Optimal power flow control method and system for power distribution network containing hybrid power flow controller

Abstract

The invention relates to an optimal power flow control method and system for a power distribution network containing a hybrid power flow controller, and the method comprises the steps: firstly, considering a dynamic uncertainty risk, and constructing a three-phase four-wire system power distribution network optimal power flow model containing the hybrid power flow controller by taking the control cost and network loss minimization of the power distribution network as targets; and solving the optimal power flow model of the three-phase four-wire power distribution network to obtain an optimal power flow control strategy of the power distribution network containing the hybrid power flow controller. According to the method, the distributed power supply and the hybrid power flow controller are considered, the three-phase four-wire system power distribution network optimal power flow model is constructed by taking the control cost and the network loss of the power distribution network as the target, and the safe operation and the electric energy quality of the power grid are effectively guaranteed through local real-time adjustment; and the economical efficiency and the reliability when the high-proportion distributed energy is accessed to the power distribution system are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of optimal power flow control of distribution network, and particularly relates to an optimal power flow control method and system of distribution network containing a hybrid power flow controller. BACKGROUND

[0002] With the promotion of the "double carbon" goal, the complexity and uncertainty of new power systems have increased dramatically. The access of high proportion of distributed energy and asymmetric load in distribution network makes the distribution network of power system face challenges such as voltage out-of-limit and three-phase imbalance. How to control the power flow distribution in the distribution network becomes the key content to effectively deal with the strong randomness of distributed energy output and load fluctuation. The traditional power flow control method relies on centralized communication architecture and cannot solve the complex problems faced by the power flow control of distribution network, which has the bottleneck of scalability and reliability. SUMMARY

[0003] The application aims at the above-mentioned problems existing in the prior art, and provides an optimal power flow control method and system of distribution network containing a hybrid power flow controller.

[0004] To achieve the above-mentioned purposes, the technical scheme of the application is as follows:

[0005] In the first aspect, the application provides an optimal power flow control method of distribution network containing a hybrid power flow controller, which comprises the following steps:

[0006] S1, considering the dynamic uncertainty risk, taking the minimum control cost and network loss of the distribution network as the target, constructing an optimal power flow model of three-phase four-wire distribution network containing a hybrid power flow controller;

[0007] S2, solving the optimal power flow model of three-phase four-wire distribution network to obtain the optimal power flow control strategy of distribution network containing a hybrid power flow controller.

[0008] In the S1, the objective function of the optimal power flow model of three-phase four-wire distribution network comprises:

[0009] ;

[0010] ;

[0011] ;

[0012] ;

[0013] ;

[0014] ;

[0015] ;

[0016] In the above formula, For the scene The total cost below For the expectation operator based on the quantum probability model, As a control variable in total cost, For high-dimensional scene space, A trade-off coefficient between economy and reliability. As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, The total number of eigenvalues. density matrix The 1 eigenvalue, For control variables The density matrix; , These are the weighting coefficients for control costs and network loss costs, respectively. To control costs, For network loss costs, For the total time, For the phase of the power distribution system, , , These are the three phases of a three-phase four-wire power distribution system. For the number of nodes, For the scene Next node Unit active cost, For the scene Next node of Xiangzai The available margin of active power at any given moment. For the scene Next node Unit reactive power cost at the location For the scene Next node of Xiangzai The cost of reactive power compensation at any moment The total number of hybrid power flow controllers. For the first Unit active power regulation cost of a hybrid power flow controller For the first A hybrid power flow controller in Active power regulation at any given time For the time-varying quantity, For the number of branches, branch road of Xiangzai Constant network loss For the scene Next node of Xiangzai Maximum active power at any given time. For the scene Next node of Xiangzai Active power at any given time For the scene Next node of Xiangzai Reactive power at any given moment.

[0017] In S1, the constraints of the optimal power flow model for a three-phase four-wire distribution network include quantum risk constraints and Wasserstein distance constraints.

[0018] The quantum risk constraint is:

[0019] ;

[0020] In the above formula, For temperature Next The dynamic uncertainty set of a hybrid power flow controller The radius is the Wasserstein radius constant. For random disturbances or uncertain loads in the system, This represents the probability distribution characteristics of a random variable. Let be the probability distribution space. Lipschitz constant controls the upper limit of the rate of change of the probability distribution. For temperature Wasserstein distance below Given probability distribution 1, For the first Wasserstein radius constant of a hybrid power flow controller As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, For Wasserstein radius coefficient, This is the upper bound of the random demand.

[0021] The Wasserstein distance constraint is:

[0022] ;

[0023] In the above formula, For temperature Below Wasserstein distance of order For the expectation operator corresponding to the coupled distribution, for and The set of all coupling distributions, Used to measure the difference between two random variables For from random variables, It is the order of the moment.

[0024] The hybrid power flow controllers in the three-phase four-wire distribution network power flow model include STATCOM mode and SVC mode. The unit active power regulation cost of each hybrid power flow controller is calculated using the following formula:

[0025] ;

[0026] In the above formula, For the first Unit active power regulation cost of a hybrid power flow controller The total investment cost of the STATCOM unit. As a capital recovery factor, Annual operating and maintenance costs, This is the rated reactive power compensation. This is the loss penalty coefficient. To reduce power loss, To reduce the cost of thyristor-controlled reactors, The cost of thyristor-switched capacitors, For maximum reactive power compensation capacity, This represents the total cost of the switchgear. This represents the average number of times the device is switched on and off per year. For equipment lifespan;

[0027] Each hybrid power flow controller in The active power regulation at any given time is calculated using the following formula:

[0028] ;

[0029] In the above formula, For the first A hybrid power flow controller in Active power regulation at any given time For the first A hybrid power flow controller in Reactive power reference value at any given time. For the first A hybrid power flow controller in DC side current at time , For the first A hybrid power flow controller in DC side voltage at time t, For nodes voltage amplitude, For SVC equivalent reactance, This is the trigger delay angle.

[0030] The method also includes a mode conversion step, which occurs before S1;

[0031] The mode conversion is achieved through a modular multilevel converter, using the following formula to select the topology mode of the hybrid power flow controller:

[0032] ;

[0033] ;

[0034] In the above formula, For the first The topology mode selection variable for a hybrid power flow controller, and , , For reference voltage, For nodes The actual voltage, The time derivative of the state variable. , , , All are system matrices of the relevant mode. For the capacitor voltage and bridge arm current states of the modular multilevel converter submodule, For the switching modulation signal of the modular multilevel converter, For the robust feedback matrix of the relevant mode switching, To output the target value for the desired system output, This is the output of the converter.

[0035] Secondly, this invention proposes an optimal power flow control system for distribution networks containing a hybrid power flow controller, including a model building module and a model solving module;

[0036] The model building unit is used to consider dynamic uncertainty risks and, with the goal of minimizing the control cost and network loss of the distribution network, construct an optimal power flow model for a three-phase four-wire distribution network containing a hybrid power flow controller.

[0037] The model solving unit is used to solve the optimal power flow model of a three-phase four-wire distribution network and obtain the optimal power flow control strategy of the distribution network including a hybrid power flow controller.

[0038] The model building module includes an objective function building unit;

[0039] The objective function construction unit is used to construct the objective function of the optimal power flow model for a three-phase four-wire distribution network as follows:

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] ;

[0047] In the above formula, For the scene The total cost below For the expectation operator based on the quantum probability model, As a control variable in total cost, For high-dimensional scene space, A trade-off coefficient between economy and reliability. As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, The total number of eigenvalues. density matrix The 1 eigenvalue, For control variables The density matrix; , These are the weighting coefficients for control costs and network loss costs, respectively. To control costs, For network loss costs, For the total time, For the phase of the power distribution system, , , These are the three phases of a three-phase four-wire power distribution system. For the number of nodes, For the scene Next node Unit active cost, For the scene Next node of Xiangzai The available margin of active power at any given moment. For the scene Next node Unit reactive power cost at the location For the scene Next node of Xiangzai The cost of reactive power compensation at any moment The total number of hybrid power flow controllers. For the first Unit active power regulation cost of a hybrid power flow controller For the first A hybrid power flow controller in Active power regulation at any given time For the time-varying quantity, For the number of branches, branch road of Xiangzai Constant network loss For the scene Next node of Xiangzai Maximum active power at any given time. For the scene Next node of Xiangzai Active power at any given time For the scene Next node of Xiangzai Reactive power at any given moment.

[0048] The model building module also includes a quantum risk constraint building unit and a Wasserstein distance constraint building unit;

[0049] The quantum risk constraint building unit is used to construct the following quantum risk constraints:

[0050] ;

[0051] In the above formula, For temperature Next The dynamic uncertainty set of a hybrid power flow controller The radius is the Wasserstein radius constant. For random disturbances or uncertain loads in the system, This represents the probability distribution characteristics of a random variable. Let be the probability distribution space. Lipschitz constant controls the upper limit of the rate of change of the probability distribution. For temperature Wasserstein distance below Given probability distribution 1, For the first Wasserstein radius constant of a hybrid power flow controller As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, For Wasserstein radius coefficient, This is the upper bound of the random demand.

[0052] The Wasserstein distance constraint building unit is used to construct the following Wasserstein distance constraints:

[0053] ;

[0054] In the above formula, For temperature Below Wasserstein distance of order For the expectation operator corresponding to the coupled distribution, for and The set of all coupling distributions, Used to measure the difference between two random variables For from random variables, It is the order of the moment.

[0055] The hybrid power flow controllers in the three-phase four-wire distribution network power flow model include STATCOM mode and SVC mode. The unit active power regulation cost of each hybrid power flow controller is calculated using the following formula:

[0056] ;

[0057] In the above formula, For the first Unit active power regulation cost of a hybrid power flow controller The total investment cost of the STATCOM unit. As a capital recovery factor, for annual operation and maintenance cost, for rated reactive power compensation, for loss penalty coefficient, for loss power, for thyristor controlled reactor cost, for thyristor switched capacitor cost, for maximum reactive power compensation capacity, for total cost of switching device, for average switching times per year, for device life;

[0058] active power regulation of each hybrid power flow controller at time is calculated by the following formula:

[0059]

[0060] In the above formula, is the active power regulation of the th hybrid power flow controller at time, is the reactive power reference value of the th hybrid power flow controller at time, is the DC side current of the th hybrid power flow controller at time, is the DC side voltage of the th hybrid power flow controller at time, is the voltage amplitude of node , is the equivalent reactance of SVC, is the trigger delay angle.

[0061] The system further comprises a mode conversion module, which is located before the model construction module;

[0062] The mode conversion module is used to select the topology mode of the hybrid power flow controller by modular multilevel converter, by the following formula:

[0063]

[0064]

[0065] In the above formula, is the topology mode selection variable of the th hybrid power flow controller, and , , is the reference voltage, ​​​the actual voltage of the node the time derivative of the state variable , , , , the system matrix of the relevant mode the capacitor voltage and bridge arm current state of the modular multilevel converter sub-module the switching modulation signal of the modular multilevel converter the switching robust feedback matrix of the relevant mode the expected system output target value the output quantity of the converter

[0066] Compared with the prior art, the present application has the following beneficial effects:

[0067] 1. The power distribution network optimal power flow control method and system containing a hybrid power flow controller are proposed, which firstly considers the dynamic uncertainty risk, takes the minimum control cost and network loss of the power distribution network as the target, and constructs a three-phase four-wire power distribution network optimal power flow model containing a hybrid power flow controller; then the optimal power flow control strategy of the power distribution network containing the hybrid power flow controller is obtained by solving the three-phase four-wire power distribution network optimal power flow model. The method considers the distributed power supply and the hybrid power flow controller, takes the minimum control cost and network loss of the power distribution network as the target, constructs a three-phase four-wire power distribution network optimal power flow model, effectively guarantees the safe operation and power quality of the power grid through local real-time adjustment, and improves the economy and reliability of the high-proportion distributed energy access to the power distribution system.

[0068] 2. The power distribution network optimal power flow control method and system containing a hybrid power flow controller are proposed, which pre-considers the uncertainty risk of load and output, constructs quantum risk constraints and Wasserstein distance constraints, effectively handles the strong randomness of distributed energy output and the volatility of load, reduces the calculation complexity under the high-dimensional scene space, solves the power flow change problem caused by the access of distributed energy to the power distribution network, and improves the compatibility of the power distribution network with high-proportion distributed energy.

[0069] 3. The power distribution network optimal power flow control method and system containing a hybrid power flow controller are proposed, which introduces a modular multilevel converter as the core hardware carrier of the hybrid power flow controller, realizes seamless switching of STATCOM mode and SVC mode through dynamic switching of sub-modules, supports online topology reconstruction through dynamic adjustment of the reactive power compensation mode of the hybrid power flow controller, solves the power flow control problem after the introduction of the hybrid power flow controller into the power distribution system, and better realizes the economic and efficient operation of the power distribution network. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 This is an overall flowchart of the method described in this invention.

[0071] Figure 2 The graph shows the comparison results of active power loss of the method described in Example 1.

[0072] Figure 3 This is a comparison diagram of the node voltage results of the method described in Example 1.

[0073] Figure 4 This is a structural diagram of the system described in this invention. Detailed Implementation

[0074] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0075] This invention proposes an optimal power flow control method and system for distribution networks incorporating a hybrid power flow controller. First, aiming to minimize distribution network control costs and network losses, an optimal power flow model for a three-phase four-wire distribution network with a hybrid power flow controller is constructed. Then, a modular multilevel converter (MLC) is introduced as the core hardware carrier of the hybrid controller. Seamless switching between STATCOM and SVC is achieved through dynamic switching of sub-modules, supporting online topology reconfiguration. Finally, a spatiotemporal graph attention mechanism is designed to capture the coupling relationship between three-phase imbalance and topology evolution in the distribution network, achieving decentralized control. A quantum-inspired robust optimization is also designed, combining quantum probability models and deep reinforcement learning to construct a dynamic uncertainty set, addressing the computational complexity problem in high-dimensional scenarios. This effectively ensures the safe operation and power quality of the power grid, and improves the economy and reliability when a high proportion of distributed energy resources are integrated into the distribution system.

[0076] Example 1:

[0077] like Figure 1 As shown, the optimal power flow control method for a distribution network containing a hybrid power flow controller is carried out in the following steps:

[0078] 1. Considering dynamic uncertainty risks, and aiming to minimize the control cost and network loss of the distribution network, construct an optimal power flow model for a three-phase four-wire distribution network with a hybrid power flow controller.

[0079] Considering distributed generation and hybrid power flow controller (HFC), the objective function of the optimal power flow model for a three-phase four-wire distribution network includes:

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] ;

[0085] In the above formula, For the scene The total cost below , These are the weighting coefficients for control costs and network loss costs, respectively. To control costs, For network loss costs, For the total time, For the phase of the power distribution system, , , These are the three phases of a three-phase four-wire power distribution system. For the number of nodes, For the scene Next node Unit active cost, For the scene Next node of Xiangzai The available margin of active power at any given moment. For the scene Next node Unit reactive power cost at the location For the scene Next node of Xiangzai The cost of reactive power compensation at any moment The total number of hybrid power flow controllers. For the first Unit active power regulation cost of a hybrid power flow controller For the first A hybrid power flow controller in Active power regulation at any given time For the time-varying quantity, For the number of branches, branch road of Xiangzai Constant network loss For the scene Next node of Xiangzai Maximum active power at any given time. For the scene Next node of Xiangzai Active power at any given time for the scenario lower node of phase in instantaneous reactive power.

[0086] In order to deal with the complex operation state of distribution network and the limitation of single power flow controller, the static synchronous compensator (STATCOM) is combined with the active power flow controller (SVC) as a hybrid power flow controller of three-phase four-wire distribution system;

[0087] When the distribution network faces the risk of load surge and voltage collapse, the static synchronous compensator is preferred to quickly compensate for the reactive power gap and raise the voltage. In actual engineering, the PI controller is introduced to realize the steady-state accuracy and anti-interference ability of the static synchronous compensator, and the STATCOM model is:

[0088] ;

[0089] In the above formula, is the reactive power of the static synchronous compensator, , are the PI controller parameters, is the reference voltage, is the actual voltage of the node ;

[0090] The active power flow controller is the key device for dynamic reactive power regulation in the distribution network, and its core function is achieved through the coordinated control of series reactors and variable capacitor banks, and the SVC model is:

[0091] ;

[0092] In the above formula, is the reactive power of the active power flow controller, is the SVC parallel voltage amplitude, is the SVC reactor rated reactance, is the SVC capacity, is the maximum SVC capacity, is the angle of the node , is the reference angle, is the angle difference, used to adjust the direction of reactive power.

[0093] In summary, the unit active regulation cost of each hybrid power flow controller is calculated using the following formula:

[0094] ;

[0095] In the above formula, is the first Unit active power regulation cost of a hybrid power flow controller The total investment cost of the STATCOM unit. As a capital recovery factor, Annual operating and maintenance costs, This is the rated reactive power compensation. This is the loss penalty coefficient. To reduce power loss, To reduce the cost of thyristor-controlled reactors, The cost of thyristor-switched capacitors, For maximum reactive power compensation capacity, This represents the total cost of the switchgear. This represents the average number of times the device is switched on and off per year. For equipment lifespan;

[0096] Each hybrid power flow controller in The active power regulation at any given time is calculated using the following formula:

[0097] ;

[0098] In the above formula, For the first A hybrid power flow controller in Active power regulation at any given time For the first A hybrid power flow controller in Reactive power reference value at any given time. For the first A hybrid power flow controller in DC side current at time , For the first A hybrid power flow controller in DC side voltage at time 10:00 For nodes voltage amplitude, For SVC equivalent reactance, This is the trigger delay angle.

[0099] A modular multilevel converter (MMC) model is introduced as the core hardware carrier of the hybrid power flow controller. Seamless switching between STATCOM and SVC modes is achieved through dynamic switching of sub-modules, supporting online topology reconfiguration.

[0100] ;

[0101] ;

[0102] In the above formula, For the first The topology mode selection variable for a hybrid power flow controller, and , , For reference voltage, For nodes The actual voltage, The time derivative of the state variable. , , , All are system matrices of the relevant mode. For the capacitor voltage and bridge arm current states of the modular multilevel converter submodule, For the switching modulation signal of the modular multilevel converter, For the robust feedback matrix of the relevant mode switching, To output the target value for the desired system output, This is the output of the converter.

[0103] The constraints of the optimal power flow model for a three-phase four-wire distribution network include three power balance constraints, distributed generation output constraints, branch current and network loss constraints, node voltage quality constraints, and neutral line current constraints (the core constraints of a three-phase four-wire system). If the distribution system contains distribution transformers, the constraints also include transformer capacity constraints.

[0104] The three power balance constraints are:

[0105] ;

[0106] In the above formula, , They are nodes of Xiangzai Active power and reactive power at any given time , They are nodes of Xiangzai time Active power and reactive power , They are nodes of Xiangzai The active and reactive power of a fixed load at all times. , They are nodes of Xiangzai The active and reactive power that constantly interacts with the external power grid have a positive inflow rate. , Branch roads of Xiangzai Active power and reactive power at any given time;

[0107] The output constraint of the distributed power source is:

[0108] ;

[0109] In the above formula, , They are nodes of Maximum active power and maximum reactive power of a phase;

[0110] The branch current and network loss constraints are as follows:

[0111] ;

[0112] ;

[0113] In the above formula, branch road of Xiangzai Current at any moment Based on the conductor's current carrying capacity, the branch of The maximum permissible current of the phase, For the total time, The total number of branches, branch road of The phase resistance value, This represents the maximum total active power loss.

[0114] The node voltage quality constraint is:

[0115] ;

[0116] In the above formula, For nodes of Xiangzai Voltage at time, , For nodes of The lower and upper limits of the phase voltage amplitude;

[0117] The neutral line current constraint is as follows:

[0118] ;

[0119] In the above formula, For nodes exist The effective value of the neutral current at time t. the current effective value of the node at the moment , the maximum allowable current of the neutral line;

[0120] The transformer capacity constraint is:

[0121] ;

[0122] In the above formula, is the apparent power of the first transformer at the moment , , , is the active power and the reactive power of the first transformer at the moment , , is the maximum apparent power of the first transformer.

[0123] 2. Considering the risk of dynamic uncertainty, a quantum heuristic robust optimization is designed, which combines a robust stochastic optimization model, a quantum probability model and deep reinforcement learning, to construct dynamic uncertainty constraints, including quantum risk constraints and Wasserstein distance constraints, to solve the problem of computational complexity in high-dimensional scene space.

[0124] The general form of the robust stochastic optimization model is:

[0125] ;

[0126] ;

[0127] In the above formula, , is the decision variable of the uncertain quantity, , , , , , , is the uncertain set, is the adjustment range of the uncertain set, is the event linear approximation rule, is the approximation order, is the event indication parameter.

[0128] Through the quantum probability model, the traditional robust stochastic optimization model is improved in combination with the Wasserstein distance model:

[0129] ​​Consider a data-driven setup where Wasserstein fuzzy sets are based on empirical distributions. Centered on, among which Number each data point. To construct an experience distribution Historical data points The weights in the distribution, For the first Given a set of observations, provide a tractable distance variable. ,in For the first The dimension of uncertainty is defined through an optimization problem. and The Wasserstein k-norm measure between them is used to construct the Wasserstein distance constraint as follows:

[0130] ;

[0131] In the above formula, For temperature Below Wasserstein distance of order For the expectation operator corresponding to the coupled distribution, for and The set of all coupling distributions, Used to measure the difference between two random variables For from random variables, It is the order of the moment;

[0132] in, and These are two mutually exclusive cases. Since we are considering probability theory here, separating the finite and infinite cases is standard practice. The physical meaning here can be interpreted as: The corresponding behavior should consider the worst-case scenario. This corresponds to the behavior that considers the average case.

[0133] Since traditional robust optimization only considers the uncertainty of the probability space, this invention introduces a quantum probability model. The density matrix represents the quantum state of the state variable, and the trace of the covariance matrix is ​​equal to the sum of all eigenvalues ​​of the density matrix. The trace is used to calculate the quantum uncertainty risk, thus achieving the goal of considering uncertainty.

[0134] ;

[0135] ;

[0136] ;

[0137] In the above formula, is the total risk measure of quantum uncertainty, the larger the value, the higher the uncertainty, is the covariance matrix of the control variable reflects the uncertainty distribution, is the total number of eigenvalues, is the first eigenvalue of the density matrix characterizes the probability distribution of the quantum state, is the density matrix of the control variable , is the equivalent temperature parameter, is the first energy level, is the first energy level, is the scenario probability of the scenario , is the quantum state under the scenario ;

[0138] In combination with the quantum risk measure, the dynamic uncertainty risk is considered, including both the empirical distribution deviation of distributed power output (Wasserstein distance) and the uncertainty risk of quantum state. The quantum risk constraint is constructed as:

[0139] ;

[0140] In the above formula, is the dynamic uncertainty set of the first hybrid flow controller under temperature , is the Wasserstein radius constant, is the random disturbance or uncertain load in the system, represents the probability distribution characteristics of the random variable, is the probability distribution space, is the Lipschitz constant, which controls the upper limit of the change rate of the probability distribution, is the Wasserstein distance under temperature , is the probability distribution 1, is the Wasserstein radius constant of the first hybrid flow controller, is the total risk measure of quantum uncertainty, is the covariance matrix of the control variable , is the Wasserstein radius coefficient, is an upper bound of the random demand;

[0141] To cope with the computational complexity of high-dimensional scenario space, deep reinforcement learning (DRL) is used to approximate the optimal control strategy. The policy network learns the transition probability and reward function between scenarios through Q function, and dynamically adjusts the control variables to minimize the total cost:

[0142] ;

[0143] In the above formula, is the state , the probability of taking action is determined by the policy network parameters , is the Q value of action under state , which is approximated by a neural network, is a temperature parameter that adjusts the balance between exploration and exploitation, the larger the exploration probability is;

[0144] The final optimization objective of the three-phase four-wire distribution network optimal power flow model is to meet the quantum risk constraint and Wasserstein distance constraint model:

[0145] ;

[0146] In the above formula, is the total cost under scenario , is the control variable in the total cost, i.e. the variable after considering the uncertainty modeling, including scenario parameters (renewable energy output) in modeling, control parameters (HFC control quantity, generator output), economic parameters (unit cost coefficient), etc. is a high-dimensional scenario space, is an expected operator based on a quantum probability model, is a trade-off coefficient between economic efficiency and reliability, adjusting the priority of economic efficiency and reliability, is the total risk measure of quantum uncertainty, is the covariance matrix of the control variable .

[0147] 3. A deep reinforcement learning framework is used to solve the three-phase four-wire distribution network optimal power flow model, and the optimal power flow control strategy of the distribution network containing the mixed type flow controller is obtained;

[0148] Aiming at the coupling problem of three-phase imbalance (such as phase difference caused by distributed power access and nonlinear load) and topology evolution (such as changes in power grid structure caused by switch action and dynamic access / exit of distributed resources), a space-time graph attention mechanism (ST-GNN) is designed to process network topology and dynamic operation data, optimize three-phase imbalance in distribution network of power system, realize decentralized control, reduce response time, improve response speed, save comprehensive cost, reduce computational complexity and communication delay, and improve the scalability of power flow control.

[0149] The core of ST-GNN is to generate node state updates through space-time joint modeling, including spatial dimension graph attention mechanism and time dimension GRU feature fusion.

[0150] First, in the spatial dimension, the distribution network is converted into a graph structure , where is the node, which specifically represents the bus, distributed power and load node, is the edge, which represents the branch impedance and topological connection relationship; and the measurement data of the continuous time window is input in the time dimension.

[0151] Then the spatial attention calculation and time feature fusion are combined, including:

[0152] The space-time joint state update formula is:

[0153] ;

[0154] Among them, the spatial dimension graph attention mechanism, i.e. the spatial feature modeling formula, is:

[0155] ;

[0156] Edge features directly encode the physical characteristics of the distribution network, such as line impedance: affecting three-phase current distribution; topology state: switch open / close, identifying changes in network connectivity, etc.

[0157] The time dimension GRU feature fusion, i.e. the time dimension dynamic modeling formula, is:

[0158] ;

[0159] The gated recurrent unit (GRU) is used as the time series modeling module, with the input being the node state at the previous time and the state updated by the current spatial attention, and the output being the hidden state at the current time. The update gate (Update Gate) and reset gate (Reset Gate) control the preservation of historical information and the fusion of new information, capturing the time series dependence of the dynamic process of the distribution network, such as the trend of three-phase imbalance over time and the continuity of topology evolution.

[0160] In the above formula, is a node The hidden state of the first layer contains node features, is an activation function, is a node The neighbor set of the node, considering the physical connection relationship of the power distribution network, is the attention weight of the neighbor node to the neighbor node, The attention weight of the neighbor node is the learnable weight matrix of the first layer, used for feature transformation, is a concatenation operation, which integrates the node's own features, neighbor features and edge features, is a node The hidden state of the first layer, is a node The hidden state of the first layer, is the edge feature between the node and the node , such as line impedance, topology state identifier, fault signal, etc. is a learnable parameter vector of the attention mechanism, is a gated recurrent unit, is a node The hidden state of the first layer.

[0161] After multi-layer propagation, the node-level output voltage regulation instruction of the final layer, the edge-level output power flow control signal, and the global-level output system operation state evaluation.

[0162] The spatial dimension captures the real-time interaction between nodes through the attention mechanism, and the time dimension models the state evolution through the GRU. The two are updated alternately (first space and then time, or space-time parallel), realizing the joint representation of “static topology structure + dynamic running state”.

[0163] In order to verify the effectiveness of the application, the local control strategy described in the application is compared with the traditional centralized control strategy. The active loss comparison result is as shown in Figure 2 , and the node voltage comparison result is as shown in Figure 3 . Figure 2In the middle, the active power change trend of the local control strategy is similar to that of the centralized control strategy, both showing double-peak characteristics in the afternoon (11-14) and at night (18-20). However, there is a significant difference in volatility between the two: the power curve of the centralized control strategy has a larger fluctuation range, especially during the period of sudden load surge (such as 12 o'clock), the power jumps by more than 30%, while the fluctuation range of the local control strategy is always controlled within ±5% (40-100 KWh). Figure 3 In the middle, the voltage fluctuation range of the local control strategy (0.995-1.000 p.u.) is significantly smaller than that of the centralized control strategy (0.992-0.999 p.u.). The centralized control strategy optimizes the voltage based on global information and stabilizes it in a narrow range (standard deviation ≤0.003 p.u.), but its mode of relying on centralized calculation leads to response delay and difficulty in dealing with high-frequency fluctuations of DER; while the local control strategy considers the uncertainty of load and output in advance through a robust stochastic optimization algorithm, and completes reactive power compensation and power flow reconstruction within milliseconds. Although the voltage instantaneous fluctuation can reach ±0.8%, the long-term voltage qualification rate is maintained above 98% through probability constraints, indicating that the local control strategy achieves higher compatibility of distributed energy resources (DER) through dynamic adjustment. This local control strategy, which uses historical data preprocessing and machine learning training classification to regulate and control the voltage of the distribution network without the need for communication infrastructure, can achieve economic and efficient operation of the distribution network.

[0164] To verify the economy of the application, the scheme without control strategy is taken as scheme 1, the scheme with centralized control strategy is taken as scheme 2, and the scheme with the local control strategy of the application is taken as scheme 3 for comparison, and the comparison results are shown in Table 1. Among them, scheme 2 is a centralized global optimization that performs fine calculation of reactive or active adjustment amount for each photovoltaic node against the most stringent voltage limit. As can be seen from Table 1, scheme 2 reduces network loss to 5.71%, which is 0.57% lower than scheme 1; three-phase imbalance is optimized to 2.13%, which is 6.19% less than scheme 1; but its comprehensive cost is as high as 45,038, which is the highest among the three schemes, and the maximum voltage is 1.089 p.u., which fails to completely avoid the risk of voltage limit, resulting in high complexity of control process, significant increase in communication and computing cost; the maximum voltage of scheme 3 is 1.068 p.u., the network loss is 5.39%, and the three-phase imbalance is 1.09%, all of which are better than scheme 2, among which the network loss is further reduced by 0.32%, the three-phase imbalance is reduced by 1.04%, and the comprehensive cost is only 31,589, which is 29.8% less than scheme 2, avoiding the calculation complexity and communication delay of the centralized control strategy, and effectively ensuring the safe operation and power quality of the power grid through local real-time adjustment, fully embodying the economic and reliable advantages of the application in the scenario of high proportion of distributed energy access.

[0165] Table 1. Relevant data for each control strategy

[0166] .

[0167] Example 2:

[0168] like Figure 4 As shown, the optimal power flow control system for distribution networks with a hybrid power flow controller includes a model building module and a model solving module.

[0169] The model building unit is used to consider dynamic uncertainty risks and, with the goal of minimizing the control cost and network loss of the distribution network, construct an optimal power flow model for a three-phase four-wire distribution network containing a hybrid power flow controller.

[0170] The model solving unit is used to solve the optimal power flow model of a three-phase four-wire distribution network and obtain the optimal power flow control strategy of the distribution network including a hybrid power flow controller.

[0171] The model building module includes an objective function building unit;

[0172] The objective function construction unit is used to construct the objective function of the optimal power flow model for a three-phase four-wire distribution network as follows:

[0173] ;

[0174] ;

[0175] ;

[0176] ;

[0177] ;

[0178] ;

[0179] ;

[0180] In the above formula, For the scene The total cost below For the expectation operator based on the quantum probability model, As a control variable in total cost, For high-dimensional scene space, A trade-off coefficient between economy and reliability. As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, The total number of eigenvalues. density matrix The 1 eigenvalue, For control variables The density matrix; , These are the weighting coefficients for control costs and network loss costs, respectively. To control costs, For network loss costs, For the total time, For the phase of the power distribution system, , , These are the three phases of a three-phase four-wire power distribution system. For the number of nodes, For the scene Next node Unit active cost, For the scene Next node of Xiangzai The available margin of active power at any given moment. For the scene Next node Unit reactive power cost at the location For the scene Next node of Xiangzai The cost of reactive power compensation at any moment The total number of hybrid power flow controllers. For the first Unit active power regulation cost of a hybrid power flow controller For the first A hybrid power flow controller in Active power regulation at any given time For the time-varying quantity, For the number of branches, branch road of Xiangzai Constant network loss For the scene Next node of Xiangzai Maximum active power at any given time. For the scene Next node of Xiangzai Active power at any given time For the scene Next node of Xiangzai Reactive power at any given moment.

[0181] The model building module also includes a quantum risk constraint building unit and a Wasserstein distance constraint building unit;

[0182] The quantum risk constraint building unit is used to construct the following quantum risk constraints:

[0183] ;

[0184] In the above formula, For temperature Next The dynamic uncertainty set of a hybrid power flow controller The radius is the Wasserstein radius constant. For random disturbances or uncertain loads in the system, This represents the probability distribution characteristics of a random variable. Let be the probability distribution space. Lipschitz constant controls the upper limit of the rate of change of the probability distribution. For temperature Wasserstein distance below Given probability distribution 1, For the first Wasserstein radius constant of a hybrid power flow controller As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, For Wasserstein radius coefficient, This is the upper bound of the random demand.

[0185] The Wasserstein distance constraint building unit is used to construct the following Wasserstein distance constraints:

[0186] ;

[0187] In the above formula, For temperature Below Wasserstein distance of order For the expectation operator corresponding to the coupling distribution, for and The set of all coupling distributions, Used to measure the difference between two random variables For from random variables, It is the order of the moment.

[0188] The hybrid power flow controllers in the three-phase four-wire distribution network power flow model include STATCOM mode and SVC mode. The unit active power regulation cost of each hybrid power flow controller is calculated using the following formula:

[0189] ;

[0190] In the above formula, For the first Unit active power regulation cost of a hybrid power flow controller The total investment cost of the STATCOM unit. As a capital recovery factor, Annual operating and maintenance costs, This is the rated reactive power compensation. This is the loss penalty coefficient. To reduce power loss, To reduce the cost of thyristor-controlled reactors, The cost of thyristor-switched capacitors, For maximum reactive power compensation capacity, This represents the total cost of the switchgear. This represents the average number of times the device is switched on and off per year. For equipment lifespan;

[0191] Each hybrid power flow controller in The active power regulation at any given time is calculated using the following formula:

[0192] ;

[0193] In the above formula, For the first A hybrid power flow controller in Active power regulation at any given time For the first A hybrid power flow controller in Reactive power reference value at any given time. For the first A hybrid power flow controller in DC side current at time , For the first A hybrid power flow controller in DC side voltage at time 10:00 For nodes voltage amplitude, For SVC equivalent reactance, This is the trigger delay angle.

[0194] The system also comprises a mode conversion module, which is located before the model construction module;

[0195] The mode conversion module is used to select the topological mode of the hybrid power flow controller by the modular multilevel converter, using the following formula:

[0196]

[0197]

[0198] In the above formula, is the topological mode selection variable of the mth hybrid power flow controller, and , , , is the reference voltage, is the actual voltage of the node , is the time derivative of the state variable, , , , are system matrices of the relevant mode, is the capacitor voltage and bridge arm current state of the modular multilevel converter submodule, is the switching modulation signal of the modular multilevel converter, is the switching robust feedback matrix of the relevant mode, is the desired system output target value, is the output quantity of the converter.​​

Claims

1. A power distribution network optimal power flow control method with hybrid power flow controllers, characterized in that, the method comprises: S1, considering the dynamic uncertainty risk, constructing a three-phase four-wire power distribution network optimal power flow model with hybrid power flow controllers, aiming at minimizing the control cost and network loss of the power distribution network; S2, solving the three-phase four-wire power distribution network optimal power flow model to obtain the optimal power flow control strategy of the power distribution network with hybrid power flow controllers.

2. The power distribution network optimal power flow control method with hybrid power flow controllers according to claim 1, characterized in that, in S1, the objective function of the three-phase four-wire power distribution network optimal power flow model includes: ； ； ； ； ； ； ； In the above formula, For the scene The total cost below For the expectation operator based on the quantum probability model, As a control variable in total cost, For high-dimensional scene space, A trade-off coefficient between economy and reliability. As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, The total number of eigenvalues. density matrix The 1 eigenvalue, For control variables The density matrix; , These are the weighting coefficients for control costs and network loss costs, respectively. To control costs, For network loss costs, For the total time, For the phase of the power distribution system, , , These are the three phases of a three-phase four-wire power distribution system. For the number of nodes, For the scene Next node Unit active cost, For the scene Next node of Xiangzai The available margin of active power at any given moment. For the scene Next node Unit reactive power cost at the location For the scene Next node of Xiangzai The cost of reactive power compensation at any moment The total number of hybrid power flow controllers. For the first Unit active power regulation cost of a hybrid power flow controller For the first A hybrid power flow controller in Active power regulation at any given time For the time-varying quantity, For the number of branches, branch road of Phase in Network loss at the moment, For the scenario Lower node The Phase in Maximum active power at the moment, For the scenario Lower node The Phase in Active power at the moment, For the scenario Lower node The Phase in Reactive power at the moment.

3. The power distribution network optimal power flow control method with hybrid power flow controllers according to claim 1, characterized in that, in S1, the constraint conditions of the three-phase four-wire power distribution network optimal power flow model include quantum risk constraints and Wasserstein distance constraints; the quantum risk constraint is: ； In the above formula, is the temperature In the following, the dynamic uncertainty set of the mixed flow controller, is the Wasserstein radius constant, is the random disturbance or uncertain load in the system, represents the probability distribution characteristics of the random variable, is the probability distribution space, is the Lipschitz constant, which controls the upper limit of the change rate of the probability distribution, is the Wasserstein distance under the temperature is the probability distribution 1, is the Wasserstein radius constant of the mixed flow controller, is the total risk measure of quantum uncertainty, is the covariance matrix of the control variable , and is the Wasserstein radius coefficient, is the upper bound of the random demand; the Wasserstein distance constraint is: ； In the above formula, For temperature Below Wasserstein distance of order For the expectation operator corresponding to the coupled distribution, for and The set of all coupling distributions, Used to measure the difference between two random variables For from random variables, It is the order of the moment.

4. The power distribution network optimal power flow control method with hybrid power flow controllers according to claim 2, characterized in that, the hybrid power flow controller in the three-phase four-wire power distribution network flow model includes STATCOM mode and SVC mode, and the unit active regulation cost of each hybrid power flow controller is calculated by the following formula: ； In the above formula, the unit active regulation cost of the first hybrid flow controller, the total investment cost of the STATCOM device, the capital recovery factor, the annual operation and maintenance cost, the rated reactive compensation amount, the loss penalty coefficient, the loss power, the cost of the thyristor-controlled reactor, the cost of the thyristor-switched capacitor, the maximum reactive compensation capacity, the total cost of the switching device, the average number of switching per year, the device life; The mixed flow controllers at the active power regulation amount at the moment of time is calculated using the following formula: ； In the above formula, For the first A hybrid power flow controller in Active power regulation at any given time For the first A hybrid power flow controller in Reactive power reference value at any given time. For the first A hybrid power flow controller in DC side current at time , For the first A hybrid power flow controller in DC side voltage at time t, For nodes voltage amplitude, For SVC equivalent reactance, This is the trigger delay angle.

5. The power distribution network optimal power flow control method with hybrid power flow controllers according to claim 4, characterized in that, the method further comprises mode conversion, which is located before S1; the mode conversion selects the topology mode of the hybrid power flow controller by the modular multilevel converter, using the following formula: ； ； In the above formula, is the topology mode selection variable of the first hybrid flow controller, and , , is a reference voltage, is an actual voltage of a node , is a time derivative of a state variable, , , , is a system matrix of a related mode, is a capacitor voltage and a bridge arm current state of a modular multilevel converter sub-module, is a switching modulation signal of the modular multilevel converter, is a switching robust feedback matrix of a related mode, is a desired system output target value, is an output quantity of the converter.

6. A power distribution network optimal power flow control system with hybrid power flow controllers, characterized in that, the system comprises a model construction module and a model solving module; the model construction unit is used to construct a three-phase four-wire power distribution network optimal power flow model with hybrid power flow controllers, considering the dynamic uncertainty risk and aiming at minimizing the control cost and network loss of the power distribution network; the model solving unit is used to solve the three-phase four-wire power distribution network optimal power flow model to obtain the optimal power flow control strategy of the power distribution network with hybrid power flow controllers.

7. The power distribution network optimal power flow control system with hybrid power flow controllers according to claim 6, characterized in that, the model construction module includes an objective function construction unit; the objective function construction unit is used to construct the objective function of the following three-phase four-wire power distribution network optimal power flow model: ； ； ； ； ； ； ； In the above formula, For the scene The total cost below For the expectation operator based on the quantum probability model, As a control variable in total cost, For high-dimensional scene space, A trade-off coefficient between economy and reliability. As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, The total number of eigenvalues. density matrix The 1 eigenvalue, For control variables The density matrix; , These are the weighting coefficients for control costs and network loss costs, respectively. To control costs, For network loss costs, For the total time, For the phase of the power distribution system, , , These are the three phases of a three-phase four-wire power distribution system. For the number of nodes, For the scene Next node Unit active cost, For the scene Next node of Xiangzai The available margin of active power at any given moment. For the scene Next node Unit reactive power cost at the location For the scene Next node of Xiangzai The cost of reactive power compensation at any moment The total number of hybrid power flow controllers. For the first Unit active power regulation cost of a hybrid power flow controller For the first A hybrid power flow controller in Active power regulation at any given time For the time-varying quantity, For the number of branches, branch road of Xiangzai Constant network loss For the scene Next node of Xiangzai Maximum active power at any given time. For the scene Next node of Xiangzai Active power at any given time For the scene Next node of Xiangzai Reactive power at any given moment.

8. The power distribution network optimal power flow control system with hybrid power flow controllers according to claim 6, characterized in that, the model construction module further includes a quantum risk constraint construction unit and a Wasserstein distance constraint construction unit; the quantum risk constraint construction unit is used to construct the following quantum risk constraint: ； In the above formula, For temperature Next The dynamic uncertainty set of a hybrid power flow controller The radius is the Wasserstein radius constant. For random disturbances or uncertain loads in the system, This represents the probability distribution characteristics of a random variable. Let be the probability distribution space. Lipschitz constant controls the upper limit of the rate of change of the probability distribution. For temperature Wasserstein distance below Given probability distribution 1, For the first Wasserstein radius constant of a hybrid power flow controller As a measure of the total risk of quantum uncertainty, For control variables The covariance matrix, For Wasserstein radius coefficient, This is the upper bound of the random demand. the Wasserstein distance constraint construction unit is used to construct the following Wasserstein distance constraint: ； In the above formula, For temperature Below Wasserstein distance of order For the expectation operator corresponding to the coupling distribution, for and The set of all coupling distributions, Used to measure the difference between two random variables For from random variables, It is the order of the moment.

9. The power distribution network optimal power flow control system with hybrid power flow controllers according to claim 7, characterized in that, The hybrid power flow controller in the three-phase four-wire power distribution network power flow model comprises a STATCOM mode and an SVC mode, and the unit active regulation cost of each hybrid power flow controller is calculated by using the following formula: ； In the above formula, is the unit active regulation cost of the th hybrid flow controller, is the total investment cost of the STATCOM device, is the capital recovery factor, is the annual operation and maintenance cost, is the rated reactive power compensation amount, is the loss penalty coefficient, is the loss power, is the cost of the thyristor-controlled reactor, is the cost of the thyristor-switched capacitor, is the maximum reactive power compensation capacity, is the total cost of the switching device, is the average number of switching per year, is the device life. The mixed flow controllers at the active power regulation amount at the moment of time is calculated using the following formula: ； In the above formula, For the first A hybrid power flow controller in Active power regulation at any given time For the first A hybrid power flow controller in Reactive power reference value at any given time. For the first A hybrid power flow controller in DC side current at time , For the first A hybrid power flow controller in DC side voltage at time t, For nodes voltage amplitude, For SVC equivalent reactance, This is the trigger delay angle.

10. The optimal power flow control system of the power distribution network with hybrid power flow controllers according to claim 9, characterized in that, The system further comprises a mode conversion module, which is located before the model construction module; The mode conversion module is used to select the topology mode of the hybrid power flow controller by using the following formula through the modular multilevel converter: ； ； In the above formula, is the topology mode selection variable of the first hybrid flow controller, and , , is a reference voltage, is an actual voltage of a node , is a time derivative of a state variable, , , , is a system matrix of a related mode, is a capacitor voltage and a bridge arm current state of a modular multilevel converter sub-module, is a switching modulation signal of the modular multilevel converter, is a switching robust feedback matrix of a related mode, is a desired system output target value, is an output quantity of the converter.

Images