A parallel candidate prediction-based transformer energy storage task conflict coordination control method and system

Abstract

The application discloses a kind of parallel candidate prediction-based district energy storage task conflict coordination control method and system, belong to power distribution automation and energy storage coordination control technical field.The method defines four types of mutually exclusive operation modes of local governance, frequency modulation, peak shaving, idle, collects grid-connected point voltage and carries out two-level safety threshold discrimination;Local governance demand is decomposed in charge and discharge direction conflict, and charging and discharging candidate scheme is constructed;In the rolling prediction time domain, the state of charge and the district operation state of four types of operation modes are recursively predicted in parallel, and a weighted comprehensive cost function containing voltage deviation, three-phase imbalance and other penalty terms is constructed;After feasible domain constraint verification, the scheme with smaller cost is selected as the candidate operation mode, combined with safety out-of-limit emergency switching and mode switching hysteresis mechanism to determine the effective mode, and the PCS generates phase-specific active and reactive power instructions.The application can realize mode coordination optimization in the scene of multiple tasks of charging and discharging of district energy storage.

Description

Technical Field

[0001] This invention belongs to the field of distribution network operation control and energy storage system coordination control, specifically a method and system for coordinating control of energy storage task conflicts in distribution areas based on parallel candidate prediction. Background Technology

[0002] With the large-scale integration of distributed power sources, electric vehicles, and flexible loads, problems such as voltage exceeding limits, three-phase imbalance, and reverse overload at the distribution network end are becoming increasingly prominent. Configuring energy storage systems at the transformer substation side, through charging and discharging regulation, participates in local voltage management. Simultaneously, it can serve as a virtual power plant (VPP) resource to participate in grid frequency regulation, peak shaving, and other ancillary services, becoming an important technical means to improve the power supply quality and asset utilization rate of transformer substations.

[0003] In actual operation, energy storage in power distribution areas often faces multiple tasks simultaneously: local governance tasks require real-time adjustment of charging and discharging power based on indicators such as grid connection voltage, three-phase imbalance, and reverse load rate; grid service tasks require tracking power setpoints according to frequency regulation or peak shaving instructions issued by virtual power plants. These tasks often conflict in their requirements for the charging and discharging direction and power level of energy storage—for example, high-voltage governance requires energy storage to charge to absorb excess photovoltaic power, while peak shaving tasks may require energy storage to discharge to support evening peak loads.

[0004] In existing technologies, distributed energy storage systems often employ a fixed-priority arbitration method to handle multi-task conflicts. This involves pre-setting the priority order of tasks such as local governance, frequency regulation, and peak shaving, with higher-priority tasks directly preempting energy storage resources. The drawbacks of this approach are: firstly, fixed priorities cannot account for the overall system optimization under different operating conditions, and frequent execution of high-priority tasks may lead to deviations in the energy storage's state of charge (SOC) from a reasonable range and accelerated battery degradation; secondly, when there are conflicts in charging and discharging directions within local governance needs (such as the simultaneous occurrence of high-voltage charging and three-phase imbalance requiring discharging), there is a lack of a systematic mechanism for comparing and selecting candidate solutions; thirdly, the mode switching process lacks hysteresis and suspension mechanisms, easily leading to frequent mode switching and interruptions and irrecoverable disruptions in virtual power plant command execution. The control results are entirely determined by the preset priority order, unable to adaptively adjust based on real-time information such as grid connection voltage levels, frequency deviations, remaining peak shaving time, uncontrolled electricity volume, and energy storage state of charge, making it difficult to achieve a dynamic balance between voltage safety, virtual power plant business completion rate, and energy storage operation economy. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for coordinating and controlling conflicts in energy storage tasks in distribution transformer areas based on parallel candidate prediction. By constructing a four-mode mutually exclusive operation model and a two-level voltage safety threshold discrimination mechanism, a forced switch to local governance is implemented under safety limit exceedance events. Within the coordinable operation domain, candidate schemes are generated in parallel for four modes: local governance, frequency regulation, peak regulation, and idle. The local governance requirements for charging and discharging direction conflicts are decomposed into two sub-candidates: charging and discharging. The energy storage state of charge and distribution transformer operation status are recursively deduced in the rolling prediction time domain. The comprehensive prediction cost function containing multiple penalties is used for evaluation. Combined with feasible domain constraint verification, the optimal operation mode is selected. Finally, a PCS phase-by-phase active and reactive power coordination command is generated, realizing the dynamic balance and optimal coordination between voltage safety, power quality management, and grid ancillary services.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0007] A method for coordinating and controlling conflicts in energy storage tasks in transformer substations based on parallel candidate prediction includes:

[0008] 101: Energy Storage Grid Connection Point Status Acquisition and Four-Mode Mutually Exclusive Operation Model Construction: Define four types of operation modes: local governance, frequency regulation, peak shaving, and idle; construct a four-mode mutually exclusive operation model that allows only one mode to be effective at any given time; collect the three-phase voltage and operation status of the energy storage grid connection point in real time; construct a two-level voltage safety threshold discrimination model; divide the control into two branches: safety over-limit events and coordinateable operation domain; construct a dynamic model of energy storage charge state; and update the energy storage charge state according to the control cycle.

[0009] 102: Identifying and Decomposing Local Governance Needs from Multiple Dimensions

[0010] Calculate the high voltage, low voltage, three-phase imbalance, and reverse load rate respectively to determine four types of local governance needs; if there is no conflict between the charging and discharging directions of the local governance needs, determine the local governance request candidate schemes based on the triggering needs; if there is a conflict between the charging and discharging directions, construct the discharge electron candidate scheme and the charging electron candidate scheme respectively, and output the three-phase target active power and three-phase target reactive power corresponding to the two sub-candidate schemes.

[0011] 103: Comprehensive Cost Calculation for Four-Mode Parallel Candidate Rolling Prediction

[0012] A comprehensive prediction cost function containing multiple penalty terms is constructed; the active power benchmarks for request candidate schemes under four operating modes—local governance, frequency regulation, peak shaving, and idle—are determined respectively; when the local governance mode includes both discharge and charging candidate schemes, the energy storage state of charge and the operating state of the distribution area are recursively predicted within a set rolling prediction time domain, the comprehensive prediction cost of the two sub-candidate schemes is calculated, and the one with the smaller comprehensive prediction cost is selected as the local governance request candidate scheme; further, rolling prediction is performed on the request candidate schemes under the four operating modes, and the comprehensive prediction cost of each request candidate scheme is calculated.

[0013] 104: Feasibility Region Constraint Verification and Optimal Operation Mode Selection

[0014] Each candidate solution is checked for feasible domain constraints, and infeasible solutions that do not meet the constraints are eliminated; among the remaining feasible candidate solutions, the solution with the minimum comprehensive prediction cost is selected as the candidate running mode.

[0015] 105: Safety Limit Exceeding Emergency Response and Operation Mode Switching:

[0016] When a safety limit violation event is triggered, the current operating mode is forcibly switched and local governance is executed in an emergency; interrupted frequency regulation or peak regulation commands of virtual power plants are suspended and saved; in the coordinateable operating domain, rolling prediction of requested candidate schemes under four types of operating modes and selection of requested candidate schemes are performed to determine the candidate operating mode;

[0017] 106: Generation of PCS phase-specific active and reactive power coordination commands

[0018] Based on the active power benchmark corresponding to the effective candidate operating mode, the PCS phase-by-phase reactive power compensation calculation is performed, and finally, the three-phase active and reactive power output commands of the energy storage grid connection point are generated.

[0019] Furthermore, the four-mode mutually exclusive operation model is constructed as follows:

[0020]

[0021] In the formula, , , , These are status flags for local governance, frequency regulation, peak shaving, and idle mode, respectively. They are all variables of 0 or 1. A value of 1 indicates that the corresponding operating mode is in effect, and a value of 0 indicates that the corresponding operating mode is not in effect.

[0022] Furthermore, the aforementioned dynamic model of the energy storage state of charge is as follows:

[0023]

[0024] For the first Control the state of charge of the energy storage cycle; For the first Active power of the energy storage grid connection point during the cycle, as agreed Indicates discharge. Indicates charging; To control the cycle; This refers to the rated capacity of the energy storage.

[0025] Furthermore, the charging / discharging direction conflict refers to the inconsistency in the charging / discharging power directions required by at least two types of local governance needs.

[0026] Furthermore, the formula for calculating the target power benchmark for local charging governance needs is as follows:

[0027]

[0028] In the formula, , These are the power requirements for high voltage regulation and reverse load rate regulation, respectively.

[0029] Furthermore, the multiple penalty items include voltage deviation, three-phase imbalance, reverse overload rate, frequency modulation tracking, peak shaving tracking, SOC deviation, battery degradation, and mode switching penalty items. The comprehensive predicted cost is obtained by weighting and summing each penalty item according to a preset weight.

[0030] Furthermore, the formula for calculating the voltage deviation penalty term is as follows:

[0031]

[0032] In the formula, , These are the minimum and maximum voltages at the grid connection point that exceed the low and high voltage thresholds of 209V and 231V respectively. The values ​​are normalized to the range of 0-1 according to the 209V-198V and 231V-235V ranges. The larger the value, the more severe the voltage deviation. , The weighting coefficients are for low and high voltage deviations.

[0033] The formula for calculating the three-phase imbalance penalty term is as follows:

[0034]

[0035] In the formula, The unbalance threshold; Take the three-phase imbalance threshold Up to the maximum severity difference; This represents the current three-phase imbalance.

[0036] The formula for calculating the reverse load rate penalty is as follows:

[0037]

[0038] In the formula, The reverse load rate threshold; Take the reverse load rate threshold Up to the maximum severity difference; This represents the current reverse load rate.

[0039] The formula for calculating the frequency modulation tracking penalty term is as follows:

[0040]

[0041] In the formula, Target power for frequency modulation; This represents the upper limit of the active power of energy storage.

[0042] The formula for calculating the peak-shaving tracking penalty term is as follows:

[0043]

[0044] In the formula, The target power for peak shaving; The current active power of energy storage; As the upper limit of the active power of energy storage, the formula for calculating the SOC deviation penalty term is:

[0045]

[0046] In the formula, This represents the current state of charge of the energy storage system. The reference state of charge is as of the day-ahead. To determine the normalized bandwidth of the charged state, take and The difference is used to relative SOC The deviation is normalized and included Penalties;

[0047] The formula for calculating the battery degradation penalty is as follows:

[0048]

[0049] In the formula: This represents the absolute value of the active power of energy storage. This represents the upper limit of the active power of energy storage.

[0050] The formula for calculating the mode switching penalty is as follows:

[0051] In the formula These are candidate patterns currently being evaluated. This refers to the currently running mode.

[0052] Furthermore, the feasible domain constraint verification includes three types of feasible domain constraint verification: state of charge operating range, charging and discharging power, and power ramp rate. Specifically, it includes: verifying whether the energy storage state of charge in each control cycle within the prediction time domain is within the preset state of charge operating range; verifying whether the charging and discharging power in each control cycle is within the preset charging and discharging power limit; and verifying whether the power change between adjacent control cycles does not exceed the preset power ramp rate limit.

[0053] Furthermore, within the coordinateable operating domain, a hysteresis verification mechanism for switching operating modes is constructed: switching to the candidate operating mode is allowed only when the difference between the comprehensive prediction cost of the current operating mode and the candidate operating mode exceeds a set threshold; when the difference is not greater than the set threshold, the current operating mode remains unchanged.

[0054] A conflict coordination and control system for energy storage tasks in a distribution area based on parallel candidate prediction, comprising:

[0055] The four-mode mutually exclusive operation model construction module is used to define four types of operation modes: local governance, frequency regulation, peak regulation, and idle, and to construct a four-mode mutually exclusive operation model that allows only one mode to be effective at any given time; the grid connection point status acquisition and voltage discrimination module is used to acquire the three-phase phase voltage and operation status of the energy storage grid connection point in real time, and divides the control branches into two parts: safety over-limit events and coordinateable operation domain based on a two-level voltage safety threshold discrimination model, and updates the energy storage charge state according to the control cycle;

[0056] The local governance demand determination and conflict decomposition module is used to calculate high-voltage governance demand, low-voltage governance demand, three-phase imbalance governance demand, and reverse load rate governance demand respectively to determine four types of local governance demands. When there is no charging / discharging direction conflict among the four types of local governance demands, the target power is directly determined based on the triggered governance demand. When there is a charging / discharging direction conflict among the four types of local governance demands, candidate schemes for discharging electrons and candidate schemes for charging electrons are constructed respectively, and compared by substituting them into the comprehensive prediction cost function. The candidate scheme with the smaller comprehensive prediction cost is selected as the local governance request candidate scheme, and the three-phase target active power and three-phase target reactive power are output.

[0057] The parallel candidate rolling prediction cost calculation module is used to determine the active power benchmark of the request candidate schemes under the four types of operation modes: local governance, frequency regulation, peak regulation, and idle. Within the set rolling prediction time domain, the module performs recursive prediction of the energy storage state of charge and the operating state of the distribution area for each request candidate scheme on a cycle basis. Based on the comprehensive prediction cost function containing multiple penalty terms, the module calculates the comprehensive prediction cost of each request candidate scheme within the rolling prediction time domain.

[0058] The feasible domain constraint verification and mode selection module is used to verify the feasible domain constraints of each request candidate solution, eliminate infeasible solutions that do not meet the constraints, and select the solution with the minimum comprehensive prediction cost from the remaining feasible request candidate solutions as the candidate running mode.

[0059] The safety limit violation emergency and mode switching module is used to forcibly switch the current operating mode to the local governance mode to perform emergency governance when a safety limit violation event is triggered, and to suspend and save the interrupted virtual power plant frequency regulation or peak regulation commands; under the coordinateable operating domain, it performs rolling prediction and selection of requested candidate schemes under four types of operating modes to determine the candidate operating mode;

[0060] The PCS phase-by-phase instruction generation module is used to perform PCS phase-by-phase reactive power compensation calculation based on the active power benchmark corresponding to the effective candidate operating mode, and generate three-phase active power output instructions and three-phase reactive power output instructions for the energy storage grid connection point.

[0061] This invention first defines four operating modes—local governance, frequency regulation, peak shaving, and idle—at the logic layer, and constructs a four-mode mutually exclusive operating model. This model constrains that only one mode is allowed to be active at any given control moment, fundamentally eliminating power command conflicts caused by parallel issuance of multiple modes. Based on this, the three-phase voltage and operating status of the energy storage grid connection point are collected in real time, and a two-level voltage safety threshold discrimination model is constructed: when the grid connection point voltage exceeds the safety boundary, it enters the "safety limit exceedance event" branch; when it does not exceed the limit, it enters the "coordinateable operating domain" branch, thus separating the "safety issues that must be addressed immediately" from the "scheduling issues that can be optimized among multiple objectives." Simultaneously, the dynamic model of the energy storage state of charge is updated recursively according to the control cycle, providing initial state values ​​and a recursive baseline for subsequent rolling predictions.

[0062] This invention calculates four types of local governance demands within the local governance channel: high voltage, low voltage, three-phase imbalance, and reverse load rate. When there is no conflict in charging / discharging direction among these demands, the target active and reactive power are directly determined by the triggering demand. When different governance demands contradict each other in charging / discharging direction (e.g., simultaneous charging and discharging), instead of a simple compromise, candidate schemes for discharging electrons and charging electrons are constructed separately. Both are then substituted into the comprehensive prediction cost function, and the one with the smaller prediction cost is selected as the candidate scheme for the local governance request. The corresponding three-phase target active power and three-phase target reactive power are output, thereby transforming the directional conflict into two comparable and preferred candidates.

[0063] Under the coordinateable operation domain branch, this invention performs parallel candidate prediction for four types of operation modes: First, the active power benchmarks for requested candidate schemes are determined for local governance, frequency regulation, peak shaving, and idle modes respectively; then, within the set rolling prediction time domain, the energy storage state of charge and distribution area operation status are recursively calculated for each requested candidate scheme on a cycle-by-cycle basis to obtain the state trajectory of each requested candidate scheme in the future time domain. Based on this, a comprehensive prediction cost function is constructed, whose cost terms cover voltage deviation, three-phase imbalance, reverse overload rate, frequency regulation tracking deviation, peak shaving tracking deviation, SOC deviation, battery degradation, and mode switching penalty. This unifies electrical safety, grid tracking, energy storage lifetime, and switching costs into comparable scalar costs, and calculates the comprehensive prediction cost of each requested candidate scheme in the rolling prediction time domain accordingly.

[0064] This invention requests that candidate solutions undergo feasible region constraint verification before being selected as the best option. This involves sequentially verifying three types of hard constraints: the state-of-charge (SOC) operating range, the upper limit of charge / discharge power, and the power ramp-up rate. Any infeasible solution that fails to meet any of these constraints is eliminated. From the remaining feasible candidate solutions, the one with the lowest overall prediction cost is selected as the candidate operating mode, ensuring that the final decision simultaneously satisfies the energy storage safety operation boundary and multi-objective optimization preferences.

[0065] This invention establishes two types of control branch exit arbitration logic. When the two-level voltage threshold discrimination model determines that a safety over-limit event has occurred, the safety priority is the highest: the current operating mode is forcibly switched, the system immediately switches to local governance, and the PCS performs emergency control with the goal of restoring the grid connection point voltage; at the same time, the virtual power plant frequency regulation or peak regulation commands interrupted by this forced switch are not discarded, but suspended and saved, and can enter the recovery arbitration process when the over-limit is lifted and the system is restored to the coordinateable operating domain, to decide whether and how to continue the original grid commands, so as to avoid grid service interruption caused by emergency switch.

[0066] Within the coordinateable operating domain, the system obtains candidate operating modes based on rolling predictions and optimization results, but does not switch immediately: a mode switching hysteresis verification mechanism is constructed, allowing switching to the candidate operating mode only when the difference between the comprehensive prediction cost of the current operating mode and the candidate operating mode exceeds a set threshold; if the cost difference is insufficient to cover the disturbance and battery degradation cost introduced by the switch, the current mode is maintained. This mechanism is used to suppress frequent mode jitter caused by small fluctuations in voltage and load.

[0067] This invention takes the corresponding active power benchmark as the active power given for the candidate operating mode that finally takes effect, and performs PCS phase-by-phase reactive power compensation calculation on this basis. It generates three-phase active power output command and three-phase reactive power output command for the energy storage grid connection point according to the three phases, so that the phase-by-phase active and reactive power can respond in a coordinated manner to the needs of local governance and grid dispatch.

[0068] Compared with the prior art, the present invention, through the above technical solution, can achieve the following beneficial technical effects:

[0069] First, by constructing a mutually exclusive operation model with four operating modes—local governance, frequency regulation, peak regulation, and idle—only one mode is effective at any given time. This avoids conflicts between active and reactive power commands caused by the parallel issuance of multiple modes at the control logic level, ensuring the uniqueness and consistency of the power given at the energy storage grid connection point.

[0070] Second, by using a two-level voltage safety threshold discrimination model, the operating status of the grid connection point is divided into two control branches: safety over-limit events and coordinateable operating domains. This enables hierarchical control that prioritizes the handling of safety issues and coordinates optimization issues, ensuring the safety of equipment and distribution areas when voltage exceeds limits, and then performing multi-objective economic scheduling when the voltage is normal.

[0071] Third, by performing parallel calculations on four types of local governance needs—high voltage, low voltage, three-phase imbalance, and reverse load rate—and constructing candidate schemes for discharging and charging electrons respectively when there is a conflict in the charging and discharging directions, and substituting them into the comprehensive prediction cost function for selection, the directional conflicts within local governance can be transformed into quantifiable and comparable candidate choices. This avoids simple compromises that lead to mutual cancellation of governance capabilities and is conducive to improving the overall satisfaction of local governance needs.

[0072] Fourth, by recursively calculating the state of charge and the operating status of the distribution area for the four types of operating modes in the rolling prediction time domain, and constructing a comprehensive prediction cost function covering voltage deviation, three-phase imbalance, reverse overload rate, frequency regulation tracking, peak regulation tracking, SOC deviation, battery degradation and mode switching penalty, electrical safety, grid tracking performance, energy storage life and switching cost can be uniformly incorporated into the same decision framework, making mode selection forward-looking and globally optimized, rather than reacting only to the instantaneous state at the current moment.

[0073] Fifth, by first eliminating infeasible solutions through three types of feasible domain constraints—charge state operation range, charging and discharging power, and power ramp rate—and then selecting the solution with the lowest comprehensive predicted cost from the remaining feasible solutions, we can ensure the safe operation boundary of the energy storage body with hard constraints and achieve multi-objective coordination with soft costs, taking into account both safety and optimization.

[0074] Sixth, by forcibly switching to local governance when a safety over-limit event is triggered, and suspending and saving the interrupted virtual power plant frequency regulation or peak shaving commands, it is possible to quickly respond to grid connection point voltage safety issues in emergency situations. At the same time, it avoids the loss of grid dispatch commands due to emergency switching, which is conducive to continuing the original frequency regulation and peak shaving services after the over-limit is cleared, and ensures the continuity of grid-side tasks.

[0075] Seventh, by constructing a mode switching hysteresis verification mechanism, switching is only allowed when the difference between the comprehensive predicted cost of the current operating mode and the candidate operating mode exceeds a set threshold. This can suppress frequent fluctuations in operating mode caused by small fluctuations in grid connection voltage and transformer load, and reduce power disturbances and battery cycle losses introduced by mode switching.

[0076] Eighth, by performing PCS phase-by-phase reactive power compensation calculations based on the active power benchmark of the effective mode and generating active and reactive power output commands for each of the three phases, it is possible to achieve phase-by-phase level active and reactive power coordinated control, which is beneficial for simultaneously responding to three-phase voltage deviation and three-phase imbalance problems and improving the precision of power quality management at the grid connection point. Attached Figure Description

[0077] Figure 1 This is a schematic diagram comparing the grid connection point Umin / Umax of the present invention with the voltage safety threshold of the fixed priority strategy.

[0078] Figure 2 This is a schematic diagram illustrating the mutual exclusion operation results of the four modes of the present invention and the fixed priority strategy;

[0079] Figure 3 This is a schematic diagram comparing the technical effects of the present invention and a fixed priority strategy;

[0080] Figure 4 This is a flowchart of the method of the present invention. Detailed Implementation

[0081] Example 1:

[0082] A method for coordinating and controlling conflicts in energy storage tasks in transformer substations based on parallel candidate prediction includes:

[0083] 101: Energy Storage Grid Connection Point Status Acquisition and Four-Mode Mutually Exclusive Operation Model Construction: Define four types of operation modes: local governance, frequency regulation, peak shaving, and idle; construct a four-mode mutually exclusive operation model that allows only one mode to be effective at any given time; collect the three-phase voltage and operation status of the energy storage grid connection point in real time; construct a two-level voltage safety threshold discrimination model; divide the control into two branches: safety over-limit events and coordinateable operation domain; construct a dynamic model of energy storage charge state; and update the energy storage charge state according to the control cycle.

[0084] Furthermore, in one embodiment of the present invention, 101 specifically includes:

[0085] 1.1 Energy Storage Grid Connection Point Status Acquisition and Four-Mode Mutually Exclusive Operation Model Construction

[0086] Construct a four-mode mutually exclusive operation model, defining 0-1 state variables corresponding to four operation modes: local governance, frequency regulation, peak shaving, and idle. Only one mode is allowed to be active at any given time.

[0087] (1)

[0088] In the formula, , , , These are status flags for local governance, frequency regulation, peak shaving, and idle mode, respectively. They are all variables of 0 or 1. A value of 1 indicates that the corresponding operating mode is in effect, and a value of 0 indicates that the corresponding operating mode is not in effect.

[0089] Real-time acquisition of three-phase voltage and operating status at the energy storage grid connection point, and real-time calculation of the extreme values ​​of the three-phase voltage at the grid connection point, serving as a quantitative basis for subsequent voltage safety assessment and mitigation needs detection:

[0090] (2)

[0091] (3)

[0092] In the formula, , , The three-phase voltage at the energy storage grid connection point; , These are the minimum and maximum values ​​of the three-phase phase voltage at the energy storage grid connection point, respectively.

[0093] 1.2 Construct a two-level voltage safety threshold discrimination model, dividing it into two control branches: safety over-limit events and the coordinateable operating domain.

[0094] The criteria for determining safety violation events are:

[0095] (4)

[0096] The criteria for a secure and reconcilable operational domain are:

[0097] (5)

[0098] In equations (4) and (5), 198V and 235V are the lower and upper safety voltage thresholds, respectively.

[0099] Equation (4) is the voltage safety limit criterion, used to identify safety limit events that require immediate emergency local management; if Equation (4) is true, the safety limit flag is set, and emergency local management is directly triggered, skipping the request candidate solution optimization process in the four types of operation modes.

[0100] Equation (5) is the voltage safety and coordinateable operation domain criterion, used to determine whether it is allowed to enter the rolling prediction of request candidate schemes and the selection of candidate operation modes in the subsequent four types of operation modes. If Equation (5) is true, the request candidate scheme optimization process in the four types of operation modes will be entered.

[0101] The safety voltage threshold of 198V-235V is distinguished from the ordinary voltage management threshold of 209V-231V: the former is used for the hierarchical switching of safety over-limit events and the coordinateable operating domain, and the latter is used for the detection of ordinary high and low voltage local management needs as shown in equations (7)-(8).

[0102] 1.3. Construct a dynamic model of the energy storage state of charge (SOC) using equation (6), and update the SOC according to the control cycle to provide a state benchmark for feasible domain constraint verification and comprehensive prediction cost calculation:

[0103] (6)

[0104] In equation (6), For the first Control the state of charge of the energy storage cycle; For the first Active power of the energy storage grid connection point during the cycle, as agreed Indicates discharge. Indicates charging; To control the cycle; This refers to the rated capacity of the energy storage.

[0105] 102: Identifying and Decomposing Local Governance Needs from Multiple Dimensions

[0106] Calculate the high voltage, low voltage, three-phase imbalance, and reverse load rate respectively to determine four types of local governance needs; if there is no conflict between the charging and discharging directions of the local governance needs, determine the local governance request candidate schemes based on the triggering needs; if there is a conflict between the charging and discharging directions, construct the discharge electron candidate scheme and the charging electron candidate scheme respectively, and output the three-phase target active power and three-phase target reactive power corresponding to the two sub-candidate schemes.

[0107] Furthermore, the charging / discharging direction conflict refers to the inconsistency in the charging / discharging power directions required by at least two types of local governance needs.

[0108] Furthermore, in one embodiment of the present invention, 102 specifically includes:

[0109] 2.1 Identify four types of local governance needs

[0110] The high voltage, low voltage, three-phase imbalance, and reverse load rates are calculated using equations (7) to (10) respectively, to determine the four types of local governance requirements:

[0111] High voltage mitigation triggering criteria:

[0112] (7)

[0113] Governance needs: i.e. Triggered at time, the required amplitude is recorded as The direction is for charging.

[0114] Low voltage mitigation trigger criteria:

[0115] (8)

[0116] Governance needs: i.e. Triggered at time, the required amplitude is recorded as The direction is discharge.

[0117] Three-phase unbalance calculation:

[0118] (9)

[0119] In the formula: This represents the maximum absolute deviation of the three-phase voltage amplitude from its average value. It is the arithmetic mean of the three-phase voltage amplitude of the bus.

[0120] Three-phase imbalance exceeding threshold management requirement: By using PCS to output reactive power in phases, positive reactive power is compensated for the phase with low voltage and negative reactive power is output for the phase with high voltage, thus correcting the three-phase voltage imbalance; the algebraic sum of the three phases is zero, and the direction of net active power is not changed.

[0121] Transformer reverse load rate calculation:

[0122] (10)

[0123] A transformer reverse load rate exceeding the threshold indicates a reverse overload, requiring mitigation measures: Over-limit trigger, reverse overload charging power demand amplitude This amplitude is from The degree of exceeding the limit is determined.

[0124] In equations (7) to (10), 231V and 209V are the threshold values ​​for ordinary high and low voltage management, respectively. This refers to the three-phase imbalance. This refers to the transformer's reverse load rate. This refers to the rated capacity of the transformer.

[0125] 2.2 Local governance conflict resolution and request candidate solution output:

[0126] Determine if there are charging / discharging direction conflicts among various local governance needs (simultaneous triggering of low-voltage discharge and high-voltage / reverse load rate charging). If there are no conflicts, directly determine the target power based on the triggering needs and output the power. ;

[0127] When multiple types of local governance needs conflict in charging and discharging directions, the target power benchmark for charging-related local governance needs is calculated according to equation (11):

[0128] (11)

[0129] In the formula, , These represent the power requirements for high-voltage charging and reverse overload charging, respectively.

[0130] Constructing candidate schemes for electron discharge (by (for the target) and candidate charging schemes (by (for the goal) , , The power demand is directly provided by the on-site proportional control based on the grid connection point measurements during this cycle, which falls under the existing voltage-power droop method. The calculation first determines whether the minimum voltage, maximum voltage, and reverse load rate exceed the acceptable threshold: if not, the corresponding demand is zero; if they have exceeded the threshold, the current measurement is considered to fall between the acceptable threshold and the hard limit. Demand is relatively low when just exceeding the threshold, and increases to the rated active power of energy storage when approaching the hard limit. From this, the low-voltage discharge demand power is obtained. High-voltage charging power requirements Reverse overload charging power requirement High voltage and reverse overload are both in the charging direction; the larger amplitude is taken and they are not added together. Phase-specific reactive power is still generated or absorbed locally according to the voltage limit of each phase. This cycle uses both the charging and discharging sub-schemes.

[0131] Electron emission candidate scheme The total active power is taken from the discharge power required for low-voltage management, and is initially distributed equally across the three phases; reactive power is still calculated using the phase reactive power for this cycle. This scheme only manages low voltage and does not perform charging required for high voltage or reverse overload. (Candidate charging scheme) The total active power is determined by the larger charging amplitude between high voltage and reverse overload (the larger one is used in the same direction, and they are not added together), and the three phases are initially evenly distributed; the reactive power and discharge schemes are the same. This scheme only addresses high voltage and / or reverse overload, and does not perform low voltage discharge.

[0132] Substitute the target power of the two sub-candidate schemes into the comprehensive prediction cost function constructed by equations (17) and (18), respectively, and calculate the comprehensive prediction cost of each within the set rolling prediction time domain, including voltage management effect, SOC change, equipment constraint feasibility, etc. Compare the comprehensive prediction costs of the discharge electron candidate scheme and the charging electron candidate scheme, and select the candidate scheme with the smaller comprehensive prediction cost as the local management request candidate scheme. The target active power corresponding to the request candidate scheme is taken as the final three-phase target active power. Output, to resolve conflicts in local governance needs.

[0133] Target power benchmark for charging-related governance needs It is the power setting value of the candidate charging scheme, which is used to calculate the comprehensive prediction cost of the scheme within the set rolling prediction time domain, and compare it with the candidate discharging scheme to finally decide whether the local governance is charging or discharging; taking the maximum value is to use a charging power to cover both high voltage and reverse overload rate governance needs at the same time.

[0134] After conflict resolution is completed, equation (12) outputs a unique local governance request candidate solution:

[0135] (12)

[0136] In the formula, The target active power for the three phases; The target reactive power for the three phases; Target active power of phases A / B / C; Target reactive power of phases A / B / C.

[0137] : Output structure of local governance request candidate solutions;

[0138] : The current set of triggered governance projects, which records the local governance requirement projects activated within this control cycle, including: high voltage governance, low voltage governance, three-phase imbalance governance, and reverse overload rate governance.

[0139] The conflict decomposition results show that when low-voltage discharge demand and high-voltage and / or reverse overload rate charging demand coexist, candidate discharge schemes are constructed respectively. (by (for the target) and candidate charging schemes (by (As the objective), the target active power corresponding to the candidate scheme for local governance requests is selected as... Output to resolve conflicts within local requirements.

[0140] This is a set of unresolved projects, recording governance needs that were not addressed in this cycle due to conflicting directions or insufficient power margins; these are not included in the current cycle. The output is left for re-detection and decision-making in the next control cycle.

[0141] This serves as a feasibility indicator for SOC and equipment constraints. A Boolean flag indicates whether the current local power management solution is feasible under physical constraints such as SOC upper and lower limits, charge / discharge power upper and lower limits, and ramp rate. =1 is feasible =0 means it is not feasible.

[0142] 103: Comprehensive Cost Calculation for Four-Mode Parallel Candidate Rolling Prediction

[0143] A comprehensive prediction cost function containing multiple penalty terms is constructed; the active power benchmarks for request candidate schemes under four operating modes—local governance, frequency regulation, peak shaving, and idle—are determined respectively; when the local governance mode includes both discharge and charging candidate schemes, the energy storage state of charge and the operating state of the distribution area are recursively predicted within a set rolling prediction time domain, the comprehensive prediction cost of the two sub-candidate schemes is calculated, and the one with the smaller comprehensive prediction cost is selected as the local governance request candidate scheme; further, rolling prediction is performed on the request candidate schemes under the four operating modes, and the comprehensive prediction cost of each request candidate scheme is calculated.

[0144] Furthermore, the multiple penalty items include voltage deviation, three-phase imbalance, reverse overload rate, frequency modulation tracking, peak shaving tracking, SOC deviation, battery degradation, and mode switching penalty items. The comprehensive predicted cost is obtained by weighting and summing each penalty item according to a preset weight.

[0145] Furthermore, in one embodiment of the present invention, 103 specifically includes:

[0146] 3.1 Equations (13)-(16) determine the active power baseline for the candidate schemes of the four types of operation modes:

[0147] (13)

[0148] (14)

[0149] (15)

[0150] (16)

[0151] The active power benchmark for local energy storage solutions. The active power benchmark for energy storage in frequency regulation schemes. The active power benchmark for peak-shaving schemes. The active power benchmark for idle energy storage schemes. It is the local governance request output structure The output local governance target is active power. , These are the target power for frequency regulation and peak shaving of the virtual power plant, respectively, while the target power for the idle scheme is zero.

[0152] 3.2 Recursive Prediction of Operating Status in the Rolling Prediction Time Domain

[0153] Let the current control period be The rolling prediction time domain is One cycle, cycle duration For each request candidate solution , Generate future power sequences ,in : No. Predict the steps in the following order, without skipping any:

[0154] (1) Substituting into the dynamic model of energy storage state of charge, the predicted state of charge is obtained. .

[0155] (2) Substitute this power into the three-phase power flow calculation of the transformer area to obtain the predicted phase voltage, and then calculate the predicted voltage extreme value. , The predicted three-phase imbalance is obtained by calculation. The predicted transformer reverse load rate is obtained through calculation. .

[0156] No. Predictive step frequency modulation target active power It is obtained by point-by-point synthesis of the AGC command prediction sequence and the primary frequency modulation droop power (calculated based on the frequency prediction sequence), and then power limiting.

[0157] No. Peak-shaving target active power in the prediction step The calculation is determined by the time-of-use pricing period (peak discharge / off-peak charging / flat period zero), or by net load peak shaving and valley filling calculation, or by dispatch plan interpolation, and then obtained through SOC energy verification and power limiting.

[0158] 3.3 Calculation of the comprehensive predicted cost of each candidate solution

[0159] Step 3-1: Construct a comprehensive prediction cost function using equations (17) and (18). Based on the recursive prediction of the energy storage charge state and the operating state of the distribution area for the four request candidate schemes in the rolling prediction time domain, calculate the comprehensive prediction cost of each request candidate scheme in the rolling prediction time domain.

[0160] (17)

[0161] (18)

[0162] In equations (17) and (18), This represents the total cost of the solution accumulated over the rolling time domain. For the first The overall predicted cost of candidate solutions for the same request; , , , This represents the comprehensive predicted cost of the four candidate solutions for local governance, frequency regulation, peak shaving, and idle access.

[0163] , , , , , , , These are voltage deviation, three-phase imbalance, reverse load rate, frequency modulation tracking, peak shaving tracking, SOC deviation, battery degradation, and mode switching penalty. to This represents the weighting coefficient for the corresponding penalty item.

[0164] 3.4 Each penalty item is quantified using the following formula. All penalty items are normalized to ensure that the dimensions are additive and comparable:

[0165] Voltage deviation penalty:

[0166] (19)

[0167] In the formula , These are the minimum and maximum voltages at the grid connection point that exceed the low and high voltage thresholds of 209V and 231V respectively. The values ​​are normalized to the range of 0-1 according to the 209V-198V and 231V-235V ranges. The larger the value, the more severe the voltage deviation. , The weighting coefficients are for low and high voltage deviations.

[0168] Three-phase imbalance penalty:

[0169] (20)

[0170] In the formula, The unbalance threshold; Take the three-phase imbalance threshold Up to the maximum severity difference; This represents the current three-phase imbalance.

[0171] Reverse load rate penalty:

[0172] (twenty one)

[0173] In the formula, The reverse load rate threshold; Take the reverse load rate threshold Up to the maximum severity difference; This represents the current reverse load rate.

[0174] Frequency modulation tracking penalty items:

[0175] (twenty two)

[0176] In the formula, Target power for frequency modulation; This represents the upper limit of the active power of energy storage. Current active power for energy storage

[0177] Peak shaving tracking penalty items:

[0178] (twenty three)

[0179] In the formula, The target power for peak shaving; The current active power of energy storage; This represents the upper limit of the active power of energy storage.

[0180] SOC deviation penalty item:

[0181] (twenty four)

[0182] In the formula, The reference state of charge is as of the day-ahead. To determine the normalized bandwidth of the charged state, take and The difference is used to relative SOC The deviation is normalized and included Penalties;

[0183] Battery degradation penalty:

[0184] (25)

[0185] In the formula: This represents the absolute value of the active power of energy storage. This is the upper limit of the active power of energy storage.

[0186] The formula for calculating the mode switching penalty is as follows: (26)

[0187] In the formula These are candidate patterns currently being evaluated. This refers to the currently running mode.

[0188] 104: Feasibility Region Constraint Verification and Optimal Operation Mode Selection

[0189] Each candidate solution is checked for feasible domain constraints, and infeasible solutions that do not meet the constraints are eliminated; among the remaining feasible candidate solutions, the solution with the minimum comprehensive prediction cost is selected as the candidate running mode.

[0190] Furthermore, the feasible domain constraint verification includes three types of feasible domain constraint verification: state of charge operating range, charging and discharging power, and power ramp rate. Specifically, it includes: verifying whether the energy storage state of charge in each control cycle within the prediction time domain is within the preset state of charge operating range; verifying whether the charging and discharging power in each control cycle is within the preset charging and discharging power limit; and verifying whether the power change between adjacent control cycles does not exceed the preset power ramp rate limit.

[0191] Furthermore, in one embodiment of the present invention, 104 specifically includes:

[0192] For each candidate solution, three types of feasible domain constraints are checked: energy storage state-of-charge operation range, charge / discharge power, and power ramp-up rate. Infeasible solutions that do not meet the constraints are eliminated.

[0193] Energy storage State of Charge (SOC) operating range constraints:

[0194] (27)

[0195] In the formula, , These are the lower and upper limits of the permissible state of charge, respectively.

[0196] Charge and discharge power constraints:

[0197] (28)

[0198] In the formula, , These are the maximum charging power and the maximum discharging power, respectively. Current active power for energy storage

[0199] Power ramp rate constraint:

[0200] (29)

[0201] In the formula, This represents the upper limit of the power ramp-up rate.

[0202] After filtering out infeasible solutions, the solution with the lowest overall prediction cost is selected from the remaining feasible candidate solutions as the candidate operating mode:

[0203] (30)

[0204] In the formula, This is the candidate operating mode with the lowest overall prediction cost; , , , The comprehensive predicted cost of the request candidate schemes under four operating modes: local governance, frequency regulation, peak regulation, and idle.

[0205] 105: Safety Limit Exceeding Emergency Response and Operation Mode Switching:

[0206] When a safety limit violation event is triggered, the current operating mode is forcibly switched and local governance is executed in an emergency; interrupted virtual power plant frequency regulation or peak regulation commands are suspended and saved to facilitate subsequent recovery arbitration; in the coordinateable operating domain, rolling prediction of requested candidate schemes under four types of operating modes and selection of requested candidate schemes are performed to determine the candidate operating mode;

[0207] Furthermore, in the aforementioned coordinateable operating domain state, an operating mode switching hysteresis verification mechanism is constructed, allowing switching to the candidate operating mode only when the difference between the comprehensive prediction cost of the current operating mode and the candidate operating mode exceeds a set threshold; when the difference is not greater than the set threshold, the current operating mode remains unchanged.

[0208] Furthermore, in one embodiment of the present invention, 105 specifically includes:

[0209] 5.1 When a security violation event is triggered, the current operating mode is forcibly switched, and local governance is executed as an emergency measure;

[0210] That is, satisfy or If this occurs, it is considered a safety violation event. In this case, skip directly. In the mode selection process, the virtual power plant frequency regulation / peak regulation command is suspended, and emergency local governance is performed according to formula (11)-formula (12).

[0211] 5.2 In the coordinateable operating domain, perform rolling prediction of request candidate solutions and optimization of request candidate solutions under four types of operating modes to determine the candidate operating mode;

[0212] That is, satisfy and At that time, the request candidate schemes under the four types of operation modes are predicted in the future control time domain by rolling according to equations (17)-(26), and infeasible schemes are filtered according to equations (27)-(29). Finally, the selected scheme is obtained according to equation (30). .

[0213] 5.3 Construct a hysteresis verification mechanism for switching operating modes, allowing switching to the candidate operating mode only when the difference between the comprehensive prediction cost of the current operating mode and the candidate operating mode exceeds a set threshold;

[0214] Mode switching hysteresis check: In the coordinateable operating domain, mode switching is only allowed when the difference between the comprehensive cost of the current operating mode and the comprehensive cost of the candidate operating mode exceeds a set threshold, thus suppressing frequent switching caused by operating condition fluctuations.

[0215] (31)

[0216] In equation (31), where, The predicted cost for the current operating mode; The cost of predicting candidate operating modes; This is the mode switching threshold.

[0217] The emergency local governance mode switching triggered by voltage safety exceeding the limit is not subject to the above mode switching hysteresis verification constraint, and is executed first, while the currently executing frequency modulation / peak shaving command is suspended.

[0218] 5.4 Suspending and saving interrupted virtual power plant control commands: When the virtual power plant frequency regulation or peak regulation mode is interrupted by emergency execution of local governance, record the information of unfinished control tasks and save the current operating parameters according to formula (32). , and Suspend and save the data for later resumption of arbitration:

[0219] (32)

[0220] In equation (32), The amount of electricity not regulated was not completed. The target power is suspended and saved; The remaining execution steps; To control the cycle.

[0221] Suspended command recovery rules: After the emergency execution of local governance ends and the voltage returns to the safe domain, the suspended frequency modulation / peak shaving commands are not directly resumed. Instead, they are re-participated as corresponding request candidate schemes in the comprehensive prediction cost calculation and candidate operation mode selection in the next control cycle. The result of the candidate operation mode selection determines whether to enable them. When the communication interruption exceeds the set time limit or the equipment interlock is triggered, the command suspension status is cleared and the corresponding request candidate scheme is set to invalid.

[0222] 106: Generation of PCS phase-specific active and reactive power coordination commands

[0223] Based on the active power benchmark corresponding to the effective candidate operating mode, the PCS phase-by-phase reactive power compensation calculation is performed, and finally, the three-phase active and reactive power output commands of the energy storage grid connection point are generated.

[0224] Furthermore, in one embodiment of the present invention, 106 specifically includes:

[0225] The PCS phase-by-phase active and reactive power coordination execution model is constructed using equations (33) and (34) to perform three-phase active power allocation. The first step is equal allocation: the current active power of the energy storage has been determined in the current mode. Then, first order the three phases to each take... When there is no three-phase imbalance and no abnormal phase voltage, the command for each phase is this average value. The second step is phase correction: only when the imbalance degree If the voltage exceeds 10%, or if one phase is below 209V and another above 231V, inter-phase active power transfer is added to the average result. Using the average three-phase voltage as a benchmark, the lower-voltage phase increases discharge, and the higher-voltage phase increases charging; the algebraic sum of the three-phase corrections is zero, therefore the total active power remains equal to... Each phase is further limited to its single-phase active power upper and lower limits; if the sum of the three phases after limiting exceeds the upper limit of the total energy storage power, it is scaled proportionally to the upper limit. This ensures that the sum of the three phases' active power equals the target total active power of the energy storage.

[0226] (33)

[0227] In the formula, , , These represent the three-phase output active power of the PCS. This represents the current active power of the energy storage corresponding to the current operating mode.

[0228] Phase-by-phase reactive power compensation calculation generates corresponding reactive power commands based on the voltage deviation and unbalance of each phase, achieving voltage management and unbalance suppression:

[0229] (34)

[0230] In the formula, This is for phase-specific reactive power compensation. This represents the voltage vector at the three-phase grid connection point. , These are the normal low and high voltage thresholds, respectively; This refers to the reactive power distribution rules for each phase.

[0231] The phase-specific reactive power allocation rule is as follows: for voltages lower than... The phase, according to its relationship The deviation ratio outputs positive reactive power to boost the voltage; for voltages higher than... The phase, according to its relationship The deviation ratio outputs negative reactive power to suppress voltage; when the three-phase imbalance... When the threshold is exceeded, a three-phase unbalanced reactive power correction is applied; the reactive power output of each phase is ultimately limited to... Within the range.

[0232] Example 2:

[0233] A conflict coordination and control system for energy storage tasks in a distribution area based on parallel candidate prediction, comprising:

[0234] The four-mode mutually exclusive operation model construction module is used to define four types of operation modes: local governance, frequency regulation, peak regulation, and idle, and to construct a four-mode mutually exclusive operation model that allows only one mode to be effective at any given time; the grid connection point status acquisition and voltage discrimination module is used to acquire the three-phase phase voltage and operation status of the energy storage grid connection point in real time, and divides the control branches into two parts: safety over-limit events and coordinateable operation domain based on a two-level voltage safety threshold discrimination model, and updates the energy storage charge state according to the control cycle;

[0235] The local governance demand determination and conflict decomposition module is used to calculate high-voltage governance demand, low-voltage governance demand, three-phase imbalance governance demand, and reverse load rate governance demand respectively to determine four types of local governance demands. When there is no charging / discharging direction conflict among the four types of local governance demands, the target power is directly determined based on the triggered governance demand. When there is a charging / discharging direction conflict among the four types of local governance demands, candidate schemes for discharging electrons and candidate schemes for charging electrons are constructed respectively, and compared by substituting them into the comprehensive prediction cost function. The candidate scheme with the smaller comprehensive prediction cost is selected as the local governance request candidate scheme, and the three-phase target active power and three-phase target reactive power are output.

[0236] The parallel candidate rolling prediction cost calculation module is used to determine the active power benchmark of the request candidate schemes under the four types of operation modes: local governance, frequency regulation, peak regulation, and idle. Within the set rolling prediction time domain, the module performs recursive prediction of the energy storage state of charge and the operating state of the distribution area for each request candidate scheme on a cycle basis. Based on the comprehensive prediction cost function containing multiple penalty terms, the module calculates the comprehensive prediction cost of each request candidate scheme within the rolling prediction time domain.

[0237] The feasible domain constraint verification and mode selection module is used to verify the feasible domain constraints of each request candidate solution, eliminate infeasible solutions that do not meet the constraints, and select the solution with the minimum comprehensive prediction cost from the remaining feasible request candidate solutions as the candidate running mode.

[0238] The safety limit violation emergency and mode switching module is used to forcibly switch the current operating mode to the local governance mode to perform emergency governance when a safety limit violation event is triggered, and to suspend and save the interrupted virtual power plant frequency regulation or peak regulation commands; under the coordinateable operating domain, it performs rolling prediction and selection of requested candidate schemes under four types of operating modes to determine the candidate operating mode;

[0239] The PCS phase-by-phase instruction generation module is used to perform PCS phase-by-phase reactive power compensation calculation based on the active power benchmark corresponding to the effective candidate operating mode, and generate three-phase active power output instructions and three-phase reactive power output instructions for the energy storage grid connection point.

[0240] Application example:

[0241] To enable those skilled in the art to better understand the present invention, the following examples are provided:

[0242] I. Example Description and Simulation Result Analysis:

[0243] To verify its effectiveness, this invention uses the IEEE 33-node distribution system as the test system and performs a case study analysis. In the test system, an energy storage system with a rated power of 0.5MW and a rated capacity of 2MWh is connected to node 18. Distributed photovoltaic power of 0.5MW, 1.5MW, and 2.2MW is connected to nodes 18, 25, and 33, respectively. The voltage monitoring object of the energy storage grid connection point is the three-phase voltage of the node where the grid connection point is located. Umin and Umax are obtained from equations (2)-(3). The safety voltage threshold is 198V-235V. The normal voltage management threshold is 209V-231V. The upper limit of the ramp rate is set to 0.25MW / 5min. This limit is used for power smoothing control under normal operating conditions. Equipment fault interlocking and safety over-limit emergency control of Umin<198V or Umax>235V are not subject to this ramp constraint to ensure timely execution of safety actions. Voltage deviation benchmark weight Wv Three-phase unbalanced weights Reverse load factor Frequency modulation tracking weight Peak shaving tracking weight SOC reference tracking weight Battery degradation weight Mode switching penalty weight Hard constraint penalty value Mode switching confirmation threshold is set .

[0244] To fully demonstrate the effectiveness of the proposed method for coordinating and controlling conflicts in energy storage tasks in transformer substations based on parallel candidate prediction, two schemes, Scheme A and Scheme B, are set up for comparison in the case study section.

[0245] Option A: Fixed priority coordination strategy, with the priority order as follows: security overrun event → frequency adjustment → peak adjustment → local governance → idle.

[0246] Solution B: The parallel candidate prediction coordination strategy proposed in this invention directly governs the safe over-limit situation locally, and constructs four request candidate schemes in parallel for comprehensive prediction cost optimization when the limit is not exceeded.

[0247] Both schemes were executed simultaneously on the IEEE 33-node test system. The simulation duration was 0:00–24:00, the control cycle was 5 minutes, and the initial and final daily cycle SOC targets were both 50%. The virtual power plant's peak-shaving command periods included 6:45–7:30, 10:35–11:15, 12:30–14:00, and 18:00–22:00, while the frequency regulation command periods included 9:00–10:30, 12:00–13:00, and 18:30–20:00. The periods from 12:30–13:00 and 18:30–20:00 were overlapping periods for frequency regulation and peak-shaving commands, used to simulate internal task conflicts within the virtual power plant. The coordination performance of the two schemes is shown in Table 1. Among them, the cumulative duration of safety threshold overrun statistics is the cumulative value of control steps Umin<198V or Umax>235V, which reflects the duration of safety overrun events as shown in equation (4); the cumulative duration of ordinary governance threshold overrun statistics is the cumulative value of control steps Umin<209V or Umax>231V, which reflects the degree of overrun in the ordinary local governance triggering intervals shown in equations (7)-(8); the root mean square error of frequency regulation and peak regulation power tracking is calculated only during the period when the virtual power plant issues frequency regulation and peak regulation commands, based on the root mean square of the difference between the actual output of energy storage and the target power, with the unit being MW. The smaller the value, the better the tracking.

[0248] Table 1

[0249]

[0250] As shown in Table 1, the cumulative duration of safety threshold exceedance in Scheme B decreased from 20 minutes in Scheme A to 0 minutes, indicating that the present invention can effectively eliminate severe voltage safety exceedance events. The duration of virtual power plant command suppression of local governance decreased from 75 minutes to 15 minutes, a reduction of 80%. The number of mode switching times decreased from 41 to 25, a reduction of 39.0%, indicating that the parallel candidate prediction and mode switching hysteresis mechanism can effectively alleviate task conflicts and suppress frequent mode switching. The cumulative duration of ordinary governance threshold exceedance in Scheme B is 235 minutes, an increase of 15 minutes compared to 220 minutes in Scheme A, but the voltage exceedance deviation integral decreased from 0.0587 pu·h to 0.0344 pu·h, a reduction of 41.4%. This indicates that although the exceedance duration near the ordinary governance threshold is slightly increased in Scheme B, it significantly reduces the voltage exceedance depth and prevents it from further developing into a safety exceedance event. Regarding virtual power plant service tracking, only the time periods during which the corresponding instructions were actually received were statistically analyzed. The root mean square errors (RMS) for frequency regulation power tracking of Scheme A and Scheme B were 0.270MW and 0.025MW, respectively, and the RMS errors for peak shaving power tracking were 0.198MW and 0.023MW, respectively. This indicates that Scheme B has higher tracking accuracy for received frequency regulation and peak shaving instructions. In summary, this invention can dynamically coordinate voltage safety, local governance, and virtual power plant services, thereby eliminating severe voltage safety exceedances, reducing the degree of voltage exceedances, decreasing task conflicts, reducing the number of mode switching operations, and improving the tracking accuracy of received instructions.

[0251] Figure 1 The comparison between the grid connection point Umin / Umax of the energy storage and the safety threshold is shown, demonstrating that the present invention can accurately identify the voltage safety over-limit process; Figure 2 The results of the mutually exclusive operation of the four modes of local governance, frequency regulation, peak shaving and idle are shown, indicating that only one mode can obtain the power output right in any control cycle; Figure 3 The technical effects shown in Table 1 were verified from a quantitative perspective. Figure 4 The overall process of this method is shown.

[0252] In this specification, the illustrative descriptions of the invention are not necessarily directed at the same embodiments or examples. Those skilled in the art can combine and integrate the different embodiments or examples described herein. Furthermore, the embodiments described in this specification are merely examples of inventive concepts and are not intended to limit the scope of protection claimed by the invention.

Claims

1. A method for coordinated control of conflicts in energy storage tasks in transformer substations based on parallel candidate prediction, characterized in that, include: 101: Energy Storage Grid Connection Point Status Acquisition and Four-Mode Mutually Exclusive Operation Model Construction: Define four types of operation modes: local governance, frequency regulation, peak shaving, and idle; construct a four-mode mutually exclusive operation model that allows only one mode to be effective at any given time; collect the three-phase voltage and operation status of the energy storage grid connection point in real time; construct a two-level voltage safety threshold discrimination model; divide the control into two branches: safety over-limit events and coordinateable operation domain; construct a dynamic model of energy storage charge state; and update the energy storage charge state according to the control cycle. 102: Identifying and Decomposing Local Governance Needs from Multiple Dimensions Calculate the high voltage, low voltage, three-phase imbalance, and reverse load rates respectively to determine the four types of local governance needs; If there is no conflict between the charging and discharging directions in the local governance requirements, the candidate schemes for the local governance requests are determined based on the triggering requirements; if there is a conflict between the charging and discharging directions, candidate schemes for discharging electrons and candidate schemes for charging electrons are constructed respectively, and the three-phase target active power and three-phase target reactive power corresponding to the two sub-candidate schemes are output. 103: Comprehensive Cost Calculation for Four-Mode Parallel Candidate Rolling Prediction A comprehensive prediction cost function containing multiple penalty terms is constructed; the active power benchmarks for request candidate schemes under four operating modes—local governance, frequency regulation, peak shaving, and idle—are determined respectively; when the local governance mode includes both discharge and charging candidate schemes, the energy storage state of charge and the operating state of the distribution area are recursively predicted within a set rolling prediction time domain, the comprehensive prediction cost of the two sub-candidate schemes is calculated, and the one with the smaller comprehensive prediction cost is selected as the local governance request candidate scheme; further, rolling prediction is performed on the request candidate schemes under the four operating modes, and the comprehensive prediction cost of each request candidate scheme is calculated. 104: Feasibility Region Constraint Verification and Optimal Operation Mode Selection Each candidate solution is checked for feasible domain constraints, and infeasible solutions that do not meet the constraints are eliminated; among the remaining feasible candidate solutions, the solution with the minimum comprehensive prediction cost is selected as the candidate running mode. 105: Safety Over-Limit Emergency Response and Operation Mode Switching When a safety limit violation event is triggered, the current operating mode is forcibly switched and local governance is executed in an emergency; interrupted frequency regulation or peak regulation commands of virtual power plants are suspended and saved; in the coordinateable operating domain, rolling prediction of requested candidate schemes under four types of operating modes and selection of requested candidate schemes are performed to determine the candidate operating mode; 106: Generation of PCS phase-specific active and reactive power coordination commands Based on the active power benchmark corresponding to the effective candidate operating mode, the PCS phase-by-phase reactive power compensation calculation is performed, and finally, the three-phase active and reactive power output commands of the energy storage grid connection point are generated.

2. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction as described in claim 1, characterized in that, The four-mode mutually exclusive operation model is constructed as follows: In the formula, , , , These are status flags for local governance, frequency regulation, peak shaving, and idle mode, respectively. They are all variables of 0 or 1. A value of 1 indicates that the corresponding operating mode is in effect, and a value of 0 indicates that the corresponding operating mode is not in effect.

3. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction as described in claim 1, characterized in that, The aforementioned dynamic model of energy storage state of charge is as follows: For the first Control the state of charge of the energy storage cycle; For the first Active power of the energy storage grid connection point during the cycle, as agreed Indicates discharge. Indicates charging; To control the cycle; This refers to the rated capacity of the energy storage.

4. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction as described in claim 1, characterized in that, The charge / discharge direction conflict refers to the fact that the charge / discharge power directions required by at least two types of local governance needs are inconsistent.

5. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction according to claim 1, characterized in that, The formula for calculating the target power benchmark for local charging governance needs is as follows: In the formula, , These are the power requirements for high voltage regulation and reverse load rate regulation, respectively.

6. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction according to claim 1, characterized in that, The multiple penalty items include voltage deviation, three-phase imbalance, reverse overload rate, frequency modulation tracking, peak shaving tracking, SOC deviation, battery degradation, and mode switching penalty items. The comprehensive predicted cost is obtained by weighting and summing each penalty item according to a preset weight.

7. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction as described in claim 6, characterized in that, The formula for calculating the voltage deviation penalty term is as follows: In the formula, , These are the minimum and maximum voltages at the grid connection point that exceed the low and high voltage thresholds of 209V and 231V respectively. The values ​​are normalized to the range of 0-1 according to the 209V-198V and 231V-235V ranges. The larger the value, the more severe the voltage deviation. , The weighting coefficients are for low and high voltage deviations. The formula for calculating the three-phase imbalance penalty term is as follows: In the formula, The unbalance threshold; Take the three-phase imbalance threshold Up to the maximum severity difference; This represents the current three-phase imbalance. The formula for calculating the reverse load rate penalty is as follows: In the formula, The reverse load rate threshold; Take the reverse load rate threshold Up to the maximum severity difference; This represents the current reverse load rate. The formula for calculating the frequency modulation tracking penalty term is as follows: In the formula, Target power for frequency modulation; This represents the upper limit of the active power of energy storage. The formula for calculating the peak-shaving tracking penalty term is as follows: In the formula, The target power for peak shaving; The current active power of energy storage; As the upper limit of the active power of energy storage, the formula for calculating the SOC deviation penalty term is: In the formula, This represents the current state of charge of the energy storage system. The reference state of charge is as of the day-ahead. To determine the normalized bandwidth of the charged state, take and The difference is used to relative SOC The deviation is normalized and included Penalties; The formula for calculating the battery degradation penalty is as follows: In the formula: This represents the absolute value of the active power of energy storage. This represents the upper limit of the active power of energy storage. The formula for calculating the mode switching penalty is as follows: In the formula These are candidate patterns currently being evaluated. This refers to the currently running mode.

8. The method for coordinated control of energy storage task conflicts in a distribution area based on parallel candidate prediction according to claim 1, characterized in that, The feasible domain constraint verification includes three types of feasible domain constraint verification: state of charge operating range, charging and discharging power, and power ramp rate. Specifically, it includes: verifying whether the energy storage state of charge in each control cycle within the prediction time domain is within the preset state of charge operating range; verifying whether the charging and discharging power in each control cycle is within the preset charging and discharging power limit; and verifying whether the power change between adjacent control cycles does not exceed the preset power ramp rate limit.

9. The method for coordinated control of energy storage task conflicts in transformer substations based on parallel candidate prediction according to claim 1, characterized in that, Within the coordinateable operating domain, a hysteresis verification mechanism for switching operating modes is constructed: switching to the candidate operating mode is allowed only when the difference between the comprehensive prediction cost of the current operating mode and the candidate operating mode exceeds a set threshold; when the difference is not greater than the set threshold, the current operating mode remains unchanged.

10. A conflict coordination and control system for energy storage tasks in a distribution area based on parallel candidate prediction, characterized in that, include: The four-mode mutually exclusive operation model construction module is used to define four types of operation modes: local governance, frequency regulation, peak regulation, and idle, and to construct a four-mode mutually exclusive operation model that allows only one mode to be effective at any given time; the grid connection point status acquisition and voltage discrimination module is used to acquire the three-phase phase voltage and operation status of the energy storage grid connection point in real time, and divides the control branches into two parts: safety over-limit events and coordinateable operation domain based on a two-level voltage safety threshold discrimination model, and updates the energy storage charge state according to the control cycle; The local governance demand determination and conflict decomposition module is used to calculate high-voltage governance demand, low-voltage governance demand, three-phase imbalance governance demand, and reverse load rate governance demand respectively to determine four types of local governance demands. When there is no charging / discharging direction conflict among the four types of local governance demands, the target power is directly determined based on the triggered governance demand. When there is a charging / discharging direction conflict among the four types of local governance demands, candidate schemes for discharging electrons and candidate schemes for charging electrons are constructed respectively, and compared by substituting them into the comprehensive prediction cost function. The candidate scheme with the smaller comprehensive prediction cost is selected as the local governance request candidate scheme, and the three-phase target active power and three-phase target reactive power are output. The parallel candidate rolling prediction cost calculation module is used to determine the active power benchmark of the request candidate schemes under the four types of operation modes: local governance, frequency regulation, peak regulation, and idle. Within the set rolling prediction time domain, the module performs recursive prediction of the energy storage state of charge and the operating state of the distribution area for each request candidate scheme on a cycle basis. Based on the comprehensive prediction cost function containing multiple penalty terms, the module calculates the comprehensive prediction cost of each request candidate scheme within the rolling prediction time domain. The feasible domain constraint verification and mode selection module is used to verify the feasible domain constraints of each request candidate solution, eliminate infeasible solutions that do not meet the constraints, and select the solution with the minimum comprehensive prediction cost from the remaining feasible request candidate solutions as the candidate running mode. The safety limit violation emergency and mode switching module is used to forcibly switch the current operating mode to the local governance mode to perform emergency governance when a safety limit violation event is triggered, and to suspend and save the interrupted virtual power plant frequency regulation or peak regulation commands; under the coordinateable operating domain, it performs rolling prediction and selection of requested candidate schemes under four types of operating modes to determine the candidate operating mode; The PCS phase-by-phase instruction generation module is used to perform PCS phase-by-phase reactive power compensation calculation based on the active power benchmark corresponding to the effective candidate operating mode, and generate three-phase active power output instructions and three-phase reactive power output instructions for the energy storage grid connection point.

Images