Micro-grid multi-objective dispatching optimization control method
By constructing a three-layer time-scale framework, adaptive adjustment of multi-objective weights, and differentiated instruction allocation, the scheduling problem of microgrids under uncertainty is solved, multi-objective collaborative optimization and safe and stable operation are achieved, and the operating efficiency and equipment utilization of microgrids are improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIAN SI TOP ELECTRIC CO LTD
- Filing Date
- 2026-05-06
- Publication Date
- 2026-06-05
AI Technical Summary
Existing microgrid multi-timescale scheduling methods suffer from problems such as large power fluctuations, imbalance between economic efficiency and low carbon emissions, unreasonable energy storage margin, mismatched equipment response, and easy overstepping of rigid constraints when facing uncertainties. They also lack flexible interaction and collaborative iteration mechanisms, resulting in low operating efficiency.
By collecting data and dividing the time domain, a three-layer decreasing time scale framework of daily planning, intraday rolling correction and real-time adjustment is constructed. Multi-objective weight adaptive adjustment is adopted, virtual charge and discharge margin is introduced, differentiated instruction allocation is implemented, and a dynamic tightening mechanism for safety constraints is established to form a closed-loop verification and update system, so as to achieve multi-objective collaborative optimization and safe and stable operation.
It significantly improves the scheduling targeting and robustness of microgrids in complex and uncertain environments, smooths power fluctuations, optimizes energy storage utilization, delays equipment lifespan degradation, improves equipment response efficiency and operational safety, and achieves synergistic optimization of economy, low carbon emissions and stability.
Smart Images
Figure CN122159385A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of optimization control, and specifically relates to a multi-objective scheduling optimization control method for microgrids. Background Technology
[0002] Multi-objective scheduling optimization control methods for microgrids generally adopt a hierarchical architecture with multiple time scales, dividing the operation process into three stages: day-ahead planning, intraday rolling, and real-time adjustment. While this decoupling of the time dimension reduces the complexity of single-stage optimization problems, it fundamentally introduces the inherent problem of cross-time scale coordination mismatch.
[0003] During the daytime dispatch phase, the optimization model relies excessively on forecast data for photovoltaic, wind power, and load over the next 24 hours, while long-term forecasts inevitably contain significant biases. In pursuit of economic optimization, the model often arranges key adjustment resources such as energy storage state of charge, interruptible load capacity, and gas turbine start-up and shutdown on the time axis in an extremely compact manner, compressing its adjustability margin to near the boundary and failing to reserve sufficient flexibility for the intraday phase.
[0004] After entering the intraday rolling optimization phase, with the shortening of the prediction time domain and the improvement of accuracy, the actual operating trajectory has deviated significantly from the day-ahead plan. Although intraday optimization theoretically has the ability to correct the plan every 15 to 30 minutes, because the energy storage charging and discharging power has been arranged to the limit and the controllable power supply ramp-up capability has been exhausted in the day-ahead phase, intraday optimization has no sufficient adjustment capacity available when facing actual fluctuations. It can only be forced to perform deep charging and discharging near the lower limit of energy storage capacity, which greatly accelerates the degradation of energy storage life and completely sacrifices the key indicator among the multiple objectives of battery health status.
[0005] During the real-time control phase, when encountering rapid fluctuations such as a sudden drop in photovoltaic power within minutes, although millisecond-level real-time control has the ability to respond, its adjustment range is limited to the local power point adjustment of fast-moving equipment such as energy storage converters and static synchronous compensators. It cannot substantially change discrete decision variables such as the unit start-up and shutdown status and large-capacity load switching plans that have been fixed in the day-ahead and intraday phases. It can only make fine adjustments near the weakened operating point and cannot restore the overall coordination of multi-objective optimization at the system level.
[0006] The unidirectional nature of information interaction and the fragmented nature of constraint transmission across multiple time scales make day-ahead optimization results a strong constraint, while intraday optimization and real-time control become passive local correction links, forming a typical "tight at the beginning and loose at the end" cooperative mismatch pattern. From the perspective of the essence of multi-objective optimization, objectives such as economic cost, energy storage life, and voltage quality have drastically different response characteristics and adjustment costs at different time scales. Existing methods mechanically divide the three time scales, constructing independent multi-objective functions for each stage, lacking a dynamic coordination mechanism. This leads to the day-ahead stage pushing the operating point to the edge to save a small amount of fuel cost, while the real-time stage is forced to maintain stability at the cost of several times the equipment wear and tear, resulting in global operation that is far from optimal.
[0007] A deeper problem lies in the fact that current multi-timescale scheduling methods handle uncertainty statically. During the day-ahead phase, robust optimization or stochastic programming generates a deterministic operating baseline curve, while the intraday phase relies on rolling optimization based on model predictive control. Neither framework possesses a closed-loop feedback mechanism. Intraday optimization cannot effectively feed back deviation information observed at the real-time layer to the day-ahead plan correction, causing prediction errors to accumulate and amplify over time. Under extreme conditions, such as when a microgrid is forced into islanded operation, the day-ahead scheduling results optimized based on grid-connected mode are simply not feasible. Furthermore, the intraday and real-time control layers lack the ability to reconstruct the multi-objective optimization framework in a short time, only managing to maintain power balance by cutting non-critical loads, completely disregarding economic, environmental, and other objectives.
[0008] Existing methods generally employ a rigid nesting approach when dealing with the coupling constraints of optimization models at different time scales. This treats day-ahead decision variables as fixed boundaries for intraday optimization and intraday optimization results as rigid reference trajectories for real-time control, neglecting the flexible interactions and collaborative iterative relationships that should exist between different time scales. This rigid coupling mechanism is highly susceptible to instability and oscillations in the command chain when faced with non-ideal conditions such as communication delays, measurement errors, and execution deviations. This manifests as abnormal operating conditions such as frequent switching of charging and discharging modes in energy storage and repeated start-ups and shutdowns of gas turbines, severely reducing equipment lifespan and significantly diminishing overall operating efficiency.
[0009] The core problem of multi-timescale coordination mismatch lies in the fact that the existing architecture artificially divides the time dimension, which should be continuous and unified, into multiple independent optimization stages. The lack of two-way information interaction, flexible constraint transmission, and adaptive goal coordination mechanism between each stage leads to a vicious cycle of rigid coupling and gradual dissipation of adjustment margin among the daily planning, intraday correction, and real-time control. Summary of the Invention
[0010] This invention proposes a multi-objective scheduling optimization control method for microgrids, which solves the problems in existing technologies such as large power fluctuations, imbalance between economy and low carbon emissions, unreasonable energy storage margin, equipment response mismatch, and easy over-limit of constraints in microgrid scheduling under different uncertainty scenarios, and achieves multi-objective collaborative optimization and safe and stable operation.
[0011] The technical solution of this invention is implemented as follows: a multi-objective scheduling optimization control method for microgrids, the method comprising the following steps: Data acquisition and time domain division: Collect core operation data of the microgrid according to a set period; combine the data error distribution to divide the future scheduling period into high and low uncertainty time domains, corresponding to the time periods when the prediction error exceeds the first threshold and is lower than the second threshold, respectively; Layered optimization framework construction: A three-layer framework is constructed, consisting of a daily planning layer, an intraday rolling correction layer, and a real-time adjustment layer. The time scales are successively decreasing to realize millisecond-to-second power compensation for initial scheduling, dynamic correction, and rapid response equipment, respectively. Multi-objective weight adaptive adjustment: The intraday correction layer constructs a multi-objective model that minimizes cost, carbon emissions, and power fluctuations, and adjusts the weights according to the time domain type. In the high-uncertainty time domain, the power fluctuation weight is increased to suppress shocks; in the low-uncertainty time domain, the cost and carbon emission weights are increased. Energy storage charge / discharge margin correction: Introduce virtual charge / discharge margin, which is calculated based on energy storage status, remaining lifetime and high uncertainty time domain duration; use the high uncertainty time domain to compress the upper and lower limits of charge and discharge for reserve, and release the margin through the low uncertainty time domain to restore rated capacity; Differentiated instruction allocation: The equipment is divided into a basic group and a fast group according to the response speed and accuracy. When the instantaneous deviation of the steady-state deviation exceeds the third threshold in the basic group, the droop control of the fast group responds directly and is not constrained by the optimization cycle. Dynamic tightening of safety constraints: Establish a dynamic tightening mechanism to adjust constraints according to operating status and time domain type; reduce the state of charge of energy storage through the high uncertainty time domain, and use the upper and lower limits of the power exchange of the large power grid to reserve margins, and restore the rated value when using the low uncertainty time domain; Closed-loop verification and update: After the end of each daily rolling cycle, the actual output is compared with the scheduling plan. When the deviation exceeds the fourth threshold, a feedback correction term is introduced. The multi-objective weight adaptive adjustment stage is repeated until the safety constraint dynamic tightening stage, forming a closed-loop control system.
[0012] Most existing microgrid dispatch optimization technologies adopt fixed time domain division and single time scale optimization schemes, which cannot accurately adapt to the uncertainty differences caused by prediction error fluctuations. High uncertainty periods are prone to causing severe power surges, while low uncertainty periods are difficult to fully tap the potential for economic and low-carbon operation. At the same time, multi-objective optimization generally adopts fixed weight settings, which cannot dynamically balance cost, carbon emissions and power fluctuations according to real-time operating scenarios. This results in dispatch results that either excessively pursue stability at the expense of economy, or focus on cost control, causing excessive grid fluctuations, making it difficult to achieve multi-objective synergistic optimization. Traditional energy storage scheduling is mostly based on rated charging and discharging capacity, without dynamically adjusting the charging and discharging margin by taking into account the remaining lifespan, the duration of uncertainty and the real-time status. In high uncertainty scenarios, insufficient energy storage reserves can easily lead to power deficits, while in low uncertainty scenarios, idle margins reduce equipment utilization. Moreover, it is difficult to balance energy storage lifespan and operational safety. Existing equipment command allocation mostly adopts a unified scheduling cycle and control mode, without implementing differentiated allocation according to response speed and accuracy. Steady-state deviations and instantaneous disturbances cannot be handled in a layered manner, making it difficult to take advantage of fast-response equipment. Slow-speed basic equipment is prone to increased losses due to frequent adjustments. At the same time, safety constraints mostly adopt fixed threshold settings, without dynamically tightening or loosening according to operating status and time-domain uncertainty. Insufficient constraint margin during high uncertainty periods can easily trigger the risk of exceeding limits, while excessively loose constraints during low uncertainty periods cannot fully utilize the capacity of the power grid and equipment. In addition, most scheduling schemes are open-loop execution modes, lacking a closed-loop verification and feedback correction mechanism for actual output and planned deviations. Accumulated deviations can easily cause scheduling commands to deviate from actual operating conditions, making it difficult to form a continuous and stable adaptive control system. This technology addresses the aforementioned pain points by accurately identifying uncertainties through data acquisition and time-domain partitioning, constructing a three-layer decreasing time scale framework to adapt to different scheduling needs, adopting multi-objective weight adaptive adjustment to achieve scenario-based objective trade-offs, introducing virtual charge and discharge margins to achieve dynamic energy storage backup and capacity release, implementing differentiated command allocation to improve equipment collaborative response efficiency, establishing a dynamic tightening mechanism for safety constraints to ensure operational safety and improve capacity utilization, and achieving deviation feedback and iterative optimization through closed-loop verification and updates. This overcomes the core technical challenges of dynamic coordination of multiple objectives in uncertain microgrid scenarios, the inability to coordinate and optimize energy storage margins and lifetimes, the mismatch between equipment response characteristics and scheduling cycles, the rigidity of safety constraints leading to easy exceeding of limits, and the accumulation of deviations in open-loop scheduling. It breaks through the technical bottlenecks of traditional scheduling methods in dealing with complex and variable operating scenarios, such as poor adaptability, weak robustness, multi-objective imbalance, and the difficulty in balancing safety and efficiency, providing a feasible solution for the efficient and stable operation of microgrids.
[0013] As a preferred implementation, in the data acquisition and time-domain partitioning steps, the core operating data includes photovoltaic power generation output, wind power generation output, load demand, energy storage status of charge, real-time electricity price, and the exchange power between the microgrid and the main grid; the set periodic acquisition time is 1-5 minutes, the first threshold setting range is 15% to 25%, and the second threshold setting range is 5% to 10%.
[0014] As a preferred implementation, the first threshold is defined as follows: when the error between the predicted output and the actual output of renewable energy exceeds 15% to 25%, the corresponding time period is divided into a high uncertainty time domain, and power fluctuations need to be suppressed first; when the error between the predicted output and the actual output of renewable energy is less than 5% to 10%, the corresponding time period is divided into a low uncertainty time domain; when dividing the time domain, the predicted and actual output data of at least 3 months in history are collected, the error is statistically analyzed and a distribution curve is fitted, and the high and low uncertainty time domains are divided accordingly.
[0015] As a preferred implementation, in the hierarchical optimization framework construction step, the day-ahead planning layer adopts a 24-hour time scale and generates an initial scheduling plan based on the previous day's forecast data, including equipment output, energy storage charging and discharging, and power grid exchange power; the intraday rolling correction layer adopts an hourly updated time scale and updates the strategy in each cycle based on actual data and time domain results; the real-time adjustment layer adopts a second-level time scale and performs power compensation for fast-response equipment.
[0016] As a preferred implementation, in the multi-objective weight adaptive adjustment step, the weight of the power fluctuation minimization objective is increased first in the time domain with high uncertainty, and the sum of the weights of the cost and carbon emission minimization objectives is increased first in the time domain with low uncertainty; a fuzzy PID algorithm is used, combined with prediction error to dynamically correct the weights, to avoid system fluctuations.
[0017] As a preferred implementation, in the energy storage charge and discharge margin correction step, the virtual charge and discharge margin is calculated using a preset formula based on energy storage-related parameters and weighting coefficients; the upper and lower limits of energy storage charge and discharge are reserved for backup in the high uncertainty time domain, and the margin is released in the low uncertainty time domain at fixed intervals to restore the rated energy storage capacity.
[0018] As a preferred embodiment, the virtual charge / discharge margin calculation formula is M=k1×(SOC-SOC) min ) / (SOC max -SOC min M is the virtual charge / discharge margin; k1, k2, and k3 are weighting coefficients, and satisfy k1+k2+k3=1; SOC is the current state of charge of the energy storage system; min State of Charge (SOC) is the lower limit of the rated state of charge of the energy storage system. maxNremaining is the upper limit of the rated state of charge of the energy storage system; Nrated is the remaining charge-discharge cycle life of the energy storage system; Nrated is the rated charge-discharge cycle life of the energy storage system; and Thigh is the duration of the high uncertainty time domain.
[0019] As a preferred implementation, in the differentiated instruction allocation step, the basic adjustment group is responsible for the steady-state power deviation, and the fast adjustment group responds faster; the third threshold is the standard for judging the rated power deviation, and the droop coefficient is adjusted according to the equipment adjustment accuracy.
[0020] After adopting the above technical solution, the beneficial effects of the present invention are: by dividing the time domain into high and low uncertainty according to the data error distribution, the present invention can accurately identify the risk level of future scheduling cycles, avoid the coarse handling of different uncertainty scenarios by the traditional unified scheduling mode, and significantly improve the pertinence and rationality of the scheduling scheme. A three-tiered optimization framework with decreasing time scales—day-ahead planning, intraday rolling correction, and real-time adjustment—is constructed. This framework enables long-cycle initial planning, medium-cycle dynamic correction, and short-cycle rapid power compensation, achieving full-scale coverage from hourly to millisecond-second levels. It balances scheduling planning with real-time response capabilities, effectively mitigating the impact of load and distributed power source fluctuations. The multi-objective weight adaptive adjustment mechanism dynamically allocates optimization weights for cost, carbon emissions, and power fluctuations based on the level of time-domain uncertainty. During periods of high uncertainty, it strengthens power stability to suppress grid impacts, while during periods of low uncertainty, it focuses on economic and low-carbon operation to improve overall efficiency. This solves the problem of not being able to balance multiple objectives under fixed weights, achieving a synergistic improvement in scheduling efficiency and operational stability. By introducing virtual charge and discharge margin and combining it with energy storage status, remaining lifespan, and time domain duration for dynamic correction, the upper limit of charge and discharge is compressed during periods of high uncertainty to reserve backup capacity, ensuring power support capability under extreme scenarios. During periods of low uncertainty, the margin is released to restore rated operating capability, improving energy storage utilization while delaying lifespan degradation, thus achieving a holistic optimization of energy storage safety, lifespan, and operating efficiency. The equipment is divided into basic and fast groups based on response speed and accuracy, and differentiated command allocation is implemented. The basic group is responsible for steady-state deviation to ensure overall power balance. When instantaneous exceedance occurs, the fast group responds without delay through droop control, breaking through the optimization cycle limit to quickly smooth out disturbances. This reduces the losses from frequent operation of basic equipment and fully leverages the adjustment advantages of fast equipment, improving overall response speed and control accuracy. The dynamic tightening mechanism of safety constraints can flexibly adjust the energy storage charge state and the power exchange limit of the main grid according to the operating status and time domain type. During periods of high uncertainty, the constraint threshold is proactively reduced to reserve a safety margin to avoid over-limit faults. During periods of low uncertainty, the rated constraints are restored to improve grid capacity utilization. This maximizes the potential of microgrid operation while ensuring operational safety. The closed-loop verification and update process compares the actual output with the scheduling plan in real time. When deviations exceed the limit, feedback correction and iterative optimization are initiated to form a closed-loop control system. This effectively suppresses the accumulation of deviations and ensures that scheduling instructions always conform to actual operating conditions, significantly improving the robustness and reliability of the scheduling scheme. Overall, this method comprehensively enhances the adaptive scheduling capability of microgrids in complex and uncertain environments, achieving multi-dimensional synergistic optimization of economy, low carbon emissions, stability, and security, and providing strong technical support for the efficient, reliable, and green operation of microgrids. Attached Figure Description
[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 skilled in the art without creative effort are within the scope of protection of the present invention.
[0024] Example 1, as Figure 1 As shown, a multi-objective scheduling optimization control method for microgrids includes the following steps: Data acquisition and time domain division: Collect core operation data of the microgrid according to a set period; combine the data error distribution to divide the future scheduling period into high and low uncertainty time domains, corresponding to the time periods when the prediction error exceeds the first threshold and is lower than the second threshold, respectively; Layered optimization framework construction: A three-layer framework is constructed, consisting of a daily planning layer, an intraday rolling correction layer, and a real-time adjustment layer. The time scales are successively decreasing to realize millisecond-to-second power compensation for initial scheduling, dynamic correction, and rapid response equipment, respectively. Multi-objective weight adaptive adjustment: The intraday correction layer constructs a multi-objective model that minimizes cost, carbon emissions, and power fluctuations, and adjusts the weights according to the time domain type. In the high-uncertainty time domain, the power fluctuation weight is increased to suppress shocks; in the low-uncertainty time domain, the cost and carbon emission weights are increased. Energy storage charge / discharge margin correction: Introduce virtual charge / discharge margin, which is calculated based on energy storage status, remaining lifetime and high uncertainty time domain duration; use the high uncertainty time domain to compress the upper and lower limits of charge and discharge for reserve, and release the margin through the low uncertainty time domain to restore rated capacity; Differentiated instruction allocation: The equipment is divided into a basic group and a fast group according to the response speed and accuracy. When the instantaneous deviation of the steady-state deviation exceeds the third threshold in the basic group, the droop control of the fast group responds directly and is not constrained by the optimization cycle. Dynamic tightening of safety constraints: Establish a dynamic tightening mechanism to adjust constraints according to operating status and time domain type; reduce the state of charge of energy storage through the high uncertainty time domain, and use the upper and lower limits of the power exchange of the large power grid to reserve margins, and restore the rated value when using the low uncertainty time domain; Closed-loop verification and update: After the end of each daily rolling cycle, the actual output is compared with the scheduling plan. When the deviation exceeds the fourth threshold, a feedback correction term is introduced. The multi-objective weight adaptive adjustment stage is repeated until the safety constraint dynamic tightening stage, forming a closed-loop control system.
[0025] This embodiment provides a multi-objective scheduling optimization control method for microgrids, applicable to grid-connected microgrids that include photovoltaic power generation systems, wind power generation systems, energy storage systems, micro gas turbines, and interruptible loads. The method is implemented in the following steps.
[0026] First, data acquisition and time domain segmentation are performed. A data acquisition server is deployed in the microgrid energy management system to collect data on photovoltaic (PV) output, wind power output, load demand, energy storage status of charge, real-time electricity price, and power exchange between the microgrid and the mains grid at a set interval of 3 minutes. Historical predicted and actual output data for the past 6 months are collected, and the prediction error for each time period is statistically analyzed and fitted with a normal distribution curve. Time periods with prediction errors exceeding 20% are classified as high uncertainty time domains, time periods with prediction errors below 8% are classified as low uncertainty time domains, and the remaining time periods are classified as regular time domains. The high uncertainty time domain corresponds to periods of severe PV power fluctuations under cloudy weather conditions, while the low uncertainty time domain corresponds to periods of stable PV power output under sunny weather conditions.
[0027] Secondly, a hierarchical optimization framework is constructed. The day-ahead planning layer uses a 24-hour timescale. Based on the previous day's 24-hour forecast data, it employs a mixed-integer linear programming algorithm to generate an initial scheduling plan, determining the start-stop status of micro gas turbines, the charging and discharging power sequence of the energy storage system, the interruptible load dispatch plan, and the power exchange curve with the main grid within each hourly segment. The intraday rolling correction layer updates every hour. At the beginning of each cycle, it collects current actual operating data and, combined with the latest ultra-short-term forecast data and the pre-defined high and low uncertainty time domain results, re-optimizes the scheduling strategy for the following four hours. The real-time adjustment layer is deployed in the local controller, with a response period of 200 milliseconds, to perform power compensation for the energy storage converter and static synchronizing compensator.
[0028] Then, multi-objective weight adaptive adjustment is performed. An intraday rolling correction layer constructs a multi-objective optimization model with the objectives of minimizing operating costs, carbon emissions, and power fluctuations. When entering the high-uncertainty time domain, the weight coefficient of the power fluctuation minimization objective is dynamically increased from 0.2 to 0.6, while the sum of the weight coefficients of the two objectives, minimizing operating costs and carbon emissions, is decreased from 0.8 to 0.4, prioritizing the suppression of power surges. When entering the low-uncertainty time domain, the weight coefficient of the power fluctuation minimization objective is decreased to 0.1, while the sum of the weight coefficients of the two objectives, minimizing operating costs and carbon emissions, is increased to 0.9, fully exploring the potential for economic operation. A fuzzy PID controller is used during the weight adjustment process, with the rate of change of prediction error as input, dynamically correcting the weight coefficients to avoid system oscillations caused by sudden weight changes.
[0029] Next, the energy storage charge / discharge margin is corrected. A virtual charge / discharge margin M is introduced, calculated based on the current state of charge (SOC) of the energy storage system, the remaining charge / discharge cycle life (N_remaining), and the duration of high uncertainty in the time domain (T_high). The specific calculation formula is M = 0.4 × (SOC - SOC) min ) / (SOC max -SOC min ) + 0.3 × (N_remaining / N_rated) + 0.3 × T_high / T_base, where SOC min Take 0.2, SOC max The value is set to 0.9, with N (rated) provided by the energy storage manufacturer, and T (baseline) set to 2 hours. When M is below 0.3, the energy storage reserve is deemed insufficient. In the high uncertainty time domain, the upper limit of the energy storage charging and discharging power is compressed from 100% of the rated value to 60% of the rated value, and the lower limit of charging and discharging is increased from -100% of the rated value to -60%, reserving 40% of the adjustment reserve capacity. After entering the low uncertainty time domain, the rated charging and discharging capacity of the energy storage is gradually restored by releasing 10% of the compression ratio every 15 minutes.
[0030] Subsequently, differentiated command allocation is implemented. Micro gas turbines and interruptible loads are assigned to the basic regulation group, whose commands are updated every 15 minutes, responsible for the slow adjustment of steady-state power deviations. Energy storage systems and static synchronizing compensators are assigned to the fast regulation group, whose droop control loop operates continuously. When a system frequency deviation exceeds 0.1Hz or a grid connection point power deviation exceeds 5% of the rated power, the fast regulation group responds directly within 100 milliseconds, without waiting for optimization cycle commands. The droop coefficient of the fast regulation group is dynamically allocated based on the equipment's regulation accuracy; equipment with higher regulation accuracy is assigned a larger droop coefficient to undertake more regulation tasks.
[0031] Then, dynamically tighten the safety constraints. Establish a dynamic tightening mechanism: when the operating state is determined to be in the high uncertainty time domain, dynamically raise the lower limit of the energy storage state of charge from 0.2 to 0.3 and lower the upper limit from 0.9 to 0.8, reserving a buffer margin of 10% for the upper and lower limits of the power grid's exchange power. That is, the upper limit of the exchange power is reduced from 100% of the rated value to 90%, and the lower limit is raised from -100% of the rated value to -90%. When the operating state changes to the low uncertainty time domain, gradually restore the lower limit of the energy storage state of charge to 0.2 and the upper limit to 0.9, and restore the power grid's exchange power to the rated value.
[0032] Finally, a closed-loop verification and update is performed. After each daily rolling cycle, the actual output of the energy storage system, micro gas turbine, and power exchange of the main power grid are compared with the dispatch plan. When the deviation of any device exceeds 8% of the rated power, a feedback correction term is introduced. The deviation value is multiplied by a decay coefficient of 0.3 and then added to the prediction correction term of the multi-objective optimization model in the next cycle. The multi-objective weight adaptive adjustment is repeated until the safety constraint dynamic tightening stage, forming a closed-loop control system.
[0033] Example 2 provides another microgrid multi-objective scheduling optimization control method, which differs from Example 1 in that it has been adapted for islanded operation scenarios, specifically in the following aspects.
[0034] In the data acquisition and time-domain partitioning steps, when a disconnection between the microgrid and the main grid is detected, the system automatically switches to islanded operation mode. The data acquisition cycle is reduced from 3 minutes to 1 minute, and periods with prediction errors exceeding 15% are designated as high-uncertainty time domains to identify risky periods under islanded operation earlier. Simultaneously, in islanded mode, the safe operating range of the energy storage state of charge is dynamically adjusted to 0.3 to 0.85, reserving sufficient adjustment margin to cope with operating conditions without main grid support within the island.
[0035] In the construction steps of the hierarchical optimization framework, the time scale of the day-ahead planning layer in islanded mode remains 24 hours, but the power exchange capacity of the large power grid is removed as a decision variable and replaced by strict power self-balancing constraints; the update cycle of the intraday rolling correction layer is shortened from 1 hour to 30 minutes to adapt to faster state changes under islanded operation; the real-time adjustment layer still maintains a 200-millisecond response cycle, but the droop control dead zone of the energy storage converter is narrowed from 5% of the rated power to 2%, improving the frequency support capability under islanded operation.
[0036] In the multi-objective weight adaptive adjustment step, the baseline weight of the power fluctuation minimization objective is increased from 0.2 in the grid-connected mode to 0.4 in the islanded mode, and further dynamically increased to 0.7 in the high uncertainty time domain to prioritize the stability of islanded operation. At the same time, frequency deviation is introduced as an additional optimization objective. When the system frequency deviates from 50Hz by more than 0.2Hz, the weight of the frequency deviation objective is increased to 0.5, temporarily suppressing the weights of other objectives to prioritize the restoration of frequency stability.
[0037] In the energy storage charge / discharge margin correction step, the calculation formula for the virtual charge / discharge margin in islanded mode is adjusted to M=0.5×(SOC-SOC) min ) / (SOC max -SOC min The weighting factor of the SOC term is increased by 0.4 × (N_remaining / N_rated) + 0.1 × T_high / T_baseline to strengthen the impact of state of charge on the safety of islanded operation. When M is below 0.4, the energy storage reserve is considered severely insufficient. In the high uncertainty time domain, the upper limit of energy storage charging and discharging power is compressed to 50% of the rated value, the lower limit of charging and discharging is raised to -50%, and 50% of the regulation reserve capacity is reserved.
[0038] In the differentiated instruction allocation step, in islanded mode, interruptible loads are adjusted from the basic adjustment group to the fast response reserve group. When the frequency drop rate is detected to exceed 0.5Hz / second, the interruptible load is automatically disconnected within 200 milliseconds without waiting for upper-level scheduling instructions, so as to quickly prevent frequency collapse.
[0039] In the dynamic tightening step of safety constraints, a dynamic tightening mechanism for frequency safety constraints is established in islanded mode. When the system frequency is below 49.5Hz, the lower limit of the energy storage state of charge is dynamically increased from 0.3 to 0.4, and energy storage discharge is prohibited. At the same time, the threshold for cutting off interruptible loads is lowered to 80% of the rated power to release the load-side regulation capacity in advance. When the frequency recovers to above 49.8Hz and lasts for 5 minutes, the constraints are gradually restored to the normal value.
[0040] The closed-loop verification and update steps are consistent with those in Example 1, but the attenuation coefficient of the feedback correction term is increased from 0.3 to 0.5 to enhance the response strength to deviations in islanded mode and accelerate the closed-loop convergence process.
[0041] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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 multi-objective scheduling and optimization control method for microgrids, characterized in that, The method includes the following steps: Data acquisition and time domain division: Collect core operation data of the microgrid according to a set period; combine the data error distribution to divide the future scheduling period into high and low uncertainty time domains, corresponding to the time periods when the prediction error exceeds the first threshold and is lower than the second threshold, respectively; Layered optimization framework construction: A three-layer framework is constructed, consisting of a daily planning layer, an intraday rolling correction layer, and a real-time adjustment layer. The time scales are successively decreasing to realize millisecond-to-second power compensation for initial scheduling, dynamic correction, and rapid response equipment, respectively. Multi-objective weight adaptive adjustment: The intraday correction layer constructs a multi-objective model that minimizes cost, carbon emissions, and power fluctuations, and adjusts the weights according to the time domain type. In the high-uncertainty time domain, the power fluctuation weight is increased to suppress shocks; in the low-uncertainty time domain, the cost and carbon emission weights are increased. Energy storage charge / discharge margin correction: Introduce virtual charge / discharge margin, which is calculated based on energy storage status, remaining lifetime and high uncertainty time domain duration; use the high uncertainty time domain to compress the upper and lower limits of charge and discharge for reserve, and release the margin through the low uncertainty time domain to restore rated capacity; Differentiated instruction allocation: The equipment is divided into a basic group and a fast group according to the response speed and accuracy. When the instantaneous deviation of the steady-state deviation exceeds the third threshold in the basic group, the droop control of the fast group responds directly and is not constrained by the optimization cycle. Dynamic tightening of safety constraints: Establish a dynamic tightening mechanism to adjust constraints according to operating status and time domain type; reduce the state of charge of energy storage through the high uncertainty time domain, and use the upper and lower limits of the power exchange of the large power grid to reserve margins, and restore the rated value when using the low uncertainty time domain; Closed-loop verification and update: After the end of each daily rolling cycle, the actual output is compared with the scheduling plan. When the deviation exceeds the fourth threshold, a feedback correction term is introduced. The multi-objective weight adaptive adjustment stage is repeated until the safety constraint dynamic tightening stage, forming a closed-loop control system.
2. The microgrid multi-objective scheduling optimization control method as described in claim 1, characterized in that: In the data acquisition and time-domain partitioning steps, the core operating data include photovoltaic power generation output, wind power generation output, load demand, energy storage status of charge, real-time electricity price, and the exchange power between the microgrid and the main grid; the set periodic acquisition time is 1-5 minutes, the first threshold setting range is 15% to 25%, and the second threshold setting range is 5% to 10%.
3. The microgrid multi-objective scheduling optimization control method as described in claim 2, characterized in that: The first threshold is that when the error between the predicted output and the actual output of renewable energy exceeds 15% to 25%, the corresponding time period is divided into a high uncertainty time domain, and power fluctuations need to be suppressed first; when the error between the predicted output and the actual output of renewable energy is less than 5% to 10%, the corresponding time period is divided into a low uncertainty time domain; when dividing the time domain, the predicted and actual output data of at least 3 months of history are collected, the error is statistically analyzed and the distribution curve is fitted, and the high and low uncertainty time domains are divided accordingly.
4. The microgrid multi-objective scheduling optimization control method as described in claim 1, characterized in that: In the hierarchical optimization framework construction steps, the day-ahead planning layer adopts a 24-hour time scale and generates an initial scheduling plan based on the previous day's forecast data, including equipment output, energy storage charging and discharging, and power grid exchange power. The intraday rolling correction layer uses an hourly update timescale, and updates the strategy based on actual data and time domain results every cycle; the real-time adjustment layer uses a second-level timescale to perform power compensation for fast-response devices.
5. The microgrid multi-objective scheduling optimization control method as described in claim 1, characterized in that: In the multi-objective weight adaptive adjustment step, the weight of the power fluctuation minimization objective is increased first in the time domain with high uncertainty, while the sum of the weights of the cost and carbon emission minimization objectives is increased first in the time domain with low uncertainty. A fuzzy PID algorithm is used, combined with prediction error to dynamically correct the weights, to avoid system fluctuations.
6. The microgrid multi-objective scheduling optimization control method as described in claim 1, characterized in that: In the energy storage charge and discharge margin correction step, the virtual charge and discharge margin is calculated using a preset formula based on energy storage-related parameters and weighting coefficients; in the high uncertainty time domain, the upper and lower limits of energy storage charge and discharge are reserved for backup, and in the low uncertainty time domain, the margin is released at fixed intervals to restore the rated capacity of energy storage.
7. The microgrid multi-objective scheduling optimization control method as described in claim 6, characterized in that: The formula for calculating the virtual charge / discharge margin is M=k1×(SOC-SOC). min ) / (SOC max -SOC min ) + k2 × (N_remaining / N_rated) + k3 × T_height; where M represents the virtual charge / discharge margin; k1, k2, and k3 are all weighting coefficients, and they satisfy k1+k2+k3=1; SOC is the current state of charge of the energy storage system. SOC min State of Charge (SOC) is the lower limit of the rated state of charge of the energy storage system. max Nremaining is the upper limit of the rated state of charge of the energy storage system; Nrated is the remaining charge-discharge cycle life of the energy storage system; Nrated is the rated charge-discharge cycle life of the energy storage system; and Thigh is the duration of the high uncertainty time domain.
8. The microgrid multi-objective scheduling optimization control method as described in claim 1, characterized in that: In the differentiated instruction allocation process, the basic adjustment group is responsible for steady-state power deviation, while the fast adjustment group responds more quickly; the third threshold is the standard for judging rated power deviation, and the droop coefficient is adjusted according to the equipment adjustment accuracy.