Multi-source load response oriented smart power distribution network hierarchical collaborative control method and system

Abstract

This invention discloses a hierarchical collaborative control method and system for intelligent distribution networks oriented towards multi-source load response, relating to the field of power system planning and operation technology. The method includes: calculating the power deviation and flexibility requirements of each node according to the distribution network operation scenario type to obtain corresponding load response requirement indicators; acquiring the state information of distribution network flexibility resources and constructing a distributed resource model; based on the load response requirement indicators and the distributed resource model, constructing a two-layer decoupled scheduling model including a global objective function and regional constraints; iteratively solving the two-layer decoupled scheduling model to obtain global scheduling parameters and a set of execution tasks, and distributing the execution tasks to the corresponding local control units; updating the global scheduling parameters and regional constraints based on feedback data from the local control units, and dynamically correcting the distribution network operation state.

Description

Technical Field

[0001] This invention relates to the field of power system planning and operation technology, and more specifically, to a hierarchical collaborative control method and system for intelligent distribution networks oriented towards multi-source load response. Background Technology

[0002] The operation and control of power distribution networks face severe challenges due to uncertainties on both the source and load sides. Traditional methods for power distribution network operation and control struggle to efficiently coordinate various flexible resources such as distributed photovoltaic power, energy storage, and controllable loads, and are insufficient in achieving full renewable energy consumption and ensuring system safety, stability, and economic operation. Therefore, various solutions have emerged in the field aimed at improving the coordinated control capabilities of power distribution networks.

[0003] For example, the invention patent announcement CN109559035A discloses a two-layer planning method for urban power distribution networks that considers flexibility. Its core technology lies in the collaborative optimization during the planning phase. This method generates typical flexibility demand scenarios, establishes a flexibility resource adjustment capability model, constructs a two-layer model of upper-level planning and lower-level scheduling, and uses a multi-objective particle swarm optimization algorithm to solve it, ultimately deriving an optimized planning scheme for the power distribution network and a coordinated scheduling scheme for flexibility resources.

[0004] For example, the invention patent announcement CN114142532B, which describes a method and system for coordinated control of distributed photovoltaic power generation, grid, load, and energy storage, focuses on real-time coordination during the operation phase. This method predicts the adjustable capacity required by the power generation, grid, load, and energy storage at the next moment and compares it with actual operating data. Then, based on a hierarchical and zoning principle, it determines a coordinated and optimized scheduling strategy, aiming to achieve full absorption of distributed photovoltaic power and economical operation of energy storage.

[0005] However, the aforementioned existing technical solutions still have significant limitations in addressing the real-time control challenges brought about by the integration of high proportions of renewable energy. CN109559035A provides a static solution oriented towards long-term planning, and its optimization results are essentially based on predictive planning for typical scenarios, unable to respond to dynamic fluctuations and uncertainties at the second or minute level in actual operation. While CN114142532B involves operational control, its control logic relies more on a feedforward open-loop process of "prediction-comparison-zonal scheduling," lacking an optimization mechanism that can dynamically close-loop correct global scheduling parameters based on real-time feedback from local control units. Therefore, when facing complex and ever-changing actual operating conditions, existing methods struggle to achieve rapid and accurate coordinated control of power deviations at each node and system voltage, hindering further improvements in the optimization level of distribution network operation.

[0006] To address the above problems, this invention proposes a solution. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a hierarchical collaborative control method for smart distribution networks oriented towards multi-source load response. By constructing load response demand indicators based on operational scenario identification and combining them with a distributed flexibility resource model, a hierarchical collaborative optimization and dynamic correction mechanism is established to solve the problems of high coupling degree of multi-source load response scheduling and difficulty in balancing supply and demand and voltage stability in existing distribution networks.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A hierarchical collaborative control method for smart distribution networks with multi-source load response includes the following steps: 1) Calculating the power deviation and flexibility requirements of each node based on the distribution network operation scenario type to obtain corresponding load response requirement indicators; 2) Obtaining the state information of distribution network flexibility resources, and introducing a dynamic safety margin coefficient to correct the adjustable capacity boundary of resources by combining historical uncertainty and real-time fluctuation characteristics of resource response, thus constructing a distributed resource model; 3) Constructing a two-layer decoupled scheduling model based on the load response requirement indicators and the distributed resource model; 4) Iteratively solving the two-layer decoupled scheduling model to obtain global scheduling parameters and a set of execution tasks, and distributing the execution tasks to the corresponding local control units; 5) Based on the feedback data from the local control units, adaptively adjusting the convergence step size based on the decay rate of the nonlinear state residual, updating the global scheduling parameters and regional constraints, and dynamically correcting the distribution network operation state.

[0009] This invention provides a hierarchical collaborative control system for intelligent distribution networks oriented towards multi-source load response, comprising: a demand calculation module, used to calculate the power deviation and flexibility requirements of each node according to the distribution network operation scenario type, and obtain the corresponding load response demand index; a resource allocation module, used to acquire the status information of distribution network flexibility resources, and introduce a dynamic safety margin coefficient to correct the adjustable capacity boundary of resources by combining the historical uncertainty and real-time fluctuation characteristics of resource response, and construct a distributed resource model; a collaborative optimization module, used to construct a two-layer decoupled scheduling model containing a global objective function and regional constraints based on the load response demand index and the distributed resource model; a collaborative control module, used to iteratively solve the two-layer decoupled scheduling model to obtain global scheduling parameters and a set of execution tasks, and distribute the execution tasks to the corresponding local control units; and a feedback correction module, used to adaptively adjust the convergence step size based on the decay rate of the nonlinear state residual according to the feedback data executed by the local control units, update the global scheduling parameters and regional constraints, and dynamically correct the distribution network operation status.

[0010] The technical effects and advantages of the multi-source load response intelligent distribution network hierarchical collaborative control method of this invention are as follows: This invention calculates the power deviation and flexibility requirements of each node according to the type of distribution network operation scenario, constructs a distributed resource model based on the flexibility resource status information, and establishes a two-layer decoupled scheduling model that includes a global objective function and regional constraints. Combined with feedback data from local control units, it achieves adaptive dynamic correction of global scheduling parameters, thereby realizing the coordinated optimization and dynamic balance of multi-source loads in the distribution network. This improves the scheduling flexibility and operational stability of the system across multiple time scales. Through a hierarchical and coordinated control mechanism, it helps improve power distribution efficiency and flexibility resource utilization, effectively solving the problems of supply-demand imbalance and the difficulty in real-time correction of voltage fluctuations at local nodes in distribution network operation. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the hierarchical collaborative control method for smart distribution networks oriented towards multi-source load response provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the power distribution network system topology provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the residual convergence curve of the augmented Lagrange algorithm provided in an embodiment of the present invention; Figure 4 This is a schematic diagram comparing the node voltage distribution provided in an embodiment of the present invention; Figure 5 This is a block diagram of a hierarchical collaborative control system for intelligent distribution networks oriented towards multi-source load response, provided in an embodiment of the present invention. Detailed Implementation

[0012] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0013] Example 1, Figure 1 This invention presents a hierarchical collaborative control method for smart distribution networks oriented towards multi-source load response, comprising the following steps: S1. Based on the distribution network operation scenario type, calculate the power deviation and flexibility requirements of each node to obtain the corresponding load response requirement index.

[0014] In this embodiment, the distribution network operation scenario type is obtained by scenario identification from the real-time operation data of the distribution network. The real-time operation data includes the voltage amplitude of each node in the distribution network. Current amplitude Active power reactive power ,frequency and line carrying capacity Monitoring quantities; Among them, nodes This refers to the group numbering of electrical connection units in a distribution network, including the low-voltage side of transformers, branch busbars, and user access points; lines Represents a node With nodes Feeders between them.

[0015] The scene recognition steps are as follows: First, the above real-time running data is input into a preset scene recognition model, which includes a threshold discrimination layer and a cluster analysis layer.

[0016] The threshold discrimination layer is used to perform a one-time screening of data based on preset index thresholds (voltage allowable range [0.94, 1.06] pu and line current carrying rate upper limit 100%) to preliminarily determine whether there are abnormalities such as low voltage, heavy overload, and insufficient reactive power; The clustering analysis layer uses the fuzzy C-means clustering algorithm to perform multidimensional feature clustering on the discriminated running data and calculates the membership function of each scene. The calculation formula is as follows: , in, For the first The feature vector of each sample (including ), (and in the denominator) ) is the first (or the cluster center of the m-th type of scenario) , The number of scene categories. The fuzzy factor (with a value of 2).

[0017] The cluster center The data was obtained through offline training using historical operational data samples. Specifically, a sample set was constructed by selecting representative multi-day typical load curves, voltage curves, and line power flow data. The clusters are initially divided according to operational scenario labels ("peak shaving scenario", "voltage regulation scenario", "line overload scenario", etc.), with the mean of features of each type of sample used as the initial cluster centers. In each iteration, the cluster centers are recalculated according to the fuzzy membership update rule, using the following formula: , Until the change in cluster centers Stop iteration when, where To set the convergence threshold.

[0018] Finally, the operating scenario type of the distribution network at the current moment is determined by the principle of maximum membership.

[0019] The load response demand index is obtained by weighted normalization of the power deviation value and flexibility demand value of each node. The weighting coefficient is determined according to the importance of the region where the node is located and the historical load fluctuation characteristics, so as to characterize the relative adjustment demand of each node in the system supply and demand balance.

[0020] The node power deviation value The calculation formula is as follows: , in, The measured active power at the node. The reference power for the scenario corresponding to this node can be calculated from the day-ahead planned load curve or historical average.

[0021] when This indicates that the load exceeds the plan and it is necessary to reduce the load or increase the distributed power supply; when This indicates insufficient load, requiring either increasing the load or decreasing the distributed power supply.

[0022] The flexibility requirement value The intensity of a node's demand for flexibility resources under system constraints is defined as: , in, For node voltage deviation, To allow the upper limit of voltage deviation, The difference in current carrying capacity between adjacent lines. For line current carrying capacity limits, , , For normalized weight coefficients, satisfying .

[0023] The load response demand index The formula used to reflect the overall adjustment needs of nodes in the current scenario is as follows: , in, The weighting coefficients for node importance. The two factors, weighted by flexibility requirements, can be adjusted according to the system's operational strategy. It represents the maximum absolute value of power deviation across all nodes in the distribution network. This represents the maximum extreme value of the flexibility requirement index among all nodes. The node importance weighting coefficient The calculation formula is determined by the node's hierarchy in the topology and its historical load fluctuation characteristics:

[0024] in, This represents the connectivity of a node, which is the number of its child nodes. This represents the historical standard deviation of node loads, used to reflect load volatility. and Let be the weighting coefficient, satisfying , This represents the maximum extreme value of the historical load fluctuation standard deviation of all nodes in the global distribution network.

[0025] The technical effect of the weighting coefficient is that by taking into account the importance and historical volatility of nodes, the load response demand index can more accurately reflect the actual adjustment needs of nodes for the system's supply and demand balance, thereby improving the accuracy of response priority determination in subsequent optimized scheduling.

[0026] The load response demand index calculated by the above method in this step quantifies the flexibility requirements of different nodes. Specifically, it can identify the potential adjustment space and response urgency of each node under the operation scenario of the distribution network, provide accurate demand input for subsequent distributed resource optimization and scheduling, and improve the observability and controllability of the distribution network under different operation scenarios.

[0027] S2 obtains the status information of the distribution network's flexibility resources, and introduces a dynamic safety margin coefficient by combining the historical uncertainty and real-time fluctuation characteristics of the resource response to correct the adjustable capacity boundary of the resources and construct a distributed resource model.

[0028] In this embodiment, the process of acquiring the status information of distribution network flexibility resources, combining the historical uncertainty and real-time fluctuation characteristics of resource response to introduce a dynamic safety margin coefficient to correct the adjustable capacity boundary of resources, and constructing a distributed resource model includes: Based on the status information of the flexibility resources, the initial adjustable capacity of the flexibility resources is obtained; The flexible resources mentioned include distributed power sources, energy storage systems, interruptible loads, and electric vehicle charging stations. Their status information includes, but is not limited to: Real-time output power Rated capacity Remaining energy State of charge (SOC) and upper and lower limits of regulated power Equipment availability status (1 indicates available, 0 indicates unavailable), response time Duration Operating mode (charging / discharging, up / down adjustment), and the access node number where the resource is located. wait.

[0029] Based on the above status information, the adjustable capacity of the flexibility resources is obtained as follows: Initial adjustable capacity of each flexibility resource It is determined by the current operating power and its adjustable range, as shown in the formula: , If the current resources are within the feasible range for upward adjustment, then If it falls within the feasible range for downward adjustment, then When resources are unavailable, The corresponding adjustable capacity .

[0030] For energy storage resources, their adjustable capacity must simultaneously meet energy constraints: , in, The upper and lower limits of the energy limit for energy storage resource r. It is the change in energy. The charging and discharging power of energy storage resource r at time t. This is the scheduling start time.

[0031] In this embodiment, considering that the real-time response capability of invited resources (such as electric vehicle fleets and temperature-controlled loads) can change drastically under extreme power grid fluctuations, using a static capacity limit could easily lead to secondary limit exceedances. Therefore, this step dynamically calculates the safety margin coefficient based on the real-time response variance of flexible resources, and uses the safety margin coefficient to lower the upper limit of the initial adjustable capacity to obtain the actual adjustable capacity. Specifically, this includes the following steps: First, the historical deviation between the actual response power and the planned power of the flexibility resource within a pre-defined time window is extracted, and the real-time response variance is calculated. The time window is set to include the previous T scheduling cycles, and the formula for calculating the real-time response variance is: , in, This represents the real-time response variance of resource r at the current time t. and They represent historical moments. The actual response power of this resource is compared with the planned scheduling power issued.

[0032] Secondly, based on the real-time response variance and the preset system fault tolerance threshold, a negatively correlated nonlinear decay function is constructed to calculate the dynamic safety margin coefficient. The specific formula is as follows: , in, This is the dynamic safety margin coefficient, with a value range of (0,1]. The preset system fault tolerance threshold characterizes the boundary of uncertainty that the system can absorb; It is a decay adjustment factor that controls the decay rate.

[0033] Finally, the upper limit of the initial adjustable capacity in the invitation control mode is adjusted downwards using the safety margin coefficient to obtain the actual adjustable capacity. Specifically: , , in, and These are the corrected actual adjustable capacity limit and the initial adjustable capacity limit, respectively; and respectively flexibility resources exist The actual adjustable capacity lower limit after time correction and the initial adjustable capacity lower limit.

[0034] The technical effect of the above-mentioned methods is that by introducing a safety margin decay mechanism based on real-time variance, the adjustable capacity upper bound of highly uncertain resources can be dynamically compressed under extreme conditions such as resource disconnection or a sharp increase in response variance. This is equivalent to establishing a dynamic safety redundancy fallback at the system's bottom layer, ensuring that scheduling instructions issued by the optimization layer always fall within the "absolutely executable" safety range, effectively preventing the risk of secondary limit violations caused by deficiencies in deterministic assumptions.

[0035] Furthermore, by combining the response speed, adjustment method, control method, and electrical characteristics of the access nodes of the flexible resources, the adjustable capacity is adjusted, and the spatial distribution of the flexible resources is adjusted according to the power grid topology information and the electrical characteristics of the nodes to obtain a distributed resource model.

[0036] The response speed The calculation formula is as follows: , A faster response time provides stronger short-term dynamic support for the system. The adjustable capacity is weighted and adjusted based on the response time, using the following formula: , in This is a reference response rate constant used for normalization.

[0037] The adjustment methods are divided into rigid adjustment and flexible adjustment. Rigid adjustment resources, such as fixed-power distributed power sources, can only be adjusted within discrete ranges, and their effective adjustable capacity is adjusted according to the discrete correction factor. For flexible adjustment resources, such as continuously adjustable variable frequency loads, a correction factor of 1 is used, and the adjustable capacity is corrected according to the following formula: , The controlled methods are divided into direct control, where the dispatch center directly issues power commands, and invitation-based control, where aggregators respond according to incentive signals. Invitation-based resources have uncertainty; their actual response rate is the response probability. Therefore, the corrected formula is as follows: , in This is obtained from statistics of historical resource response records.

[0038] The electrical characteristics of the access node include the node voltage sensitivity coefficient. With tidal current sensitivity coefficient , respectively, represent the degree of impact of node power changes on voltage and line power flow.

[0039] To ensure the electrical feasibility of resource regulation, the capacity is further weighted according to the above-mentioned node voltage and power flow sensitivity, and the calculation formula is as follows: , in This is a sensitivity weighting parameter to prevent over-adjustment at nodes with weak voltage or heavy line load.

[0040] Finally, the spatial distribution of flexibility resources is adjusted based on the power grid topology information and the electrical characteristics of the nodes to obtain a distributed resource model.

[0041] Specifically, this involves: assigning each resource r according to its access node. Mapping to the node layer, we construct node-level flexible aggregated capacity, expressed as follows: , in, For nodes A collection of resources on the internet.

[0042] Combined with node voltage sensitivity matrix Establish a resource spatial distribution matrix: , This indicates the contribution of each node's flexibility resources to the regulation of the overall network voltage distribution.

[0043] The resulting distributed resource model includes attributes such as type identifier, actual adjustable capacity, dynamic safety margin coefficient, location sensitivity (sensitivity coefficient), and response probability for each type of resource, which are used as input for the subsequent two-layer decoupled scheduling model.

[0044] This step, by comprehensively considering the dynamic safety margin, adjustment methods, and spatial distribution characteristics of flexible resources, achieves unified modeling from single resource data to system-level adjustability. Its technical advantages lie in: accurately reflecting the spatiotemporal adjustment potential and response reliability of different types of resources at different nodes, providing quantifiable resource support for subsequent two-layer decoupled scheduling models, thereby improving the utilization efficiency and scheduling accuracy of distribution network flexible resources.

[0045] S3. Based on the load response demand index and the distributed resource model, a two-layer decoupled scheduling model containing a global objective function and regional constraints is constructed.

[0046] In this embodiment, the objective function of the two-layer decoupled scheduling model is a joint objective function that minimizes grid load fluctuations, the cost of flexible resource scheduling, and the total energy efficiency loss of the system, while maximizing the efficiency of flexible resource scheduling. Specifically: The load fluctuation of the power grid Defined as the difference between the actual load on the power grid and the reference load, the objective is to minimize this fluctuation. This is based on the load response demand index calculated in step S1. The total regulation demand of the power grid can be equivalently represented by the sum of the load response demand indices of all nodes, reflecting the comprehensive fluctuation state of the entire grid under power deviation and flexibility constraints, as shown in the following formula: , Among them node The load response requirements in the current scenario reflect the power deviation and flexibility requirements of each node.

[0047] Flexible resource scheduling cost It is based on the scheduling volume of each flexible resource. and unit scheduling cost The calculated cost of flexible resource scheduling is expressed as follows: , in, Let r be the unit scheduling cost of resource r. This represents the amount of resource r to be scheduled within a given time period, such as the amount of energy storage discharge or the amount of interruptible load reduction. This cost item reflects the economic cost required to actually execute the scheduling instructions.

[0048] The total energy efficiency loss of the system This represents the energy loss caused by power grid line losses and reactive power losses, and its calculation formula is as follows: , in, and The lines are respectively The active power loss penalty factor and the reactive power loss penalty factor; For the line Active power flow on For the line The reactive power flow. This loss objective function not only reflects the physical loss of active power transmission, but also effectively suppresses the implicit performance loss caused by insufficient reactive power support or long-distance reactive power transmission during the scheduling process by penalizing reactive power flow.

[0049] The flexibility of resource scheduling efficiency The effect of flexible resource scheduling on suppressing grid load fluctuations is measured and expressed as: , in, This reflects the actual ability of each flexible resource to suppress load fluctuations, based on the adjusted flexibility resource capacity in step S2. Scheduling efficiency. The higher the value, the greater the contribution of resource scheduling to reducing grid load fluctuations.

[0050] The expression for the objective function is as follows: , in, These are weighting coefficients that determine the importance of load fluctuations, scheduling costs, energy efficiency losses, and scheduling efficiency. In the objective function, , and The weights for controlling load fluctuations, costs, and energy efficiency losses, while Prioritize resource scheduling efficiency. By adjusting the aforementioned coefficients, the optimization model can better align with actual scheduling needs and system operational requirements.

[0051] The regional constraints include node power deviation constraints, node voltage constraints, and maximum adjustable capacity limits for resources, specifically: 1) Node power deviation constraints Power deviation at node i A certain maximum limit needs to be met to prevent excessive power fluctuations at any node from affecting system stability. The expression for the node power deviation constraint is as follows: , in, This represents the maximum allowable power deviation at node i. This limit can be given by the distribution network's operating standards or system operation strategies to ensure grid stability.

[0052] 2) Node voltage constraints To ensure the safe and stable operation of the power distribution network, the voltage at each node... The node voltage constraint must be kept within the allowable range, and the expression for the constraint is as follows: , in, and These are the minimum and maximum voltage limits for node i, respectively. This constraint is a fundamental requirement for power grid operation, preventing equipment damage or system instability caused by excessively low or high voltage.

[0053] 3) Maximum adjustable capacity limit of resources Adjustment capacity of flexible resource r Constrained by equipment capacity and operational limitations. Adjustable capacity of each resource. Its maximum and minimum adjustable capacity must be met, as shown in the formula: , Based on the aforementioned objective function and constraints, a two-layer decoupled scheduling model incorporating both a global objective function and regional constraints is constructed. This model can minimize load fluctuations, scheduling costs, and energy efficiency losses while maximizing the scheduling efficiency of flexibility resources, thereby achieving optimized scheduling of the distribution network's operating state, all while ensuring stable grid operation. This step, through the combination of objective function and constraints, enables efficient scheduling of distribution network flexibility resources, ensuring the stability and economy of the grid under various operating scenarios.

[0054] S4. Iteratively solve the two-layer decoupled scheduling model to obtain global scheduling parameters and execution task set, and send the execution tasks to the corresponding local control unit.

[0055] In this embodiment, the two-layer decoupled scheduling model is solved iteratively using an augmented Lagrange-based decomposition coordination iterative algorithm; It should be noted that the augmented Lagrange method is an algorithm that optimizes constrained optimization problems by introducing Lagrange multipliers and augmented terms. In this scheme, this method is used to solve a two-level decoupled scheduling model, decomposing the global optimization problem and the regional optimization problem, and solving the subproblems of each level through cooperative iteration. Specifically: The augmented Lagrange method introduces a penalty term, which not only considers minimizing the objective function but also penalizes solutions that do not meet the constraints. The expression is as follows: , in, To augment the Lagrange function; This is the original objective function, i.e., the joint objective function in S3. This is a coupling constraint between the global and regional variables, meaning that the global variable x and the regional variable z must be consistent. This is a global-region coupling mapping matrix, whose elements represent the correspondence between global scheduling parameters and regional scheduling variables. If node i belongs to region k, then the matrix... The element in the i-th row and k-th column is 1, otherwise it is 0. Let be the Lagrange multiplier vector, used for the linear penalty of the constraints. To broaden the penalty term, used to reinforce the degree to which constraints are satisfied, This is the penalty coefficient.

[0056] In each iteration, the optimization process gradually adjusts the scheduling decisions between the global and regional sides by updating the Lagrange multipliers and penalty coefficients, so that the solution gradually approaches the optimal solution. This method is helpful for handling large-scale optimization problems, especially complex systems involving multiple levels of constraints.

[0057] The two-layer decoupled scheduling model includes a global optimization layer and a regional optimization layer, and performs the following steps in each iteration round: S401, with the equivalent power injection amount and boundary coupling variables of each region fixed, the global optimization layer solves the global objective function to update the global scheduling parameters; specifically: In each iteration, the equivalent power injection and boundary coupling variables of the regional optimization layer are first fixed. This means that the adjustment capability of nodes and the constraints within the region remain unchanged in the regional optimization layer. The equivalent power injection refers to the aggregated result of all resource scheduling within the region, reflecting the region's total contribution to system regulation. The boundary coupling variables refer to variables shared between the global layer and the regional layer, such as node power and resource scheduling.

[0058] After determining the regulation capacity and boundary variables within a fixed region, the global optimization layer updates the global scheduling parameters by solving the global objective function. The global objective function is the joint objective function defined in S3, which aims to minimize grid load fluctuations, scheduling costs, energy efficiency losses, and maximize the scheduling efficiency of flexibility resources.

[0059] The global scheduling parameters are the scheduling decisions for all nodes in the system, including power allocation and resource scheduling for each node. These global scheduling parameters are calculated and optimized globally to ensure the overall scheduling balance of the system.

[0060] S402, Based on the global scheduling parameters, the set of tasks to be executed is obtained by the regional optimization layer under regional constraints; The region optimization layer is responsible for handling resource scheduling within each region. Each region performs local scheduling based on global scheduling parameters. The goal of the region optimization layer is to optimize resource scheduling within each region while satisfying the physical constraints within that region.

[0061] The set of execution tasks includes specific scheduling tasks within each region. These tasks describe resource scheduling decisions for each region, including the power scheduling amount for each node, the energy storage discharge / charge amount, and the amount of interruptible load reduction.

[0062] S403, based on the residuals of the coupling constraints between the global and regional sides, update the multipliers and penalty coefficients of the augmented Lagrange multiplier and proceed to the next iteration; The global and regional coupling constraints are the variables and constraints that need to be shared between the global optimization layer and the regional optimization layer, such as node power scheduling and resource scheduling.

[0063] The residuals of the coupling constraints are typically quantified by comparing the global scheduling amount with the difference between the set of tasks to be executed. For example, the difference between the global scheduling amount and the regional scheduling amount is calculated using the following formula: , in, and These represent the power allocation of node i in the global and regional layers, respectively.

[0064] The updated augmented Lagrange multipliers and penalty coefficients are specifically described by adjusting the multipliers based on the residuals. and penalty coefficient This promotes the convergence of the optimized solution. The updated formula is as follows: , Where k is the number of iterations. To adjust the factor, This represents the maximum penalty coefficient.

[0065] S404: The iteration terminates when both the original residual and the dual residual are less than a preset threshold or the number of iterations reaches the upper limit, obtaining the global scheduling parameters and the set of execution tasks. The specific termination condition formula is as follows: , in, It is a preset error threshold. It is the dual residual, representing the change in the Lagrange multiplier.

[0066] This step employs an augmented Lagrange decomposition-coordination iterative algorithm, using alternating global and regional optimization layers. By combining global scheduling parameters with the task set, the scheduling parameters are updated by calculating the coupling residual between global and regional algorithms, ultimately obtaining the optimal solution that satisfies the constraints. This optimization process enables efficient scheduling of flexible resources across nodes and regions, ensuring stable grid operation and maximizing resource utilization efficiency.

[0067] S5, based on the feedback data executed by the local control unit, adaptively adjusts the convergence step size based on the nonlinear state residual decay rate, updates the global scheduling parameters and regional constraints, and dynamically corrects the distribution network operating status, specifically as follows: Real-time measurement data returned by the local control unit is collected, and a sensitivity matrix between node power and voltage / current is established by combining it with power grid topology information; the real-time measurement data includes the voltage amplitude of each node. Current amplitude And the measured values ​​of active power and reactive power. , Based on power grid topology information and power flow equations, the partial derivatives of node power with respect to node voltage and current can be obtained, and a node sensitivity matrix can be constructed. Specifically: , in, and These represent the active and reactive power of the nodes, respectively. Indicates the node voltage amplitude. This matrix represents the node phase angle. It is obtained by linearizing the power flow equations at the current operating point and can be solved using the Jacobian matrix in the Newton-Raphson method. The sensitivity matrix quantifies the impact of node power changes on voltage and current changes, enabling modeling of the propagation relationship of local disturbances under the overall network electrical conditions. This is then used for global response calculations in the subsequent dynamic correction phase. The technical effect is that it allows for fast and linear state estimation and correction of the distribution network even under complex topologies and dynamic operating conditions.

[0068] The operating state residual vector for the current round is calculated based on the real-time measurement data and global scheduling parameters; the operating state residual vector is used to reflect the deviation between the current operating state of the system and the global target state, specifically: , in, , These are respectively based on the global scheduling parameters in the previous iteration. The calculated planned values ​​of active and reactive power at the nodes, The reference voltage is used. By calculating the operating state residual vector, the deviation of the current actual operating state of the power grid from the desired dispatch state can be obtained, providing a basis for subsequent power correction.

[0069] In this embodiment, since the sensitivity matrix is ​​obtained based on the linearized solution of the power flow equations, the linear partial derivative relationship will be severely distorted when a large disturbance (such as local load shedding or short-circuit fault) causes the operating state to deviate significantly from the equilibrium point. If a fixed step size correction is used, it is very easy to cause the correction direction to be incorrect or even the system to diverge. Therefore, this embodiment introduces a damping attenuation mechanism. Based on the norm change rate of the residual vector of the operating state between the current cycle and the previous cycle, the damping attenuation mechanism is triggered to adaptively adjust the convergence step size, specifically including: First, compare the running state residual vectors of the current round (k-th iteration) with those of the previous round (k-1-th iteration) (which are respectively...) and The norm of the state residual is used to calculate the decay rate of the state residual. : , in, The L2 norm of a vector is used to represent the vector's L2 norm.

[0070] Subsequently, when the residual decay rate is less than zero (i.e. When the system deviates from the linearization equilibrium point, a damping attenuation mechanism is triggered, reducing the convergence step size by a preset ratio. The update formula is: , in, This is the convergence step size factor for the current round. The preset damping attenuation coefficient (range of values) For example, take 0.5).

[0071] The technical advantages of the above-mentioned methods are as follows: By adaptively evaluating the convergence trend of the state residuals, when the error is found to increase instead of decrease, the linearization distortion of the Jacobian matrix is ​​quickly identified. At this time, the damping attenuation is actively activated, and the amplitude of a single power correction is weakened by reducing the convergence step size. This avoids the issuance of erroneous commands due to linearization distortion, and significantly enhances the robustness and convergence stability of the closed-loop correction system under scenarios of sudden changes in grid topology or large disturbances.

[0072] The impact of local power changes on the global operating state is calculated based on the sensitivity matrix to obtain a power correction vector. The impact of local power changes on the global operating state is obtained by multiplying the sensitivity matrix by the state residual vector, as shown in the following formula: , in, Let be the power correction vector for the k-th iteration, representing the power adjustment required by each node. This vector, through the inverse mapping of the sensitivity matrix, converts the node voltage and current residuals into equivalent power correction amounts. This enables the function of converting electrical state deviations into adjustable resource power correction commands, allowing the control strategy to directly apply to distributed resources and achieving unified modeling of electrical state and control space.

[0073] Based on the power correction vector, the node power allocation and constraint boundary values ​​in the global scheduling parameters are updated, and the two-layer decoupled scheduling model is resolved to obtain the corrected execution task set, which is then distributed to the corresponding local control unit for execution. The update formula for the global scheduling parameters is: , in, This represents the set of global scheduling parameters for the k-th iteration, including the power allocation and voltage limits for each node. This is the convergence step size factor. The updated global scheduling parameters will cause corresponding changes in regional constraints (such as power boundaries and node voltage constraints). The updated parameters are used as input to resolve the two-layer decoupled scheduling model, i.e., the global-region decomposition model defined in S3 and S4, with the solution steps consistent with the aforementioned iterative process. This re-solution process ensures that the coordination relationship between the global and regional systems remains consistent after each correction, guaranteeing that the system's adaptive scheduling strategy can dynamically adjust with the operating state.

[0074] The operating state residual vector is calculated using the measurement data after execution by the local control unit. This process is repeated until the norm of the operating state residual vector is lower than a preset threshold, thereby achieving adaptive dynamic correction of the distribution network operating state. Specifically, the iterative process terminates when the following convergence condition is met: , in, This is a preset convergence threshold. If the condition is not met, local feedback data is collected again, and the above correction steps are repeated until the power grid operating state stabilizes near the target value. This process realizes dynamic closed-loop correction of the distribution network operating state, enabling the system to adaptively adjust scheduling decisions based on real-time feedback.

[0075] This step, by constructing a sensitivity matrix and a closed-loop correction mechanism for state residuals, achieves adaptive dynamic control of the distribution network from real-time measurement to dispatch commands and then to state feedback. The technical effect is that the system can maintain supply-demand balance and voltage stability even under operating disturbances, load fluctuations, and resource uncertainties, thereby improving the dynamic robustness of the power grid and the accuracy of resource dispatching.

[0076] Example 2: To verify the applicability and effectiveness of the method of the present invention, a typical urban power distribution network system was selected as a case study. The hierarchical collaborative control method for smart power distribution networks oriented towards multi-source load response described in Example 1 was used for analysis and verification. The specific steps are as follows: 1) Description of the case study scenario This embodiment selects a typical 10 kV urban distribution network containing distributed photovoltaic power, electric vehicle charging and discharging piles, and energy storage units as the research object. The system comprises 10 load nodes and 15 branch lines, with a main transformer capacity of 10 MVA. The system topology and resource distribution are as follows: Figure 2 As shown, node 1 is the substation bus node, nodes 2–10 are load nodes, nodes 4, 6, and 8 are connected to distributed photovoltaic systems, nodes 7 and 9 are configured with energy storage units, and nodes 5 and 10 are connected to electric vehicle groups.

[0077] The main operating parameters of the system are shown in Table 1.

[0078] Table 1

[0079] To demonstrate the control effect under multiple operating scenarios, three typical operating scenarios are set up: Scenario A (Peak Load Period): Maximum load, low output of distributed power sources.

[0080] Scenario B (off-peak hours): The load is low and the energy storage is in a charging state.

[0081] Scenario C (Peak PV Output): PV output is high, and some lines may experience reverse power flow risk.

[0082] Based on the above scenario, the load response demand index vector is calculated using the clustering identification method in step S1 of Example 1. Among them, the index values ​​of nodes 2, 6, and 8 are relatively high, which represents the key adjustment needs of these nodes in the system's supply and demand balance.

[0083] 2) Model calculation and optimization In this example, a two-layer decoupled scheduling model is constructed using the specific methods in steps S2–S4 of Example 1. The flexible resources in the system include photovoltaic output regulation, energy storage system charging and discharging, and interruptible load response. Distributed resources are established based on resource characteristics, and the adjustable capacity of the resources is shown in Table 2.

[0084] Table 2

[0085] Based on the objective function in step S3 of Example 1, a joint optimization objective is established: , The weighting coefficient is set as follows: , , , .

[0086] The solution is obtained using the augmented Lagrange-based decomposition coordination iterative algorithm in step S4 of Example 1, with an initial value for the penalty coefficient. Convergence threshold The maximum number of iterations is 800, and the convergence process is as follows: Figure 3 As shown, after 550 iterations, the residual curve gradually converges. The optimized global scheduling parameters ( , The active power and reactive power corrections (respectively) and the task set are shown in Tables 3 and 4, respectively.

[0087] Table 3

[0088] Table 4

[0089] 3) Closed-loop correction Five minutes after the initial execution of the regional task set, real-time measurements were performed on the system to obtain voltage data for key nodes. The voltage at node 9 was 0.989 pu, and the voltage at node 10 was 0.986 pu, still showing a deviation of 0.002 pu compared to the reference voltage of 0.988 pu. Based on this deviation, the closed-loop correction process described in S5 was triggered, the operating state residual vector was calculated, and the sensitivity coefficient of voltage to active power injection was obtained online. After linearization analysis, the sensitivity of the local active power injection at node 10 to its voltage was approximately 1.0 × 10⁻⁶. -4 The voltage coupling sensitivity of node 9 to node 10 is approximately 0.4 × 10 kW. -4 Based on this sensitivity relationship, the power correction required for node 10 is approximately 20 kW. If local resources are insufficient, node 9 will increase the discharge power by approximately 50 kW to achieve equivalent regulation.

[0090] In this example, the electric vehicle load at node 10 still has a 50 kW interruptibility margin, so the correction can be achieved simply by adding a 20 kW charge limiting amount at node 10. In the corrected global scheduling parameters, the active power scheduling amount at node 10 is updated from -130 kW to -150 kW, while the constraint boundary remains unchanged. After receiving the update task, the regional control unit executes the new charge limiting command within the remaining 55 minutes.

[0091] Measurements after the second execution showed that the voltage at node 10 rose to 0.9885 pu, and the voltage at node 9 was 0.990 pu. The overall system entered the preset tolerance range, and both the original residual and the dual residual were below the set threshold. The iterative closed loop terminated at this point. This demonstrates that the dynamic correction mechanism described in this invention can achieve rapid adaptive adjustment of the distribution network's operating state without reconstructing the global model, correcting the steady-state deviation of the terminal voltage from the initial -0.002 pu to within 0.0005 pu. This further verifies the stability and efficiency of the method in multi-timescale collaborative control.

[0092] 4) Results Comparison and Analysis To verify the effectiveness of the proposed method, the method of this invention (hereinafter referred to as the "hierarchical collaborative method") is compared and analyzed with the traditional centralized scheduling method (hereinafter referred to as the "centralized scheduling method"). The main indicators for comparison include: node voltage deviation; system active power loss; flexibility resource utilization rate; and optimized convergence speed. The calculation results are shown in Table 5.

[0093] Table 5

[0094] As shown in Table 5, the method described in this invention significantly outperforms traditional methods in all performance indicators. Specifically, node voltage deviation is effectively suppressed through hierarchical optimization and dynamic correction mechanisms; system losses are significantly reduced due to a more rational allocation of flexibility resources in the spatiotemporal dimensions; and the augmented Lagrange decomposition algorithm significantly improves iterative convergence efficiency.

[0095] Furthermore, Figure 4 The invention demonstrates a comparison of node voltage distribution in scenario A (peak load period). After adopting the method of this invention, the voltage drop problem of nodes 7–10 is significantly alleviated, and the voltage deviation is controlled within ±1.5%, while the voltage deviation of some nodes exceeds 4% under the traditional method.

[0096] Example 3, Figure 5 A hierarchical collaborative control system for intelligent distribution networks oriented towards multi-source load response is presented, including: The demand calculation module is used to calculate the power deviation and flexibility requirements of each node according to the type of distribution network operation scenario, and obtain the corresponding load response demand indicators. The resource allocation module is used to obtain the status information of the distribution network's flexibility resources. It combines the historical uncertainty and real-time fluctuation characteristics of resource response to introduce a dynamic safety margin coefficient to correct the adjustable capacity boundary of resources and construct a distributed resource model. The collaborative optimization module is used to construct a two-layer decoupled scheduling model that includes a global objective function and regional constraints based on the load response demand index and the distributed resource model. The collaborative control module is used to iteratively solve the two-layer decoupled scheduling model to obtain global scheduling parameters and a set of execution tasks, and to send the execution tasks to the corresponding local control units; The feedback correction module is used to dynamically correct the operating status of the distribution network by adaptively adjusting the convergence step size based on the decay rate of the nonlinear state residual, updating the global scheduling parameters and regional constraints, and based on the feedback data executed by the local control unit.

[0097] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0098] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0099] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0102] In conclusion, 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 hierarchical collaborative control method for intelligent distribution networks oriented towards multi-source load response, characterized in that, include: Calculate node power deviation and flexibility requirements to obtain load response requirement indicators; A dynamic safety margin coefficient is introduced to correct the adjustable capacity boundary of flexible resources, and a distributed resource model is constructed. A two-layer decoupled scheduling model is constructed based on the aforementioned demand indicators and resource model; Solve the scheduling model to obtain global scheduling parameters and execution tasks, and then send the execution tasks to the local control unit; Based on the feedback data from the local control unit, the convergence step size is adjusted according to the state residual decay rate, the global scheduling parameters and regional constraints are updated, and the operating status of the distribution network is dynamically corrected. The construction of the distributed resource model includes: The initial adjustable capacity is obtained based on the flexible resource status information; The dynamic safety margin coefficient is calculated based on the real-time response variance of flexible resources, and the initial adjustable capacity upper limit is corrected to obtain the actual adjustable capacity. The actual adjustable capacity is adjusted by combining the response speed, adjustment method, control method and electrical characteristics of the access nodes of the flexible resources, and the spatial distribution of the resources is adjusted according to the power grid topology information and the electrical characteristics of the nodes to obtain a distributed resource model; The calculation of the dynamic safety margin coefficient includes: Extract the historical deviation between the actual response power and the planned power of the flexibility resources within a preset time window, and calculate the real-time response variance. Based on the real-time response variance and the system fault tolerance threshold, a negatively correlated nonlinear decay function is constructed to calculate the dynamic safety margin coefficient. The dynamic safety margin coefficient is specifically formulated as follows: in, This is the dynamic safety margin coefficient, with a value range of (0,1]. The preset system fault tolerance threshold characterizes the boundary of uncertainty that the system can absorb; This is a decay adjustment factor that controls the decay rate. This represents the real-time response variance of resource r at the current time t.

2. The method according to claim 1, characterized in that, Before considering the power deviation and flexibility requirements of the computing nodes, the method also includes: identifying scenarios from real-time operating data of the distribution network to determine the type of operating scenario.

3. The method according to claim 1, characterized in that, The load response demand index is obtained by weighted normalization calculation of the node power deviation and flexibility demand; wherein the weighting coefficient is determined according to the importance of the region where the node is located and the historical load fluctuation characteristics.

4. The method according to claim 1, characterized in that, The two-layer decoupled scheduling model takes minimizing load fluctuation, scheduling cost, energy efficiency loss and maximizing scheduling efficiency as its joint objective function; The regional constraints include node power deviation constraints, node voltage constraints, and maximum adjustable capacity constraints of resources.

5. The method according to claim 1, characterized in that, The scheduling model is solved using a decomposition-coordination iterative algorithm based on augmented Lagrange; wherein in each iteration round: The global optimization layer solves the global objective function to update the global scheduling parameters under the condition of fixed regional equivalent power injection and boundary coupling variables; The regional optimization layer solves for the execution task based on the global scheduling parameters under the regional constraints. Based on the residuals of the coupling constraints between the global and regional systems, the multipliers and penalty coefficients are updated and the next iteration is initiated until the residuals are less than a preset threshold or the maximum number of iterations is reached. The augmented Lagrange decomposition coordination iterative algorithm includes the augmented Lagrange function, as shown in the following formula: in, To augment the Lagrange function; The original objective function is... This is a coupling constraint between the global and regional levels. This is a global-region coupling mapping matrix, whose elements represent the correspondence between global scheduling parameters and regional scheduling variables, where x and z are global and regional variables, respectively. Let Lagrange multiplier vectors be used. To broaden the range of penalties, This is the penalty coefficient.

6. The method according to claim 1, characterized in that, The dynamic correction of the distribution network operating status includes: Real-time measurement data returned by the local control unit is collected, and a sensitivity matrix is ​​established by combining it with power grid topology information; Calculate the state residual vector based on the real-time measurement data and the global scheduling parameters; The convergence step size is adaptively adjusted based on the norm change rate of the state residual vectors in adjacent rounds. The power correction vector is obtained based on the sensitivity matrix; Based on the power correction vector and the adjusted convergence step size, the global scheduling parameters and the regional constraints are updated and re-solved to obtain the corrected execution task and issue it until the norm of the state residual vector is lower than the preset threshold.

7. The method according to claim 6, characterized in that, The adaptive adjustment of the convergence step size includes: Compare the norm of the current state residual vector with that of the previous round, and calculate the state residual decay rate; When the state residual decay rate is less than zero, it is determined that the system deviates from the equilibrium point, and the convergence step size is reduced according to a preset ratio. The formula for calculating the state residual decay rate is as follows: in, The attenuation rate is... and These are the state residual vectors for kth and k-1th iterations, respectively; The L2 norm of a vector is used to represent the vector's L2 norm.

8. A hierarchical collaborative control system for intelligent distribution networks oriented towards multi-source load response, characterized in that, include: The demand calculation module is used to calculate node power deviation and flexibility requirements to obtain load response demand indicators. The resource allocation module is used to introduce a dynamic safety margin coefficient to correct the adjustable capacity boundary of flexible resources and build a distributed resource model. The collaborative optimization module is used to construct a two-layer decoupled scheduling model based on the aforementioned demand indicators and resource model; The collaborative control module is used to solve the scheduling model, obtain global scheduling parameters and execution tasks, and send the execution tasks to the local control unit; The feedback correction module is used to adjust the convergence step size based on the state residual decay rate according to the feedback data of the local control unit, update the global scheduling parameters and regional constraints, and dynamically correct the operation status of the distribution network. The construction of the distributed resource model includes: The initial adjustable capacity is obtained based on the flexible resource status information; The dynamic safety margin coefficient is calculated based on the real-time response variance of flexible resources, and the initial adjustable capacity upper limit is corrected to obtain the actual adjustable capacity. The actual adjustable capacity is adjusted by combining the response speed, adjustment method, control method and electrical characteristics of the access nodes of the flexible resources, and the spatial distribution of the resources is adjusted according to the power grid topology information and the electrical characteristics of the nodes to obtain a distributed resource model; The calculation of the dynamic safety margin coefficient includes: Extract the historical deviation between the actual response power and the planned power of the flexibility resources within a preset time window, and calculate the real-time response variance. Based on the real-time response variance and the system fault tolerance threshold, a negatively correlated nonlinear decay function is constructed to calculate the dynamic safety margin coefficient. The dynamic safety margin coefficient is specifically formulated as follows: in, This is the dynamic safety margin coefficient, with a value range of (0,1]. The preset system fault tolerance threshold characterizes the boundary of uncertainty that the system can absorb; This is a decay adjustment factor that controls the decay rate. This represents the real-time response variance of resource r at the current time t.

Images