Intelligent regulation and control method and system for new energy grid-connected inverter group

Abstract

The invention is suitable for the technical field of new energy grid connection, and provides an intelligent regulation and control method and system for a new energy grid-connected inverter group, and the method comprises the following steps: obtaining the measurement data and prediction data of each node of a power distribution network, carrying out the absorption capability evaluation, and generating a dynamic absorption capability distribution diagram; constructing a self-adaptive multi-objective optimization function based on the absorption situation according to the dynamic absorption capacity distribution diagram; optimizing the network topology structure of the power distribution network to obtain an optimal network topology structure; based on the optimal network topology structure, generating an inverter group optimization control instruction according to the adaptive multi-objective optimization function; and regulating and controlling the new energy grid-connected inverter group according to the inverter group optimization control instruction. According to the method, the safe bearing capacity, the economic operation level and the real-time consumption rate of renewable energy sources of new energy grid connection are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of new energy grid connection technology, and in particular relates to an intelligent control method and system for a new energy grid-connected inverter group. Background Technology

[0002] Currently, the penetration rate of distributed renewable energy sources, represented by photovoltaics and wind power, in power distribution networks is increasing rapidly. These renewable energy sources connected to inverter interfaces are highly intermittent and volatile, and their large-scale grid connection poses multiple challenges to the safe, stable, and economical operation of the power system, mainly in the following aspects:

[0003] I. Inverters inherently possess the ability to rapidly adjust active and reactive power output, serving various control objectives such as suppressing voltage deviations, reducing network losses, increasing renewable energy absorption, and providing low-voltage / high-voltage ride-through support. However, these objectives often conflict with each other. For example, during voltage dips, providing reactive power support to restore voltage may require reducing active power output, which contradicts the goal of maximizing energy absorption. Existing strategies mostly target single or pre-defined scenarios, lacking the ability to adaptively coordinate multiple objectives based on the real-time state of the power grid.

[0004] Second, existing technologies mainly focus on optimization under given boundary conditions, lacking dynamic and accurate assessment of the real-time absorption capacity of the distribution network. The formulation of control instructions is often based on historical data or simple rules, resulting in control that is either too conservative (unutilizing the absorption capacity) or too aggressive (causing safety problems).

[0005] Therefore, there is an urgent need for an intelligent control method that can sense the grid absorption status in real time and adaptively coordinate multiple control objectives to improve the safety and absorption level of the new energy distribution network. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent control method for a new energy grid-connected inverter group, aiming to solve the above-mentioned technical problems.

[0007] This invention is implemented as follows: an intelligent control method for a new energy grid-connected inverter group, comprising the following steps:

[0008] The system acquires measurement and forecast data from each node of the distribution network, assesses its absorption capacity, and generates a dynamic absorption capacity distribution map. This dynamic absorption capacity distribution map describes the spatiotemporal distribution of renewable energy power that the distribution network can accommodate.

[0009] Based on the dynamic absorption capacity distribution map, an adaptive multi-objective optimization function based on the absorption status is constructed;

[0010] The optimal network topology of the distribution network is obtained by optimizing the network topology.

[0011] Based on the optimal network topology, an inverter group optimization control command is generated according to the adaptive multi-objective optimization function.

[0012] The inverter group is regulated and controlled according to the inverter group optimization control command.

[0013] Further steps include acquiring measurement and forecast data from each node of the distribution network, assessing its absorption capacity, and generating a dynamic absorption capacity distribution map. These steps specifically include:

[0014] The system collects measurement data from each node of the distribution network in real time and obtains the predicted data of the distribution network. The measurement data includes node voltage, branch current, active power and reactive power. The predicted data includes future renewable energy power prediction data and load prediction data.

[0015] Using real-time collected measurement data as the initial state, combined with predicted data, the baseline operating state of the current distribution network is solved based on the online power flow calculation method.

[0016] Based on the current baseline operating state of the distribution network, the maximum active power injection that each node can increase within a future preset time period is calculated, and the maximum active power injection increment is used as the active power dynamic absorption capacity of the node.

[0017] Based on the current baseline operating state of the distribution network, calculate the maximum reactive power support range that each node can provide while maintaining its current active power output, and use this range as the node's dynamic reactive power absorption capacity.

[0018] Aggregate the active power dynamic absorption capacity and reactive power dynamic absorption capacity of all nodes to generate a dynamic absorption capacity distribution map.

[0019] Furthermore, based on the dynamic absorption capacity distribution map, the steps for constructing an adaptive multi-objective optimization function based on the absorption situation specifically include:

[0020] Based on the dynamic absorption capacity distribution map, the set of objectives to be optimized is determined; the set of objectives includes the following objective functions: the absorption increment maximization objective function with the goal of maximizing the increase in the total active power output of the inverter group, the network loss minimization objective function with the goal of minimizing the total active power loss of the distribution network, and the node voltage deviation minimization objective function with the goal of minimizing the sum of squares of voltage deviations of all nodes.

[0021] Adaptive weight adjustment is performed on each objective function in the objective set;

[0022] The adaptive multi-objective optimization function at the current time is constructed by linearly weighting and summing the objective functions after adaptive weight adjustment.

[0023] Furthermore, the step of adaptively adjusting the weights of each objective function in the objective set specifically includes:

[0024] Based on the absorption capacity margin of each node in the dynamic absorption capacity distribution map, the weight coefficients of the objective function for maximizing the absorption increment are dynamically adjusted.

[0025] The weighting coefficients of the objective function for minimizing node voltage deviation are dynamically adjusted based on the degree of deviation between the real-time voltage and the rated voltage of each node.

[0026] Furthermore, the steps to optimize the distribution network topology to obtain the optimal network topology include:

[0027] At the start of a preset time scale, predictive data for several future cycles and the current topology of the distribution network are acquired as inputs to the state space.

[0028] All feasible switch combinations that satisfy the radial operation constraints of the distribution network are mapped to an action space represented by a spanning tree.

[0029] The distribution network reconfiguration problem is modeled as a Markov decision process;

[0030] Based on a pre-trained deep reinforcement learning model, the Markov decision process is solved, and the optimal network topology with the lowest expected overall cost within the current time scale period is output from the action space according to the input in the state space.

[0031] Furthermore, the steps for generating inverter group optimization control commands based on the optimal network topology and the adaptive multi-objective optimization function specifically include:

[0032] Based on the optimal network topology, an adaptive multi-objective optimization function is used as the optimization objective to establish a complete optimization model with constraints. The constraints include power flow balance constraints, upper and lower limits of active and reactive power output of inverters, node voltage safety constraints, and line capacity constraints.

[0033] Solving the complete optimization model yields a set of baseline control commands that satisfy all constraints and optimize the adaptive multi-objective optimization function.

[0034] Based on current measurement and forecast data, the baseline control instruction set is modified to generate optimized control instructions for the inverter group.

[0035] Another objective of this invention is to provide an intelligent control system for a new energy grid-connected inverter group, used to implement the above-mentioned intelligent control method, comprising:

[0036] The absorption capacity assessment module is used to acquire measurement and prediction data of each node in the distribution network, and to assess the absorption capacity and generate a dynamic absorption capacity distribution map; the dynamic absorption capacity distribution map is used to describe the spatiotemporal distribution of renewable energy power that the distribution network can accept.

[0037] The objective function construction module is used to construct an adaptive multi-objective optimization function based on the dynamic absorption capacity distribution map.

[0038] The network topology optimization module is used to optimize the network topology of the distribution network to obtain the optimal network topology.

[0039] The control command generation module is used to generate inverter group optimization control commands based on the optimal network topology and an adaptive multi-objective optimization function.

[0040] The instruction execution module is used to regulate the grid-connected inverter group of new energy sources according to the optimized control instructions of the inverter group.

[0041] Furthermore, the absorption capacity assessment module specifically includes:

[0042] The data acquisition unit is used to collect measurement data from each node of the distribution network in real time and obtain the predicted data of the distribution network; the measurement data includes node voltage, branch current, active power and reactive power; the predicted data includes future renewable energy power prediction data and load prediction data.

[0043] The baseline operating state determination unit is used to solve the current baseline operating state of the distribution network based on the online power flow calculation method, using real-time collected measurement data as the initial state and combined with prediction data.

[0044] The active power absorption determination unit is used to calculate the maximum active power injection that each node can increase within a future preset time period based on the current baseline operating state of the distribution network, and obtain the maximum active power injection increment as the active power dynamic absorption capacity of the node.

[0045] The reactive power absorption determination unit is used to calculate the maximum reactive power support range that each node can provide while maintaining the current active power output unchanged, based on the current baseline operating state of the distribution network, as the reactive power dynamic absorption capacity of that node.

[0046] The active power absorption capacity determination unit aggregates the active power absorption capacity and reactive power absorption capacity of all nodes to generate a dynamic absorption capacity distribution map.

[0047] Furthermore, the objective function construction module specifically includes:

[0048] The target set determination unit is used to determine the target set to be optimized based on the dynamic absorption capacity distribution map. The target set includes the following objective functions: the absorption increment maximization objective function with the goal of maximizing the increase in the total active power output of the inverter group; the network loss minimization objective function with the goal of minimizing the total active power loss of the distribution network; and the node voltage deviation minimization objective function with the goal of minimizing the sum of squares of voltage deviations of all nodes.

[0049] The weight adaptive adjustment unit is used to adaptively adjust the weights of each objective function in the objective set.

[0050] The objective function fusion unit is used to linearly weight and sum the objective functions after adaptive weight adjustment to construct the adaptive multi-objective optimization function at the current time.

[0051] Furthermore, the network topology optimization module specifically includes:

[0052] The state space determination unit is used to acquire the predicted data for several future cycles and the current topology of the distribution network at the beginning of a cycle on a preset time scale, as input to the state space.

[0053] The action space determination unit is used to map all feasible switch combinations that satisfy the radial operation constraints of the distribution network into an action space represented by a spanning tree.

[0054] The distribution network reconfiguration modeling unit is used to model the distribution network reconfiguration problem as a Markov decision process.

[0055] The optimization output unit is used to solve the Markov decision process based on the pre-trained deep reinforcement learning model. Based on the state space input, it outputs the optimal network topology with the lowest expected comprehensive cost within the current time scale period from the action space.

[0056] The control command generation module specifically includes:

[0057] The complete optimization model construction unit is used to establish a complete optimization model with constraints based on the optimal network topology and with an adaptive multi-objective optimization function as the optimization objective. The constraints include power flow balance constraints, upper and lower limits of inverter active and reactive power output constraints, node voltage safety constraints, and line capacity constraints.

[0058] The baseline control command determination unit is used to solve the complete optimization model to obtain a baseline control command set that satisfies all constraints and makes the adaptive multi-objective optimization function optimal.

[0059] The control command correction unit is used to correct the baseline control command set based on the current measurement data and prediction data, and generate optimized control commands for the inverter group.

[0060] This invention provides an intelligent control method for a cluster of inverters connected to the grid for new energy sources. It dynamically quantifies and assesses the real-time absorption capacity of the power grid and generates a dynamic absorption capacity distribution map, providing a precise and adaptive control boundary for subsequent optimization. Furthermore, this invention constructs an adaptive multi-objective optimization function based on the absorption status, achieving optimal collaborative optimization of absorption, network losses, and voltage targets under different operating scenarios. While ensuring voltage safety, it effectively reduces distribution network losses and improves the overall control efficiency and economy of the inverter cluster. This method significantly improves the safe carrying capacity, economic operation level, and real-time absorption rate of renewable energy connected to the grid. Attached Figure Description

[0061] Figure 1 A flowchart illustrating the intelligent control method for a new energy grid-connected inverter group provided in an embodiment of the present invention.

[0062] Figure 2 This is a flowchart illustrating step S100 in the intelligent control method for a new energy grid-connected inverter group provided in an embodiment of the present invention.

[0063] Figure 3 This is a flowchart illustrating step S200 in the intelligent control method for a new energy grid-connected inverter group provided in an embodiment of the present invention.

[0064] Figure 4 This is a flowchart illustrating step S300 in the intelligent control method for a new energy grid-connected inverter group provided in an embodiment of the present invention.

[0065] Figure 5 This is a flowchart illustrating step S400 in the intelligent control method for a new energy grid-connected inverter group provided in an embodiment of the present invention.

[0066] Figure 6 A schematic diagram of the structure of the intelligent control system for a new energy grid-connected inverter group provided in an embodiment of the present invention.

[0067] Figure 7 This is a schematic diagram of the absorption capacity assessment module provided in an embodiment of the present invention.

[0068] Figure 8 This is a schematic diagram of the objective function construction module provided in an embodiment of the present invention.

[0069] Figure 9 This is a schematic diagram of the network topology optimization module provided in an embodiment of the present invention.

[0070] Figure 10 This is a schematic diagram of the control instruction generation module provided in an embodiment of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0072] like Figure 1 As shown, in one embodiment of the present invention, an intelligent control method for a new energy grid-connected inverter group is provided, comprising the following steps:

[0073] S100. Acquire the measurement data and prediction data of each node in the distribution network, and conduct a power absorption capacity assessment to generate a dynamic power absorption capacity distribution map; the dynamic power absorption capacity distribution map is used to describe the spatiotemporal distribution of renewable energy power that the distribution network can accept.

[0074] S200. Based on the dynamic absorption capacity distribution map, construct an adaptive multi-objective optimization function based on the absorption situation;

[0075] S300: Optimize the network topology of the distribution network to obtain the optimal network topology.

[0076] S400: Based on the optimal network topology, it generates inverter group optimization control commands according to the adaptive multi-objective optimization function;

[0077] S500 regulates the grid-connected inverter group of new energy sources according to the inverter group optimization control command.

[0078] like Figure 2 As shown, in a preferred embodiment of the present invention, the step of acquiring measurement data and prediction data of each node in the distribution network, assessing the absorption capacity, and generating a dynamic absorption capacity distribution map, namely step S100, specifically includes:

[0079] S110. Real-time acquisition of measurement data from each node of the distribution network and acquisition of forecast data for the distribution network;

[0080] Specifically, measurement units deployed at key nodes of the distribution network (new energy grid connection) and new energy power plants collect measurement data from each node of the distribution network in real time. The measurement data includes at least node voltage (from voltage sensors), branch current (from current transformers), and active and reactive power of the nodes (from inverter local controllers or grid connection point meters). It should be noted that a node refers to a connection point with the same potential in the distribution network topology, and each node may include equipment such as inverters, loads, and capacitors.

[0081] In addition, the forecast data includes at least the power forecast data and load forecast data of new energy sources within a preset time period (e.g., 1-20 minutes). It should be noted that the forecast data can be obtained directly by using existing machine learning models, or it can be obtained directly from existing power forecasting systems deployed in the cloud or locally through standard data interfaces (e.g., IEC 104, WebService). This is not an improvement of the present invention, so it will not be elaborated here.

[0082] S120. Using real-time collected measurement data as the initial state and combined with predicted data, the baseline operating state of the current distribution network is solved based on the online power flow calculation method. Specifically, the existing Newton-Raphson method can be used for online power flow calculation.

[0083] S130. Based on the current baseline operating state of the distribution network, the continuous power flow method or linearized sensitivity analysis method is used to calculate the maximum active power injection that each node can increase within a future preset time period under the constraints of all node voltage and line capacity. The maximum active power injection increment is used as the active power dynamic absorption capacity of the node.

[0084] Specifically, under the baseline operating conditions, the active power injection of a specific node is gradually increased in a preset step size, and the power flow calculation is re-performed until the voltage of any node or the power of the line exceeds the limit for the first time. At this time, the total active power added to the node is the active power dynamic absorption capacity of the node.

[0085] S140. Based on the current baseline operating state of the distribution network, calculate the maximum reactive power support range that each node can provide while maintaining its current active power output unchanged, and use this range as the node's dynamic reactive power absorption capacity.

[0086] S150. Aggregate the active power dynamic absorption capacity and reactive power dynamic absorption capacity of all nodes to generate a dynamic absorption capacity distribution map; where the dynamic absorption capacity distribution map is actually a data structure that includes all nodes and their corresponding maximum active power injection power increment and maximum reactive power support power range.

[0087] In this embodiment of the invention, a quantitative dynamic absorption capacity distribution map is generated by fusing real-time measurement data and predicted data and using the continuous power flow method. This dynamic absorption capacity distribution map accurately reveals the spatiotemporal distribution and limits of renewable energy power that each node can accept under safety constraints, providing key and forward-looking boundary conditions and decision-making basis for subsequent intelligent regulation, fundamentally changing the limitations of previous regulation that relied on static and conservative thresholds.

[0088] like Figure 3As shown, in a preferred embodiment of the present invention, the step of constructing an adaptive multi-objective optimization function based on the absorption capacity distribution map, i.e., step S200, specifically includes:

[0089] S210. Based on the dynamic absorption capacity distribution map, determine the target set to be optimized; the target set includes the following objective functions: the absorption increment maximization objective function with the goal of maximizing the increase in the total active power output of the inverter group, the network loss minimization objective function with the goal of minimizing the total active power loss of the distribution network, and the node voltage deviation minimization objective function with the goal of minimizing the sum of squares of voltage deviations of all nodes.

[0090] Specifically, the expression for the objective function f1, which maximizes the absorption increment, is:

[0091] ;

[0092] In the formula, N is the total number of inverters in the inverter group; The future active power reference value for inverter i to be optimized; Let be the active power of inverter i at time t.

[0093] The expression for the objective function f2, which minimizes network loss, is:

[0094] ;

[0095] In the formula, B is the set of all branches in the distribution network; g mn V is the conductance of branch mn; m V n These are the voltage amplitudes across branch mn at the current moment; Let m be the phase angle difference between the two ends of branch mn;

[0096] The expression for the objective function f3, which minimizes the node voltage deviation, is:

[0097] ;

[0098] In the formula, M is the total number of nodes in the distribution network; V j This refers to the real-time voltage, i.e., the voltage at the current node. j The voltage amplitude; Vr is the rated voltage.

[0099] S220. Adaptively adjust the weights of each objective function in the objective set; specifically as follows:

[0100] Based on the absorption capacity margin of each node in the dynamic absorption capacity distribution map, the weight coefficients of the objective function for maximizing the absorption increment are dynamically adjusted; when the absorption capacity margin of a node is higher than the first preset threshold, the weight coefficients of the objective function for maximizing the absorption increment are increased.

[0101] Based on the degree of deviation between the real-time voltage and the rated voltage of each node, the weight coefficient of the objective function for minimizing the node voltage deviation is dynamically adjusted; when the absolute value of the voltage deviation of a node is higher than the second preset threshold, the weight coefficient of the objective function for minimizing the node voltage deviation is increased.

[0102] In practical applications, the dynamic adjustment of the weight coefficients can be achieved through a fuzzy logic controller. The input of the fuzzy logic controller is the fuzzy quantity of the node absorption capacity margin and voltage deviation, and the output is the adjustment factor of the weight coefficient of each sub-objective function.

[0103] S230. The objective functions after adaptive weight adjustment are linearly weighted and summed to construct the adaptive multi-objective optimization function at the current time.

[0104] Specifically, the expression for the adaptive multi-objective optimization function F is:

[0105] ;

[0106] In the formula, w1, w2, and w3 are weight coefficients, which are adaptively adjusted according to the weights and satisfy the following expression:

[0107] ;

[0108] ;

[0109] ;

[0110] In the formula, The maximum active power injection increment for node j; The rated capacity of node j; The first preset threshold, i.e. the absorption margin threshold, can be set by the scheduling procedure, and is generally 0.1-0.3; The second preset threshold, i.e. the voltage over-limit threshold, is generally set to 70%-90% of the difference between the upper limit of voltage safety (from the power grid regulations) and the rated voltage; k1 and k2 are sensitivity coefficients, which can be determined by fitting historical data. Here, x is the sigmoid function. .

[0111] Traditional multi-objective optimization weights are usually fixed or preset manually, making it difficult to adapt to real-time changes in the power grid state. However, in this embodiment of the invention, by introducing a weight adaptive mechanism, the weight coefficients of targets such as absorption capacity, grid loss, and voltage can be dynamically adjusted based on the dynamic absorption capacity distribution map and real-time voltage information. This enables the control targets to intelligently match the immediate needs of the power grid: prioritizing absorption when the safety margin is large and prioritizing stabilization when the voltage is critical, thereby achieving global optimization and real-time precision of the control strategy.

[0112] like Figure 4 As shown, in a preferred embodiment of the present invention, the step of optimizing the network topology of the distribution network to obtain the optimal network topology, namely step S300, specifically includes:

[0113] S310. At the start of a preset slow time scale (e.g., 10-20 minutes), obtain the prediction data (new energy power prediction data and load prediction data) for several future cycles, as well as the current topology of the distribution network, as the state space input.

[0114] S320: Map all feasible switch combinations that satisfy the radial operation constraints of the distribution network to an action space represented by a spanning tree.

[0115] S330. The distribution network reconfiguration problem is modeled as a Markov decision process. In the Markov decision process, the state space transition is determined by the power flow equation and the result of the action execution. The reward function is the comprehensive cost considering network loss, voltage quality and estimated absorption space after considering the action space.

[0116] S340. Based on a pre-trained deep reinforcement learning model, the Markov decision process is solved. According to the input in the state space, the optimal network topology with the lowest expected comprehensive cost in the current slow time scale period is output from the action space. This optimal network topology is used to improve the overall absorption space and reduce the expected network loss.

[0117] In practical applications, the aforementioned deep reinforcement learning model can adopt an Actor-Critic architecture to handle continuous or high-dimensional discrete action spaces, which mainly includes:

[0118] Feature extraction network: Input the original state space, the feature extraction network is usually 2-3 fully connected layers, or combined with one-dimensional convolutional layers to process time series prediction data and output high-dimensional feature vectors;

[0119] Policy Network: Input a high-dimensional feature vector and output an action probability distribution; for discrete actions (switch combinations), output the probability of each possible action (spanning tree) to form a probability vector; for parameterized actions (such as simultaneously outputting switch actions and expected action durations), output the action type and the mean and variance of the corresponding parameters.

[0120] State-Value Network: Input a high-dimensional feature vector and output a state value scalar to evaluate the long-term expected return of the current state space, which is used to measure the quality of the state space.

[0121] Target network: As a copy of the policy network and state-value network, it is used to provide stable target values ​​and solve the problem of model training instability.

[0122] In addition, this deep reinforcement learning model can be trained using existing proximal policy optimization or soft actor-critic algorithms.

[0123] In this embodiment of the invention, the topology (switch combination) of the distribution network is intelligently reconstructed on a slow time scale through a deep reinforcement learning algorithm. This is equivalent to optimizing the network topology of power flow distribution at the source. From a global perspective, it actively creates a lower-level structure that is more conducive to the access of new energy sources, reduces network losses and improves voltage, laying an efficient and safe foundation for subsequent rapid regulation and control, and significantly improving the long-term carrying capacity of the distribution network.

[0124] like Figure 5 As shown, in a preferred embodiment of the present invention, the step of generating inverter group optimization control commands based on the optimal network topology and according to the adaptive multi-objective optimization function, i.e., step S400, specifically includes:

[0125] S410. Based on the optimal network topology, and with the adaptive multi-objective optimization function as the optimization objective, a complete optimization model containing constraints is established. The constraints include power flow balance (such as DistFlow power flow model) constraints, upper and lower limits of active and reactive power output of inverters, node voltage safety constraints, and line capacity constraints.

[0126] S420. Solve the complete optimization model to obtain a set of baseline control instructions that satisfy all constraints and make the adaptive multi-objective optimization function optimal.

[0127] Specifically, based on the second-order cone relaxation algorithm, the non-convex power flow constraints in the complete optimization model are made convex, transforming them into a fast-solvable second-order cone programming problem. The second-order cone programming problem is then solved using an interior-point solver, resulting in a set of reference control instructions that satisfy all constraints and optimize the adaptive multi-objective optimization function. The reference control instruction set includes the initial active power reference value and reactive power reference value for each inverter.

[0128] S430. Based on the current measurement data and prediction data, the reference control instruction set is corrected to generate inverter group optimized control instructions.

[0129] Specifically, the real-time measurement data, prediction data, and reference control command set at the current moment are used as inputs. The output of the correction amount of the reference control command is generated through a lightweight neural network compensator that can be updated online, thus generating the final inverter group optimized control command. The inverter group optimized control command includes active power reference value, reactive power reference value, and droop control parameters.

[0130] In practical applications, lightweight neural network compensators can employ fully connected feedforward neural networks, including input, hidden, and output layers. The input layer receives real-time measurement data, predicted data, and key feature vectors from the baseline control command set. The dimension is typically controlled between 20 and 50. Examples of input vectors include the baseline control command, the difference between real-time voltage measurements and predicted values ​​at key nodes, and the fluctuating gradients of short-term (1-5 minutes) renewable energy power predictions. The hidden layer consists of one to two layers, with 16 to 64 neurons per layer, using ReLU or Tanh activation functions. The output layer outputs the correction amount for the baseline control command, using a linear activation function. Furthermore, the lightweight neural network compensator can be trained and updated using an online gradient descent algorithm, with the loss function used during training being a mean squared error loss, among others.

[0131] Under the aforementioned optimal network topology, this embodiment of the invention employs second-order cone programming to convexize the large-scale, non-convex inverter control problem, achieving millisecond-level fast solutions to complex optimization problems and ensuring real-time control. Simultaneously, the introduction of a lightweight neural network compensator can correct command deviations caused by model simplification or ultra-short-term fluctuations online, thereby significantly improving the robustness of inverter group control to uncertainties and the final control accuracy while ensuring computational efficiency.

[0132] In a preferred embodiment of the present invention, the step of regulating the grid-connected inverter group of new energy sources according to the inverter group optimization control command, namely step S500, specifically includes:

[0133] The inverter group optimization control command is sent to the local controller of each renewable energy grid-connected inverter through the communication network. Each local controller sets the corresponding active and reactive power reference values ​​and droop control parameters in the inverter group optimization control command as the target setpoint of the local controller. During operation, each inverter monitors the voltage of its renewable energy grid connection point in real time, and dynamically fine-tunes the active and reactive power output of each inverter based on the deviation between the actual voltage and the target voltage, as well as the target setpoint, according to the adaptive droop control algorithm. In addition, through the distributed collaborative execution of the local controllers of all inverters, the actual operating state of the distribution network can be made closer to the solution result of the above complete optimization model, and finally, the inverter group can achieve multi-objective collaborative optimization operation under the premise of ensuring grid security.

[0134] In this embodiment of the invention, the inverter group optimization control command generated by centralized optimization is sent to the local controller of each inverter. Each inverter, in its local controller, combines the inverter group optimization control command with real-time measurement data to execute adaptive droop control. This control method not only ensures the achievement of the global optimization goal but also endows it with local autonomy in abnormal situations such as communication interruptions. Furthermore, it can form a closed loop through local feedback, enhancing the dynamic response speed and operational reliability of the entire system.

[0135] like Figure 6 As shown, in another embodiment of the present invention, an intelligent control system for a new energy grid-connected inverter group is also provided to implement the above-mentioned intelligent control method, comprising:

[0136] The absorption capacity assessment module 10 is used to acquire the measurement data and prediction data of each node of the distribution network, and to perform absorption capacity assessment to generate a dynamic absorption capacity distribution map; the dynamic absorption capacity distribution map is used to describe the spatiotemporal distribution of renewable energy power that the distribution network can accept.

[0137] The objective function construction module 20 is used to construct an adaptive multi-objective optimization function based on the dynamic absorption capacity distribution map.

[0138] The network topology optimization module 30 is used to optimize the network topology of the distribution network to obtain the optimal network topology.

[0139] The control command generation module 40 is used to generate inverter group optimization control commands based on the optimal network topology and an adaptive multi-objective optimization function.

[0140] The instruction execution module 50 is used to regulate the grid-connected inverter group of new energy according to the inverter group optimization control instructions.

[0141] like Figure 7 As shown, in a preferred embodiment of the present invention, the absorption capacity assessment module 10 specifically includes:

[0142] The data acquisition unit 11 is used to collect the measurement data of each node of the distribution network in real time and obtain the prediction data of the distribution network; the measurement data includes node voltage, branch current, active power and reactive power; the prediction data includes future renewable energy power prediction data and load prediction data.

[0143] The baseline operating state determination unit 12 is used to solve the baseline operating state of the current distribution network based on the online power flow calculation method, using the real-time collected measurement data as the initial state and combined with the predicted data.

[0144] The active power absorption determination unit 13 is used to calculate the maximum active power injection that each node can increase within a future preset scale period based on the current baseline operating state of the distribution network, and obtain the maximum active power injection increment as the active power dynamic absorption capacity of the node.

[0145] The reactive power absorption determination unit 14 is used to calculate the maximum reactive power support range that each node can provide while maintaining the current active power output unchanged, based on the current baseline operating state of the distribution network, as the reactive power dynamic absorption capacity of the node.

[0146] Unit 15, which determines the absorption capacity, aggregates the active power dynamic absorption capacity and reactive power dynamic absorption capacity of all nodes to generate a dynamic absorption capacity distribution map.

[0147] like Figure 8 As shown, in a preferred embodiment of the present invention, the objective function construction module 20 specifically includes:

[0148] The target set determination unit 21 is used to determine the target set to be optimized based on the dynamic absorption capacity distribution map. The target set includes the following objective functions: the absorption increment maximization objective function with the goal of maximizing the increase in the total active power output of the inverter group, the network loss minimization objective function with the goal of minimizing the total active power loss of the distribution network, and the node voltage deviation minimization objective function with the goal of minimizing the sum of squares of voltage deviations of all nodes.

[0149] The weight adaptive adjustment unit 22 is used to adaptively adjust the weights of each objective function in the objective set;

[0150] Objective function fusion unit 23 is used to linearly weight and sum the objective functions after adaptive weight adjustment to construct the adaptive multi-objective optimization function at the current time.

[0151] like Figure 9 As shown, in a preferred embodiment of the present invention, the network topology optimization module 30 specifically includes:

[0152] The state space determination unit 31 is used to obtain the predicted data for several future cycles and the current topology of the distribution network at the beginning of a cycle on a preset time scale, as the state space input.

[0153] Action space determination unit 32 is used to map all feasible switch combinations that satisfy the radial operation constraints of the distribution network into an action space represented by a spanning tree;

[0154] Distribution network reconfiguration modeling unit 33 is used to model the distribution network reconfiguration problem as a Markov decision process;

[0155] The optimization result output unit 34 is used to solve the Markov decision process based on the pre-trained deep reinforcement learning model. According to the state space input, it outputs the optimal network topology with the lowest expected comprehensive cost within the current time scale period from the action space.

[0156] like Figure 10 As shown, in a preferred embodiment of the present invention, the control command generation module 40 specifically includes:

[0157] The complete optimization model construction unit 41 is used to establish a complete optimization model with constraints based on the optimal network topology and with an adaptive multi-objective optimization function as the optimization objective. The constraints include power flow balance constraints, upper and lower limits of active and reactive power output of inverters, node voltage safety constraints, and line capacity constraints.

[0158] The reference control instruction determination unit 42 is used to solve the complete optimization model to obtain a set of reference control instructions that satisfy all constraints and make the adaptive multi-objective optimization function optimal.

[0159] The control command correction unit 43 is used to correct the reference control command set based on the current measurement data and prediction data, and generate the inverter group optimized control command.

[0160] It should be noted that the above modules and units can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up the modules or units, enabling the processor to execute the various steps of the above method.

[0161] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0162] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0163] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for intelligent control of a new energy grid-connected inverter group, characterized in that, Includes the following steps: Acquire measurement and forecast data for each node of the distribution network, assess the absorption capacity, and generate a dynamic absorption capacity distribution map; The dynamic absorption capacity distribution map is used to describe the spatiotemporal distribution of renewable energy power that the distribution network can accept. Based on the dynamic absorption capacity distribution map, an adaptive multi-objective optimization function based on the absorption status is constructed; The optimal network topology of the distribution network is obtained by optimizing the network topology. Based on the optimal network topology, an inverter group optimization control command is generated according to the adaptive multi-objective optimization function. The inverter group is regulated and controlled according to the inverter group optimization control command.

2. The intelligent control method for a new energy grid-connected inverter group according to claim 1, characterized in that, The steps for acquiring measurement and forecast data from each node of the distribution network, assessing its absorption capacity, and generating a dynamic absorption capacity distribution map include: The system collects measurement data from each node of the distribution network in real time and obtains the predicted data of the distribution network. The measurement data includes node voltage, branch current, active power and reactive power. The predicted data includes future renewable energy power prediction data and load prediction data. Using real-time collected measurement data as the initial state, combined with predicted data, the baseline operating state of the current distribution network is solved based on the online power flow calculation method. Based on the current baseline operating state of the distribution network, the maximum active power injection that each node can increase within a future preset time period is calculated, and the maximum active power injection increment is used as the active power dynamic absorption capacity of the node. Based on the current baseline operating state of the distribution network, calculate the maximum reactive power support range that each node can provide while maintaining its current active power output, and use this range as the node's dynamic reactive power absorption capacity. Aggregate the active power dynamic absorption capacity and reactive power dynamic absorption capacity of all nodes to generate a dynamic absorption capacity distribution map.

3. The intelligent control method for a new energy grid-connected inverter group according to claim 2, characterized in that, Based on the dynamic absorption capacity distribution map, the steps for constructing an adaptive multi-objective optimization function based on the absorption situation include: Based on the dynamic absorption capacity distribution map, the set of objectives to be optimized is determined; the set of objectives includes the following objective functions: the absorption increment maximization objective function with the goal of maximizing the increase in the total active power output of the inverter group, the network loss minimization objective function with the goal of minimizing the total active power loss of the distribution network, and the node voltage deviation minimization objective function with the goal of minimizing the sum of squares of voltage deviations of all nodes. Adaptive weight adjustment is performed on each objective function in the objective set; The adaptive multi-objective optimization function at the current time is constructed by linearly weighting and summing the objective functions after adaptive weight adjustment.

4. The intelligent control method for a new energy grid-connected inverter group according to claim 3, characterized in that, The steps for adaptively adjusting the weights of each objective function in the objective set specifically include: Based on the absorption capacity margin of each node in the dynamic absorption capacity distribution map, the weight coefficients of the objective function for maximizing the absorption increment are dynamically adjusted. The weighting coefficients of the objective function for minimizing node voltage deviation are dynamically adjusted based on the degree of deviation between the real-time voltage and the rated voltage of each node.

5. The intelligent control method for a new energy grid-connected inverter group according to claim 2, characterized in that, The steps for optimizing the network topology of a distribution network to obtain the optimal network topology include: At the start of a preset time scale, predictive data for several future cycles and the current topology of the distribution network are acquired as inputs to the state space. All feasible switch combinations that satisfy the radial operation constraints of the distribution network are mapped to an action space represented by a spanning tree. The distribution network reconfiguration problem is modeled as a Markov decision process; Based on a pre-trained deep reinforcement learning model, the Markov decision process is solved, and the optimal network topology with the lowest expected overall cost within the current time scale period is output from the action space according to the input in the state space.

6. The intelligent control method for a new energy grid-connected inverter group according to claim 2, characterized in that, The steps for generating inverter group optimization control commands based on the optimal network topology and the adaptive multi-objective optimization function specifically include: Based on the optimal network topology, an adaptive multi-objective optimization function is used as the optimization objective to establish a complete optimization model with constraints. The constraints include power flow balance constraints, upper and lower limits of active and reactive power output of inverters, node voltage safety constraints, and line capacity constraints. Solving the complete optimization model yields a set of baseline control commands that satisfy all constraints and optimize the adaptive multi-objective optimization function. Based on current measurement and forecast data, the baseline control instruction set is modified to generate optimized control instructions for the inverter group.

7. An intelligent control system for a new energy grid-connected inverter group, used to implement the intelligent control method according to any one of claims 1-6, characterized in that, include: The absorption capacity assessment module is used to acquire measurement and forecast data of each node in the distribution network, assess the absorption capacity, and generate a dynamic absorption capacity distribution map. The dynamic absorption capacity distribution map is used to describe the spatiotemporal distribution of renewable energy power that the distribution network can accept. The objective function construction module is used to construct an adaptive multi-objective optimization function based on the dynamic absorption capacity distribution map. The network topology optimization module is used to optimize the network topology of the distribution network to obtain the optimal network topology. The control command generation module is used to generate inverter group optimization control commands based on the optimal network topology and an adaptive multi-objective optimization function. The instruction execution module is used to regulate the grid-connected inverter group of new energy sources according to the optimized control instructions of the inverter group.

8. The intelligent control system for a new energy grid-connected inverter group according to claim 7, characterized in that, The absorption capacity assessment module specifically includes: The data acquisition unit is used to collect measurement data from each node of the distribution network in real time and obtain the predicted data of the distribution network; the measurement data includes node voltage, branch current, active power and reactive power; the predicted data includes future renewable energy power prediction data and load prediction data. The baseline operating state determination unit is used to solve the current baseline operating state of the distribution network based on the online power flow calculation method, using real-time collected measurement data as the initial state and combined with prediction data. The active power absorption determination unit is used to calculate the maximum active power injection that each node can increase within a future preset time period based on the current baseline operating state of the distribution network, and obtain the maximum active power injection increment as the active power dynamic absorption capacity of the node. The reactive power absorption determination unit is used to calculate the maximum reactive power support range that each node can provide while maintaining its current active power output, based on the current baseline operating state of the distribution network, and to serve as the reactive power dynamic absorption capacity of that node. The active power absorption capacity determination unit aggregates the active power absorption capacity and reactive power absorption capacity of all nodes to generate a dynamic absorption capacity distribution map.

9. The intelligent control system for a new energy grid-connected inverter group according to claim 8, characterized in that, The objective function construction module specifically includes: The target set determination unit is used to determine the target set to be optimized based on the dynamic absorption capacity distribution map. The target set includes the following objective functions: the absorption increment maximization objective function with the goal of maximizing the increase in the total active power output of the inverter group; the network loss minimization objective function with the goal of minimizing the total active power loss of the distribution network; and the node voltage deviation minimization objective function with the goal of minimizing the sum of squares of voltage deviations of all nodes. The weight adaptive adjustment unit is used to adaptively adjust the weights of each objective function in the objective set. The objective function fusion unit is used to linearly weight and sum the objective functions after adaptive weight adjustment to construct the adaptive multi-objective optimization function at the current time.

10. The intelligent control system for a new energy grid-connected inverter group according to claim 8, characterized in that, The network topology optimization module specifically includes: The state space determination unit is used to acquire the predicted data for several future cycles and the current topology of the distribution network at the beginning of a cycle on a preset time scale, as input to the state space. The action space determination unit is used to map all feasible switch combinations that satisfy the radial operation constraints of the distribution network into an action space represented by a spanning tree. The distribution network reconfiguration modeling unit is used to model the distribution network reconfiguration problem as a Markov decision process. The optimization output unit is used to solve the Markov decision process based on the pre-trained deep reinforcement learning model. Based on the state space input, it outputs the optimal network topology with the lowest expected comprehensive cost within the current time scale period from the action space. The control command generation module specifically includes: The complete optimization model construction unit is used to establish a complete optimization model with constraints based on the optimal network topology and with an adaptive multi-objective optimization function as the optimization objective. The constraints include power flow balance constraints, upper and lower limits of inverter active and reactive power output constraints, node voltage safety constraints, and line capacity constraints. The baseline control command determination unit is used to solve the complete optimization model to obtain a baseline control command set that satisfies all constraints and makes the adaptive multi-objective optimization function optimal. The control command correction unit is used to correct the baseline control command set based on the current measurement data and prediction data, and generate optimized control commands for the inverter group.

Images