Calculation method of maximum power supply capability of power grid considering dynamic voltage stability constraints

Abstract

The present invention relates to the field of power systems, and in particular to a method for calculating the maximum power supply capacity of a power grid taking into account dynamic voltage stability constraints. Step 1: Establish a maximum power supply capacity calculation model for a partitioned interconnected power grid taking into account dynamic voltage stability constraints, the steps include: establishing an objective function for the maximum power supply capacity; establishing constraints for the maximum power supply capacity, the constraints include: partitioned interconnected device constraints, AC power grid equality constraints, node voltage amplitude constraints, and dynamic voltage stability constraints; wherein, in the dynamic voltage stability constraints, an improved small disturbance analysis method is adopted, and the voltage of the load node is used instead of the equivalent potential in the electromagnetic power; Step 2: A deep deterministic policy gradient algorithm is used to solve the maximum power supply capacity calculation model for the partitioned interconnected power grid taking into account dynamic voltage stability constraints, and obtain the maximum power supply capacity value. The present invention solves the problem that voltage instability in partitioned flexible interconnected power grids is becoming more and more prominent and the power supply capacity of power grids is declining.

Description

Technical Field

[0001] The invention relates to the field of electric power systems, and in particular to a method for calculating the maximum power supply capacity of a partitioned interconnected power grid taking into account dynamic voltage stability constraints. Background Art

[0002] In order to solve the problem of excessive short-circuit current caused by the expansion of the power grid and prevent the hidden dangers of accidents in the electromagnetic ring network, large power grids usually operate in a zoned and segmented manner. However, with the development of the economy and society, the division of zones has become more and more detailed, and the open-loop operation mode between zones is difficult to cope with extreme situations of the power grid. If a flexible DC interconnection device is installed between power grid zones, real-time active and reactive power support can be achieved, which improves the reliability of power supply and enhances the power supply capacity of the power grid and zones.

[0003] The total power-supply capability (TSC) of a power grid is defined as the maximum load supply capability that can be sustained within a certain power supply area under the N-1 safety criteria and various actual operation constraints. The traditional TSC research method mainly considers constraints such as power flow equation, component thermal stability, node voltage, generator, and N-1 static safety. However, with the increasing proportion of new energy in the power grid and the increasingly obvious characteristics of "strong direct current and weak alternating current", low-frequency oscillations continue to occur, and voltage instability is more likely to occur at load points. The results obtained by the traditional TSC model are too ideal, and there is a risk of voltage instability in future power grid operation. The voltage instability process is essentially a dynamic change process, and it will be more accurate to study the voltage stability problem from a dynamic perspective. Therefore, when calculating the TSC of future power grids, the impact of dynamic voltage on stability must be considered to ensure the safe and stable operation of the power grid.

[0004] Dynamic voltage is designed to analyze voltage stability from a nonlinear perspective. The traditional small disturbance analysis method converts nonlinear differential equations into linear ones, which improves the calculation speed, but the calculation error is large. Summary of the invention

[0005] The purpose of the present invention is to provide a method for calculating the maximum power supply capacity of a power grid taking into account dynamic voltage stability constraints, which has high accuracy and fast convergence speed.

[0006] In order to solve the above technical problems, the technical solution of the present invention is: a method for calculating the maximum power supply capacity of a power grid taking into account dynamic voltage stability constraints, comprising:

[0007] Step 1: Establish a maximum power supply capacity calculation model for a partitioned interconnected power grid taking into account dynamic voltage stability constraints, the steps include: Step 1.1: Establish an objective function for maximum power supply capacity;

[0008] Step 1.2: Establishing the constraint conditions of the maximum power supply capacity, the constraint conditions include: partition interconnection device constraint, AC grid equation constraint, node voltage amplitude constraint and dynamic voltage stability constraint; wherein, in the dynamic voltage stability constraint, an improved small disturbance analysis method is adopted, using the voltage of the load node to replace the equivalent potential in the electromagnetic power;

[0009] Step 2: Use the deep deterministic policy gradient algorithm to solve the maximum power supply capacity calculation model of the partitioned interconnected power grid taking into account the dynamic voltage stability constraint to obtain the maximum power supply capacity value.

[0010] The present invention has the following beneficial effects:

[0011] The present invention discloses the calculation of the maximum power supply capacity of a partitioned interconnected power grid taking into account the dynamic voltage stability constraint, so as to solve the problems of increasingly prominent voltage instability of the partitioned flexible interconnected power grid and the decline of the power supply capacity of the power grid. Firstly, the dynamic voltage stability is analyzed by using an improved small disturbance analysis method, and a mathematical model of the maximum power supply capacity taking into account the dynamic voltage stability index and multiple constraints is established; secondly, a maximum power supply capacity calculation method framework based on a deep deterministic policy gradient algorithm is given, and its action and state space, reward function, neural network and learning process are designed; the present invention adopts an improved small disturbance analysis method to analyze the dynamic voltage stability, simplifies the calculation process while retaining the nonlinearity of the dynamic voltage, and aims at the problems that the traditional TSC model solution method is slow and easy to fall into the local optimum, and the deep deterministic policy gradient (DDPG) algorithm is applied to adopt an actuator-evaluator structure, and neural network learning is used to improve the training stability and convergence speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the process of the present invention;

[0013] Figure 2 This is a schematic diagram of flexible interconnection of urban power grid zones;

[0014] Figure 3 It is a structural schematic diagram of a zoned flexible DC interconnection device;

[0015] Figure 4 It is the equivalent circuit of the power grid under the first-order model;

[0016] Figure 5 This is the flow chart of the DDPG algorithm;

[0017] Figure 6 Provide a geographical wiring diagram for the power grid;

[0018] Figure 7is the voltage curve of node 9 under TSC with and without considering dynamic voltage constraint;

[0019] Figure 8 This is the curve of the loss function changing with the number of training times under the DDPG algorithm. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Please refer to Figures 1 to 8 The present invention is a method for calculating the maximum power supply capacity of a power grid taking into account dynamic voltage stability constraints, which comprises the following steps:

[0022] Step 1: Establish a maximum power supply capacity calculation model for a partitioned interconnected power grid taking into account dynamic voltage stability constraints, the steps include: Step 1.1: Establish an objective function for maximum power supply capacity;

[0023] Step 1.2: Establish the constraints of the maximum power supply capacity, which include: partition interconnection device constraints, AC grid equation constraints, node voltage amplitude constraints and dynamic voltage stability constraints; among them, in the dynamic voltage stability constraint, the dynamic voltage stability constraint analysis method, i.e., the improved small disturbance analysis method, is described, using the voltage U i Replace the potential E in electromagnetic power eq2 , improve the accuracy of model solution;

[0024] Step 2: The deep deterministic policy gradient algorithm (DDPG) is used to solve the maximum power supply capacity calculation model of the partitioned interconnected power grid taking into account the dynamic voltage stability constraint, design the action and state space, reward function, neural network and learning process, and obtain the maximum power supply capacity value.

[0025] In step 1.1, assuming that zone A and zone B are connected through a zone interconnection device, the objective function of the maximum power supply capacity of the power grid is:

[0026]

[0027] Where: f TSC is the total maximum power supply capacity after the A partition and the B partition are connected, that is, TSC; i∈(A+B) means that the node i is the node of A and B; P Li is the active load of node i.

[0028] In step 1.2, the constraints of the maximum power supply capacity include: partition interconnection device constraints, AC grid equality constraints, node voltage amplitude constraints, and dynamic voltage stability constraints;

[0029] The partition interconnection device constraints in step 1.2 are:

[0030] Figure 1 This is a schematic diagram of flexible interconnection of urban power grid zones. In order to save construction land and equipment costs, the device adopts a back-to-back half-bridge MMC design. Since the loss of the zone interconnection device itself has little effect on TSC research, the loss of the device is ignored here. Figure 2 This is a schematic diagram of the structure of the grid partition flexible interconnection device. The constraint equation of the partition interconnection device is as follows:

[0031] (1) Power transfer equation for partition interconnected devices

[0032] Since the device loss is negligible, therefore:

[0033]

[0034]

[0035] Where: P s , Q s U is the active and reactive exchange value between the busbar and the device; dc ,I dc is the DC voltage and current on both sides of the device; k is the transformation ratio of the connecting transformer; U s1 is the base frequency of the busbar; U m1 is the base frequency voltage output by the MMC1 side of the device; X eq is the equivalent reactance from the device to the busbar; δ1 is U s1 and U m1 The phase angle difference.

[0036] (2) Capacity constraints of partition interconnection devices

[0037] P of the device interacting with the grid s and Q s Limited by the capacity of the device:

[0038]

[0039] Where: S max The capacity limitation of the converter at one end of the device.

[0040] (3) Dynamic reactive power output range of partition interconnection device

[0041] Assume that the upper and lower limits of the voltage modulation ratio M1 of the device are M max and M min , the base frequency voltage U output by the MMC1 side of the device m1 The upper and lower limits are U mmax and U mmin , the device should satisfy the modulation ratio and voltage constraints:

[0042]

[0043]

[0044] (4) The operating domain constraints of the partition interconnection device are:

[0045]

[0046] Where X1 represents the equivalent commutation reactance.

[0047] The constraints of the partition interconnection device should satisfy equations (1)-(6), in addition to the AC grid equality constraints, node voltage amplitude constraints and dynamic voltage stability constraints in step 1.2;

[0048] The AC power grid equality constraint in step 1.2 is:

[0049] The AC power grid equality constraint is the power flow equation constraint:

[0050]

[0051] Where: P i , Q i are respectively the active and reactive injection values ​​of node i connected to the device; P si , Q si are respectively the active and reactive values ​​absorbed by the device from node i; j∈i means that node j is a node associated with node i; U i , U j are the voltage amplitudes of nodes i and j respectively; Ω is the N-1 fault set of the power grid; and are the grid conductance and susceptance parameters under the sth fault condition respectively; θ ij is the phase angle difference between the voltages at nodes i and j.

[0052] Where P i , Q i As shown in formula (9):

[0053]

[0054] Where: P Gi , Q Gi are respectively the active and reactive power injection of the generator at node i; P Li , Q Li are the active and reactive loads of node i respectively.

[0055] The node voltage amplitude constraint in step 1.2 is:

[0056] The node voltage amplitude constraints include amplitude constraints in normal operation and after N-1, branch safety constraints, generator output constraints and upper node output constraints;

[0057] Where, where the amplitude constraint is:

[0058]

[0059] Where: are the lower and upper limits of the voltage amplitude at node i, respectively;

[0060] Among them, branch safety constraints are:

[0061]

[0062] Where: is the apparent power of the branch between node i and node j in the sth state; are the thermal stability limit powers of the branch between node i and node j respectively;

[0063] Among them, the generator output constraint and the upper node output constraint are:

[0064]

[0065] Where: are the active output and reactive output of the i-th generator in the s-th state respectively; are the lower and upper limits of the active output and reactive output of the i-th generator respectively.

[0066] The dynamic voltage stability constraint in step 1.2 is:

[0067] At a certain time section of the power grid, the part of the load side looking into the power grid is subjected to the Thevenin equivalent; the load adopts a first-order electromechanical transient induction motor parallel constant impedance model, and the equivalent circuit is as follows Figure 3 As shown. Figure 3 The left side of the dotted line is subjected to the Thevenin equivalent, and the equivalent impedance is R eq2 +jX eq2 , the equivalent potential is E eq2 .

[0068] Since the traditional small disturbance method does not consider the complexity of the actual grid load points, it only analyzes the part from the load to the grid, resulting in the same equivalent potential at each load point in the large grid. However, this paper uses TSC calculation taking into account dynamic voltage stability, and the load points are different. Therefore, an improved small disturbance analysis method is proposed, which converts the equivalent potential E eq2 The voltage U of the load node iThe voltage changes with the load, which improves the accuracy of the model solution. The improved small disturbance voltage stability state indicator is marked as L r :

[0069]

[0070] The stability criterion and The expression is:

[0071]

[0072]

[0073] Where: T j is the motor rotor inertia time constant; P m P is the mechanical load power of the motor; e is the electromagnetic power of the motor; U i is the load node voltage; K H is the capacity conversion ratio, that is, the ratio of the system capacity base value to the induction motor's own capacity base value; r2+jx2 is the rotor impedance of the motor; R eq2 +jX eq2 is the equivalent impedance; K L is the load rate; a is the percentage of the mechanical load that is independent of the speed; s0 is the initial value of the motor slip; n is the load index.

[0074] When L r >0, the node small interference voltage is stable, and L r The larger the value, the higher the stability; when L r = 0, the node is at the voltage stability critical value; when L r When <0, the node small interference voltage becomes unstable.

[0075] In step 3, in order to improve the efficiency of model solving, the deep deterministic policy gradient algorithm is used to calculate the maximum power supply capacity TSC. The specific design steps are as follows:

[0076] Step 2.1: State Definition

[0077] After the dynamic voltage stability constraint is introduced, the nonlinear characteristics of the model become more obvious and the calculation process is not easy to converge. Therefore, the DDPG algorithm can be used to improve the training stability and convergence speed through the actuator-evaluator structure and neural network learning. The continuous action search TSC is used in the interaction between the agent and the environment, which greatly improves the efficiency and accuracy of the calculation. The algorithm structure of DDPG is as follows Figure 4 shown.

[0078] The state s of the power grid agent (PGA) is defined as the power grid information matrix (PGIM) which represents the power and voltage of the power grid:

[0079]

[0080] Where: P i is the node active power; Q i is the node reactive power; V i is the node voltage amplitude; is the node voltage phase angle (i=1,2,…,n).

[0081] Step 2.2: Action selection

[0082] The action matrix of PGA is defined as:

[0083]

[0084] Where: A is the action matrix. Assume that the power grid has n load nodes, that is, A is n-dimensional; It indicates that the step size of active power growth of load node i is α, α∈(-0.1, 0.1).

[0085] To prevent PGA from always being greedy when selecting actions and to balance the exploration and learning relationships in the interaction, the ε-greedy strategy is used for the agent action a k+1 choose:

[0086]

[0087] Where: η∈A means η is the value in the action matrix; ε is a uniformly distributed random number, ε∈[0,1]; ε0 is a fixed value of the greedy strategy, ε0∈[0,1]; Q ps (s k+1 ,a) is the power grid in state s k+1 The knowledge matrix of action a, i.e., the Q-value matrix shown in formula (19), uses the “state-action” Q function as the estimated value function; rand(-0.1,0.1) represents a random value in (-0.1,0.1).

[0088]

[0089] Where: E[r|s,a] is the reward matrix consisting of the reward r in the corresponding state obtained by taking action a in state s; γ is the discount factor γ∈[0,1]; Pr(s'|s,a) is the probability of going to the next state s' when taking action a in state s; V π (s') is the value function in state s'.

[0090] Use the greedy strategy π' to obtain the optimal strategy The optimal progression of the agent state can be achieved:

[0091]

[0092] Step 2.3: Policy value function update

[0093] Update the knowledge matrix Q through the value function ps :

[0094]

[0095] Where: s′ is the state of PGA in state s after action a; γ represents the importance of PGA to subsequent rewards, usually γ∈[0,1], where γ=0.99, which increases the exploration ability of PGA.

[0096] When the scale of the power grid increases, the number of state-actions of PGA increases rapidly. Designing the Q-network as a PGA knowledge network can effectively reduce the workload of updating the Q table. The Q-network inputs PGIM and outputs the corresponding Q(s,a), selects the best action according to formula (21), and continuously progresses to reach the optimal state.

[0097] Step 2.4: Environment Modeling

[0098] 1) Operating environment definition

[0099] The operating environment is defined as the power grid simulation environment, and the mathematical model is the power flow equation as well as the TSC objective function and constraints. The power flow equation is shown in formula (22):

[0100] f(x0,λ 0 )+η1e1+η2e2+…+η n e n =0 (22)

[0101] Where: (x0,λ 0 ) is a certain initial state, x0 is the initial value of the system node voltage and amplitude, λ 0 is the basic operating point of the system 0 =[P1,P2,…,P n ] where P1, P2, …, P n is the active power of the load node; e i The unit basis vector representing the active power injection at node i, which is 1 only at position i and all other elements are 0.

[0102] 2) Reward function design

[0103] After repeated trials, the reward value is set as shown in the formula:

[0104]

[0105] In the formula, |f i (x k )| max =|ΔU k , Δδ k | max is the maximum error of the kth power flow operation; given error ε = 1e-4; f' TSC The TSC value in state s′ is reached when PGA takes action a in state s; f TSC is the TSC value of PGA in state s.

[0106] If the state s′|f i (x k )| max ≤ε、L r ≥0 and f' TSC >f TSC , assuming that the R signal is 1, it indicates that the environment encourages PGA to move in the direction of power flow convergence, dynamic voltage stability and TSC value increase; if f' TSC ≤f TSC , assuming that the R signal is -1, it indicates that PGA is more inclined to move in the direction of TSC growth; if L r <0 or |f i (x k )| max >ε, assuming that the R signal is -10, it means that PGA is more inclined to move towards power flow convergence, dynamic voltage stability and TSC growth.

[0107] Simulations are performed in BPA and Python to verify the accuracy of the proposed calculation method. This paper performs TSC calculations on two partitions A and B of a 220kV partitioned flexible interconnected power grid. The two partitions contain a total of 61 nodes, including 51 load nodes, 5 power plant nodes, 5 500kV nodes, and 56 220kV nodes. The flexible DC interconnection device is located between nodes 21 and 46, with a capacity of 800MVA. The geographical wiring diagram of the power grid is shown in the figure below. Figure 5 shown.

[0108] In order to verify the necessity of calculating the maximum power supply capacity of the designed zone interconnected power grid with dynamic voltage stability constraints, the maximum power supply capacity of the zone interconnected power grid with and without dynamic voltage stability constraints is calculated. The calculation results are shown in Table 1. And when the power grid experiences a sudden load increase of 50MW, the two cases are simulated respectively. Since the load rate of node 9 is the largest, the dynamic voltage stability of node 9 is analyzed, and the voltage curve is obtained as shown in Figure 6 shown.

[0109] Table 1: TSC table with and without dynamic voltage stability constraints

[0110]

[0111] Depend on Figure 6 As shown in Table 1, when the dynamic voltage stability constraint is not taken into account, the voltage value of node 9 is less than 0.75 (pu) for more than 1.0s from 0.37s to 2.15s, and the dynamic voltage is unstable. Therefore, dynamic voltage stability needs to be taken into account when calculating the maximum power supply capacity.

[0112] In order to verify the accuracy of the proposed improved small disturbance analysis method in analyzing dynamic voltage stability, the voltage instability mode coefficient w is used. ij (i=1,2,…,n), the traditional small interference analysis method and the improved small interference analysis method are analyzed and compared. ij If it is larger, it means that there is a greater voltage instability factor at the corresponding node.

[0113] w ij The calculation process is as follows:

[0114] W=CU=[w1,w2,…,w N ] (twenty four)

[0115] Where: C is the correction coefficient matrix of voltage amplitude in rectangular coordinate form; U = [u1,u2,…,u N ] is the vector matrix of voltage amplitude.

[0116] Considering the complexity of the model in this paper, some nodes are selected for analysis. Since the load rate of nodes 9, 27, and 56 is relatively large, nodes 9, 27, and 56 are selected for analysis. According to the calculation, 1 TSC power grid calculated by the improved small interference analysis method; 2 TSC power grid calculated by the traditional small interference analysis method. The w in the corresponding cases is given respectively. ij , see Table 2:

[0117] Table 2: Correlation ratio and voltage instability mode coefficient table

[0118]

[0119] From Table 2, we can see that the w of the TSC power grid calculated by the improved small interference method is ij Compared with the TSC power grid w calculated by the traditional small disturbance method ij It is relatively small. It can be seen that the improved small disturbance method is more accurate than the traditional small disturbance method in analyzing dynamic voltage stability and can better reflect the voltage stability of the actual power grid.

[0120] The loss function obtained under the DDPG algorithm changes with the number of training times as shown below: Figure 7 As shown. Figure 7 It is concluded that with the increase in the number of training times, the loss function and the reward value function begin to stabilize after 2500 training times, and the algorithm begins to converge gradually and obtains the optimal solution at 3767 times. Therefore, the maximum power supply capacity calculation of the partitioned interconnected power grid taking into account the dynamic voltage stability constraint can effectively avoid voltage instability in the future development of the power grid.

[0121] The present invention discloses the calculation of the maximum power supply capacity of a partitioned interconnected power grid taking into account the dynamic voltage stability constraint, to solve the problems of increasingly prominent voltage instability and reduced power supply capacity of a partitioned flexible interconnected power grid. Firstly, the dynamic voltage stability is analyzed by using an improved small disturbance analysis method, and a mathematical model of the maximum power supply capacity taking into account the dynamic voltage stability index and multiple constraints is established; secondly, a maximum power supply capacity calculation method framework based on a deep deterministic policy gradient algorithm is given, and its action and state space, reward function, neural network and learning process are designed; finally, a certain power grid is used as an example for verification, and the method can effectively optimize the maximum power supply capacity in the partitioned interconnected power grid under the condition of voltage fluctuation.

[0122] The parts not involved in the present invention are the same as the prior art or are implemented by using the prior art.

[0123] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A method for calculating the maximum power supply capacity of a power grid taking into account dynamic voltage stability constraints, characterized in that: include Step 1: Establish a maximum power supply capacity calculation model for the regional interconnected power grid taking into account dynamic voltage stability constraints, including: Step 1.1: Establish the objective function of maximum power supply capability; Step 1.2: Establishing the constraint conditions of the maximum power supply capacity, the constraint conditions include: partition interconnection device constraint, AC grid equation constraint, node voltage amplitude constraint and dynamic voltage stability constraint; wherein, in the dynamic voltage stability constraint, the voltage of the load node is used to replace the equivalent potential in the electromagnetic power; Step 2: Use the deep deterministic policy gradient algorithm to solve the maximum power supply capacity calculation model of the partitioned interconnected power grid taking into account the dynamic voltage stability constraint to obtain the maximum power supply capacity value; In step 1.1, the objective function of the maximum power supply capacity is: Assuming that the A zone and the B zone are connected through a zone interconnection device, the objective function of the maximum power supply capacity of the power grid is: Where: f TSC is the total maximum power supply capacity after the A partition and the B partition are connected, that is, TSC; i∈(A+B) means that the node i is the node of A and B; P Li is the active load of node i; In the step 1.2, the partition interconnection device constraints are: the power transmission equation of the partition interconnection device, the capacity constraint of the partition interconnection device, the dynamic reactive power output range of the partition interconnection device and the operation domain constraint of the partition interconnection device; Among them, the power transmission equation of the partition interconnection device is: Since the device loss is negligible, therefore: Where: P s , Q s U is the active and reactive exchange value between the busbar and the device; dc ,I dc is the DC voltage and current on both sides of the device; k is the transformation ratio of the connecting transformer; U s1 is the base frequency of the busbar; U m1 is the base frequency voltage output by the MMC1 side of the device; X eq is the equivalent reactance from the device to the busbar; δ1 is U s1 and U m1 The phase angle difference; Among them, the capacity constraint of the partition interconnection device is: P of the device interacting with the grid s and Q s Limited by the capacity of the device: Where: S max The capacity limitation of the converter at one end of the device; Among them, the dynamic reactive power output range of the partition interconnection device is: Assume that the upper and lower limits of the voltage modulation ratio M1 of the device are M max and M min , the base frequency voltage U output by the MMC1 side of the device m1 The upper and lower limits are U mmax and U mmin , the device should satisfy the modulation ratio and voltage constraints: Among them, the operating domain constraints of the partition interconnection device are: In the formula, X1 represents the equivalent commutation reactance; The AC power grid equation constraint is the power flow equation constraint, specifically: Where: P i , Q i are respectively the active and reactive injection values ​​of node i connected to the device; P si , Q si are respectively the active and reactive values ​​absorbed by the device from node i; j∈i means that node j is a node associated with node i; U i , U j are the voltage amplitudes of nodes i and j respectively; Ω is the N-1 fault set of the power grid; and are the grid conductance and susceptance parameters under the sth fault condition respectively; θ ij is the phase angle difference between the voltages at nodes i and j; Where P i , Q i As shown in formula (9): Where: P Gi , Q Gi are respectively the active and reactive power injection of the generator at node i; P Li , Q Li are the active and reactive loads of node i respectively; The node voltage amplitude constraints include amplitude constraints in normal operation and after N-1, branch safety constraints, generator output constraints and upper node output constraints; Among them, the amplitude constraint is: Where: are the lower and upper limits of the voltage amplitude at node i, respectively; Among them, the branch safety constraints are: Where: is the apparent power of the branch between node i and node j in the sth state; are the thermal stability limit powers of the branch between node i and node j respectively; Among them, the generator output constraint and the upper node output constraint are: Where: are the active output and reactive output of the i-th generator in the s-th state respectively; are the lower and upper limits of the active and reactive outputs of the i-th generator respectively; In the dynamic voltage stability constraint, an improved small disturbance analysis method is proposed to convert the equivalent potential E eq2 The voltage U of the load node i Instead, the voltage changes with the load, improving the accuracy of the model solution; The improved small disturbance voltage stability state indicator is marked as L r : Among them, the stability criterion and The expression is: Where: T j is the motor rotor inertia time constant; P m P is the mechanical load power of the motor; e is the electromagnetic power of the motor; U i is the load node voltage; K H is the capacity conversion ratio, that is, the ratio of the system capacity base value to the induction motor's own capacity base value; r2+jx2 is the rotor impedance of the motor; R eq2 +jX eq2 is the equivalent impedance; K L is the load rate; a is the percentage of the mechanical load that is not related to the speed; s0 is the initial value of the motor slip; n is the load index; When L r >0, the node small interference voltage is stable, and L r The larger the value, the higher the stability; when L r = 0, the node is at the voltage stability critical value; when L r When <0, the node small interference voltage becomes unstable.

2. The method for calculating the maximum power supply capacity of a power grid taking into account dynamic voltage stability constraints according to claim 1, characterized in that: The steps for calculating the maximum power supply capacity TSC using the deep deterministic policy gradient algorithm are as follows: Step 2.1: State definition, The state s of the power grid agent PGA is the power grid information matrix PGIM, which represents the power and voltage of the power grid: Where: P i is the node active power; Q i is the node reactive power; V i is the node voltage amplitude; is the node voltage phase angle (i=1,2,…,n); Step 2.2: Action selection The action matrix of PGA is defined as: Where: A is the action matrix. Assume that the power grid has n load nodes, that is, A is n-dimensional; Indicates that the step size of active power growth of load node i is α, α∈(-0.1, 0.1); Use ε-greedy strategy to perform agent action a k+1 choose: Where: η∈A means η is the value in the action matrix; ε is a uniformly distributed random number, ε∈[0,1]; ε0 is a fixed value of the greedy strategy, ε0∈[0,1]; Q ps (s k+1 ,a) is the power grid in state s k+1 Take the knowledge matrix of action a, that is, the Q value matrix shown in formula (19), and use the "state-action" Q function as the estimated value function; rand (-0.1, 0.1) represents a random value in (-0.1, 0.1); Where: E[r|s,a] is the reward matrix consisting of the reward r in the corresponding state obtained by taking action a in state s; γ is the discount factor γ∈[0,1]; Pr(s's,a) is the probability of going to the next state s' when taking action a in state s; V π (s') is the value function in state s'; Use the greedy strategy π' to obtain the optimal strategy The optimal progression of the agent state can be achieved: Step 2.3: Policy value function update Update the knowledge matrix Q through the value function ps : Where: s′ is the state of PGA in state s after action a; γ represents the importance of PGA to subsequent rewards, usually γ∈[0,1], where γ=0.99, which increases the exploration ability of PGA; When the scale of the power grid increases, the number of state-actions of PGA increases rapidly. Designing the Q-network as a PGA knowledge network can effectively reduce the workload of Q-table update. The Q-network inputs PGIM and outputs the corresponding Q(s,a), selects the best action according to formula (21), and continuously progresses to reach the optimal state. Step 2.4: Environment Modeling 1) Operating environment definition The operating environment is defined as the power grid simulation environment, and the mathematical model is the power flow equation as well as the TSC objective function and constraints. The power flow equation is shown in formula (22): f(x0,λ 0 )+η1e1+η2e2+…+η n e n =0(22) Where: (x0,λ 0 ) is a certain initial state, x0 is the initial value of the system node voltage and amplitude, λ 0 is the basic operating point of the system 0 =[P1,P2,…,P n ] where P1, P2, …, P n is the active power of the load node; e i The unit basis vector representing the active power injection at node i is 1 only at position i, and all other elements are 0; 2) Reward function design Set the reward value as shown: In the formula, |f i (x k ) max =|ΔU k , Δδ k max is the maximum error of the kth power flow operation; given error ε=1e-4; f' TSC The TSC value in state s′ is reached when PGA takes action a in state s; f TSC is the TSC value of PGA in state s; If the state s′|f i (x k ) max ≤ε、L r ≥0 and f' TSC >f TSC , assuming that the R signal is 1, it indicates that the environment encourages PGA to move in the direction of power flow convergence, dynamic voltage stability and TSC value increase; if f' TSC ≤f TSC , assuming that the R signal is -1, it indicates that PGA is more inclined to move in the direction of TSC growth; if L r <0 or |f i (x k ) max >ε, assuming that the R signal is -10, it means that PGA is more inclined to move towards power flow convergence, dynamic voltage stability and TSC growth.

Images