solution screening unit.

4. The rural micro-energy cluster operation control system according to claim 1, characterized in that: The digital twin mapping unit establishes bidirectional data links with the field sensing terminal, the cluster coordination unit, and the micro-energy station autonomous unit, respectively. It relies on the multi-level cache built into the device to distinguish the refresh frequency of device data, system status data, and prediction data. It relies on the edge computing device to carry a three-level scheduling channel of day-ahead planning, intraday rolling, and real-time correction to alleviate the problems of large randomness of source load and lag in single-layer scheduling to suppress disturbances. ADMM distributed solution units are configured in township edge computing devices and village-level edge controllers, respectively. The overall optimization task is decomposed into two parallel processing links: global coupled computing and local site computing. Iterative convergence calculation is completed by relying on the dual variable data interaction link between devices, which improves the drawbacks of centralized optimization, such as concentrated computing power and excessively large interactive data volume. The cross-station multi-energy mutual assistance control unit is electrically connected to the power distribution acquisition device of the distribution area and the village-level heating network monitoring terminal. Relying on the mutual assistance path selection and revenue distribution metering logic built into the edge controller, it completes the bidirectional allocation of electric and heat energy between stations by using the existing power distribution network and heating network on site, thus solving the shortcomings of isolated energy storage and heat storage resources of each station and the inability to share them in a cluster. The weak communication autonomous distribution network adaptation unit electrically accesses the communication status detection device, relying on the edge controller's built-in logic configuration millisecond-level operation mode switching function, and simultaneously links with the field reactive power compensation device and energy storage control equipment to form a linkage control link for voltage regulation, three-phase imbalance management, and line loss optimization. It is used to quickly switch to autonomous operation mode when communication is abnormal, and smoothly switch back to cluster collaborative control after communication is restored.

5. A method for controlling the operation of rural micro-energy clusters, comprising scheduling and control using the rural micro-energy cluster operation control system described in any one of claims 1 to 4; characterized in that, Includes the following steps: S1. Complete the networking and connection of the three-level management and control architecture devices, input the device parameter calibration, initialize the parameters of each functional unit, and agree on the hierarchical data transmission protocol and the autonomous mode switching trigger conditions. S2. Load all quantitative indicator parameters into the multi-factor indicator construction unit, and input the agricultural load limit, biomass supply cycle, and distribution network carrying capacity limit exclusive constraint parameters. S3. Input the scene baseline weights and complete the offline training and calibration of the reinforcement learning model, complete the calibration of the digital twin virtual-real mapping channel and the configuration of the three-level scheduling timing parameters; S4. The cluster coordination unit relies on the township edge equipment to perform daily day-ahead optimization solutions and issue basic operation instructions for unit output, energy storage operation, and cross-station energy mutual assistance. S5. Each micro-energy station autonomous unit relies on the station-level controller to periodically perform intraday rolling optimization, and uses the weight adaptation unit to correct the optimization target weight in real time and update the control instructions of the station equipment. S6. Real-time correction control is performed based on on-site measured data to suppress instantaneous power fluctuations, and the operating data is transmitted back to the digital twin system to form a closed-loop management and control system. S7. Real-time monitoring of communication link status; automatic switching to autonomous operation and priority to ensure power supply to agricultural primary loads after communication interruption; and smooth access to the cluster for collaborative operation after communication is restored. S8. After the daily operation is completed, the operation indicators are statistically analyzed, and the parameters of the reinforcement learning model and digital twin model are iteratively optimized based on the actual test data.

6. The rural micro-energy cluster operation control method according to claim 5, characterized in that: The digital twin collects data in layers according to a preset cycle, and the three-level scheduling processes the data according to a predetermined time step to match the source load fluctuations and disturbances at different time scales.

7. The rural micro-energy cluster operation control method according to claim 5, characterized in that: In the ADMM distributed solution process, each micro-station calculates local variables in parallel, the cluster coordination unit updates global coupled variables, each station updates dual variables synchronously, and iterative calculations are performed until the convergence criteria are met.

8. The rural micro-energy cluster operation control method according to claim 5, characterized in that: The weight adjustment process involves retrieving the baseline weights for the corresponding working conditions, correcting the weight values ​​through a reinforcement learning model, and then superimposing spatiotemporal correction parameters to complete the calibration. When multiple objectives conflict, the Pareto solution set is obtained and the optimal control scheme is selected and implemented.

9. The rural micro-energy cluster operation control method according to claim 5, characterized in that: The system uses a heartbeat message combined with a data verification mechanism to identify communication faults. During the autonomous operation phase, it prioritizes the power supply to agricultural loads. Under both cluster collaboration and local autonomy operation modes, it simultaneously completes the control of transformer area voltage, suppression of three-phase imbalance, and optimization control of low-voltage line loss.

Images