Power system reliability evaluation method based on loss-free load state set

Abstract

The invention provides a power system reliability evaluation method based on a loss-free load state set, which adopts an extended cross entropy important sampling method and is divided into a pre-sampling stage and a formal sampling stage, and in the two stages, the loss-free load state set is used for calculating the loss load capacity of the state, so that the calculation amount is reduced. According to the method, the loss-free load state set realized by the hash table is designed, and the efficiency of reliability evaluation of the power system with continuous random variables in the state by using the extended cross entropy important sampling method is improved by reducing the number of times of solving the direct current load reduction model by using the loss-free load state set.

Description

Technical Field

[0001] The invention relates to the field of power systems, and in particular to a power system reliability assessment method based on a no-load loss state set. Background Art

[0002] Currently, there are two main methods for evaluating the reliability of power systems: analytical method and Monte Carlo method. The analytical method enumerates each state for calculation, which has the advantage of high accuracy. However, when evaluating the reliability of large power systems, as the number of system components increases, the number of states will grow exponentially, and the time consumption will also increase significantly. For example, a system containing N two-state components has 2 N There are 3 possible states. Therefore, the analytical method is more suitable for small power systems.

[0003] Traditional Monte Carlo methods can be divided into sequential Monte Carlo and non-sequential Monte Carlo. The sequential Monte Carlo method needs to consider factors that change over time, such as wind power generation, solar power generation, and system load changes, so its sampling process is relatively complex. There are two sampling methods in the sequential Monte Carlo method, the state duration method and the state transition method. The sampling process of the non-sequential Monte Carlo method is relatively simple, but it ignores the time series change process and cannot calculate indicators related to time and frequency. In order to obtain higher-precision indicators, the Monte Carlo method usually requires a long simulation time.

[0004] The instability of renewable energy leads to a significant decrease in the efficiency of the Monte Carlo method. If the state is continuous, the probability of sampling the same state in the sampling process is 0, and it is impossible to use the method of judging whether the state is repeated to improve efficiency. In response to this problem, the present invention proposes a set of no-loss load states based on the inclusion relationship between the feasible domains of the planning problem. By using the set of no-loss load states, it is determined whether each state needs to solve the load reduction model, reducing the number of solutions and improving efficiency. It is implemented using a hash table and combined with the cross-entropy importance sampling method and parallel computing. Reliability evaluation is performed on the RBTS and RTS79 systems, and the set of no-loss load states implemented using a hash table effectively improves the evaluation efficiency of systems with low load levels. Summary of the invention

[0005] The purpose of the present invention is to propose a power system reliability assessment method based on a no-load loss state set, design a no-load loss state set, and use the no-load loss state set to reduce the number of times a DC load reduction model is solved, thereby accelerating the efficiency of reliability assessment of a power system with continuous random variables in the state using an extended cross-entropy importance sampling method.

[0006] To achieve the above object, the technical solution of the present invention is: a method for evaluating the reliability of a power system based on a no-load loss state set, specifically comprising the following steps:

[0007] Step S1: Set the initial parameters of pre-sampling, including the number of samples N k , quantile parameter ρ, smoothing parameter α, original probability density or probability mass function g(x; Q0) = f(x; P), number of iterations k = 0, no load loss state set Among them, x represents a system state, Q0 represents the parameter vector of function g, and P represents the parameter vector of function f;

[0008] Step S2: Iteration number k++, calculate the likelihood ratio W(x; P, Q k-1 ), let the list to be calculated U s is empty;

[0009] Step S3: Using the probability density or probability mass function g(x; Q k-1 ) k Sampling times, we get N k System Status

[0010] Step S4: Calculate N in step S3 k The amount of load loss in each system state;

[0011] Step S5: Calculate the additional threshold parameter γ according to the calculation result of step S4 k And update the parameter Q k ;

[0012] Step S6: If γ k = 0, then let the parameter vector Q of function g be opt =Q k End the pre-sampling and enter the formal sampling, otherwise return to step S2;

[0013] Step S7: Set the initial parameters for formal sampling, including the number of samples N for each round of sampling m , probability density or probability mass function g(x;Q opt ), sampling round number m = 0, maximum value of variance coefficient β max , let list l LOLP and l EPNS is empty; and calculates the likelihood ratio W(x; P, Q opt );

[0014] Step S8: Sampling round number m++, let the list to be calculated U s Empty, with probability density or probability mass function g(x; Q opt ) m Sampling times, we get N m System Status

[0015] Step S9: Calculate N in step S8 m The load loss amount of each system state is obtained to obtain the load loss amount list R;

[0016] Step S10: traverse N m For the qth system state, the system state extracted in the previous m-1 rounds plus the estimated reliability indicators LOLP and EPNS of the first q system states in this round are added to the list l LOLP and l EPNS middle;

[0017] Step S11: Calculate the variance coefficient β. If β EPNS <β max Then end; otherwise return to step S8;

[0018] Step S12: From list 1 LOLP and l EPNS The last digit of obtains the reliability indicators LOLP and EPNS.

[0019] Preferably, the likelihood ratio W(x; P, Q k-1 ) and W(x;P,Q opt ) is calculated as:

[0020]

[0021] Where Q is Q k-1 or Q opt , f(x; P) is the original probability density or probability mass function.

[0022] Preferably, the steps of calculating the load loss amount in step S4 and step S9 are the same, and specifically include the following steps:

[0023] Step S31: traverse all sampled states x, and check whether there is a feasible domain of a state in the no-load state set A that is included in the feasible domain corresponding to state x; if so, mark state x as no-load; otherwise, add state x to the list U s middle;

[0024] Step S32: s The states in the CPU are divided into core_num groups, where core_num is the number of CPU cores; each group of states corresponds to a core, and the load is calculated on this core. The calculation results on each core are calculated according to the CPU core number in U s The order in is combined together to get the list R0;

[0025] Step S33: traverse U s For all states x in , if the load loss corresponding to state x is f LCIf it is 0, insert state x into A; otherwise, continue to step S34;

[0026] Step S34: R0 and the load loss amounts marked as no load loss in step S31 are combined in the order of sampling to obtain a sequence R, and R is returned.

[0027] Preferably, the construction of the no-load-loss state set A is specifically as follows:

[0028] Each time the load loss in a certain state is calculated, the DC load reduction model is used to solve the planning problem corresponding to the state:

[0029]

[0030] Among them, P d is the busbar load, P g is the active output of the generator, C g is the generator connection matrix, θ is the voltage phase angle of the node, P gmin and P gmax are the minimum and maximum active output of the generator, Y ft and Y bus are the branch admittance matrix and the node admittance matrix respectively, F min and F max are the minimum and maximum values ​​of branch power flow, P LC is the active power of load reduction, f LC is the objective function;

[0031] If f LC = 0, this state is called no load loss, and this state is inserted into A;

[0032] The query in step S31 whether the feasible domain of the state in A is included in the feasible domain corresponding to the state x is specifically:

[0033] If there is a feasible region in the load reduction model without load loss in A that is included in the feasible region of the load reduction model of the state x sampled this time, then the state x sampled this time also has no load loss;

[0034] The basis for judging whether the feasible domain is included is: if there is a discrete component of state g in the no-load loss state set A that is the same as the discrete component of state x, and the continuous components of state y are all smaller than the continuous components of state x, then it means that the feasible domain of state y is included in the feasible domain of state x.

[0035] Preferably, in step S33, the state x is inserted into the load-free state set A in A by using a hash table, and the system state x is represented as a structure, which contains x d, Power two member variables; among them, x d is the discrete component of state x, and Power is the value of the continuous random variable in the component through the wind power output model function P, that is,

[0036]

[0037] Among them, x (i) is the i-th component of x, the first n components in x d The components are discrete components; this hash table is denoted as Hash.

[0038] Preferably, in step S33, the state x is inserted into A, and the specific implementation method is as follows:

[0039] Step S331: Calculate the index value s=h(x) for state x d ), query x in the hash table Hash by the index value and delete the state whose corresponding feasible domain in A is included in the feasible domain of state x;

[0040] Step S332: In the linked list with Hash[s] as the head pointer, insert p.Power=P(x c ),px=x d ,p.next=Hash[s],Hash[s]=p;

[0041] The query operation and the delete operation are performed simultaneously, which specifically includes the following steps:

[0042] Step S331.1: Calculate the index value s=h(x d ), h(x d ) is as follows:

[0043]

[0044] Among them, prime refers to a large prime number, M i Determined by the following expression

[0045]

[0046] n d is the number of discrete random variables, n r is the number of values ​​of the rth discrete random variable;

[0047] Step S331.2: traverse the linked list with Hash[s] as the head pointer, where It is the node traversed, x c is the continuous component of state x, if Then delete It and point It to the next node;

[0048] Step S331.3: If It is empty, end; otherwise, continue with step S331.2.

[0049] Preferably, the step S5 specifically includes the following steps:

[0050] Step S51: Calculate the power generation capacity margin for each state x using formula (2)

[0051]

[0052] Among them, P g is the sum of the available capacities of the generators in state x, L d is the system load at state x, L c =f LC is the load loss in state x when there is load loss, Ω success N k The set of states without load loss in the states, Ω failure N k The set of load loss states in the state;

[0053] Step S52: Arrange the power generation capacity margin sequence from small to large to obtain

[0054] Step S53: If the first Elements Then let γ k =0, otherwise let

[0055] Step S54: Correct M(x) to M′(x)=M(x)-γ k , and get the sequence

[0056] Step S55: Calculate parameters using formula (3) to formula (9). The quantities calculated by formula (3) to formula (9) are all parameters in V, where Indicates the nth state x drawn in this round of sampling n The rth component of nd For state x n The discrete components of

[0057] Replace the indicator function H(x) of the reliability index in formula (4) to formula (9) with H M (x)W(x;P,Q k-1 ), and we can get V by substituting it into the calculation:

[0058]

[0059] For state x nThe discrete components x nd :

[0060]

[0061] in, is the indicator function, when x nd The i-th component of takes the j-th value and is 1, otherwise it is 0. The expression is as follows:

[0062]

[0063] If the state x n The rth component of Follows the Weibull distribution:

[0064]

[0065] Where p is a fixed parameter;

[0066] If the state x n The rth component of Follows a lognormal distribution:

[0067]

[0068]

[0069] If the state x n The rth component of Obeying Rayleigh distribution:

[0070]

[0071] Step S56: Calculate parameter Q k :

[0072] Q k =αV+(1-α)Q k-1 (10).

[0073] Preferably, equations (4) to (9) are obtained by minimizing H(g opt (x), g(x; Q)) is solved to obtain H(g opt (x), g(x; Q)) is the cross entropy, g(x; Q) is the unknown function to be solved, g opt (x) is obtained from the following formula:

[0074]

[0075] Among them, H(x) is the indicator function of a certain reliability index, l is the value of the reliability index of the system, which is a constant;

[0076] The following optimization problem needs to be solved

[0077]

[0078] Where Q refers to any parameter vector and Θ is the parameter space; the constraints here are:

[0079]

[0080] Using the Lagrange multiplier method, we can solve equations (4) to (9), where n d represents the number of discrete components of the system state x, represents the probability of the i-th discrete component in the j-th case, n i Represents the total number of cases corresponding to the i-th discrete component.

[0081] Preferably, the calculation formulas of LOLP and EPNS in step S10 are:

[0082] (1) Load loss probability LOLP, which indicates the probability of load loss in the system:

[0083]

[0084]

[0085] (2) Expected power shortage EPNS, which indicates the expected value of power shortage in the system:

[0086]

[0087] Among them, N sum The number of states that need to be considered for calculating the reliability index, X v The vth state of the reliability index is calculated this time; W(x v ) is the likelihood ratio W(x; P, Q) calculated in step S7 opt );ΔP is the power shortage corresponding to state x, Ω success N sum The set of states without load loss in the states, Ω failure N sum The set of load loss states in the state.

[0088] Preferably, the calculation formula of the variance coefficient β in step S11 is:

[0089]

[0090] in, and n 2 Determined by the following formula:

[0091]

[0092] in, It indicates the average reliability index when the sampling reaches the qth sample in this round. There are N samples in this round. m times; N sum The number of states that need to be considered for calculating the reliability index, x v This is the vth state of the reliability index calculated this time; H(x) is the indicator function of the reliability index, using H LOLP (x) or N EPNS (x).

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] The present invention designs a no-load loss state set, and utilizes the no-load loss state set to reduce the number of times a DC load reduction model is solved, thereby accelerating the efficiency of reliability assessment of a power system with continuous random variables in the state using an extended cross-entropy importance sampling method. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 It is a flow chart of the power system reliability assessment based on the no-load loss state set and the hash table implementation method thereof of the present invention;

[0096] Figure 2 It is a structural diagram of the hash table of the load-loss-free state set of the present invention. DETAILED DESCRIPTION

[0097] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0098] This embodiment provides a method for evaluating the reliability of a power system based on a load-free state set and implementing the same in a hash table. Figure 1 As shown, it is divided into two stages: pre-sampling and formal sampling. In the steps of calculating the loss of load in the two stages, the state set without loss of load is used to reduce the amount of calculation, which specifically includes the following steps:

[0099] Step S1: Set the initial parameters of pre-sampling, including the number of samples N k , quantile parameter ρ, smoothing parameter α, original probability density or probability mass function g(x; Q0) = f(x; P), number of iterations k = 0, no load loss state set Among them, x represents a system state, Q0 represents the parameter vector of function g, and P represents the parameter vector of function f;

[0100] Step S2: Iteration number k++, calculate the likelihood ratio W(x; P, Q k-1 ), let the list to be calculated U s is empty;

[0101] Step S3: Using the probability density or probability mass function g(x; Q k-1 ) k Sampling times, we get N k System Status

[0102]

[0103] Step S4: Calculate N in step S3 k The amount of load loss in each system state;

[0104] Step S5: Calculate the additional threshold parameter γ according to the calculation result of step S4 k And update the parameter Q k ;

[0105] Step S6: If γ k = 0, then let the parameter vector Q of function g be opt =Q k End the pre-sampling and enter the formal sampling, otherwise return to step S2;

[0106] Step S7: Set the initial parameters for formal sampling, including the number of samples N for each round of sampling m , probability density or probability mass function g(x;Q opt ), sampling round number m = 0, maximum value of variance coefficient β max , let list l LOLP and l EPNS is empty; and calculates the likelihood ratio W(x; P,Q opt );

[0107] Step S8: Sampling round number m++, let the list to be calculated U s Empty, with probability density or probability mass function g(x; Q opt ) m Sampling times, we get N m System Status

[0108] Step S9: Calculate N in step S8 m The load loss amount of each system state is obtained to obtain the load loss amount list R;

[0109] Step S10: Traverse N m For the qth system state, the system state extracted in the previous m-1 rounds plus the estimated reliability indicators LOLP and EPNS of the first q system states in this round are added to the list l LOLP and l EPNS middle;

[0110] Step S11: Calculate the variance coefficient β. If β EPNS <β max Then end; otherwise return to step S8;

[0111] Step S12: From list 1 LOLP and l EPNS The last digit of obtains the reliability indexes LOLP and EPNS;

[0112] In this embodiment, the likelihood ratio W(x; P, Q k-1 ) and W(x;P,Q opt ) is calculated as:

[0113]

[0114] Where Q is Q k-1 or Q opt , f(x; P) is the original probability density or probability mass function.

[0115] In this embodiment, the steps of calculating the load loss amount in step S4 and step S9 are the same, and specifically include the following steps:

[0116] Step S31: traverse all sampled states x, and check whether there is a feasible domain of a state in the no-load state set A that is included in the feasible domain corresponding to state x; if so, mark state x as no-load; otherwise, add state x to the list U s middle;

[0117] Step S32: s The states in the CPU are divided into core_num groups, where core_num is the number of CPU cores; each group of states corresponds to a core, and the load is calculated on this core. The calculation results on each core are calculated according to the CPU core number in U s The order in is combined together to get the list R0;

[0118] Step S33: traverse U s For all states x in , if the load loss corresponding to state x is f LC If it is 0, insert state x into A; otherwise, continue to step S34;

[0119] Step S34: R0 and the load loss amounts marked as no load loss in step S31 are combined in the order of sampling to obtain a sequence R, and R is returned.

[0120] In this embodiment, the construction of the no-load-loss state set A is specifically as follows:

[0121] Each time the load loss in a certain state is calculated, the DC load reduction model is used to solve the planning problem corresponding to the state:

[0122]

[0123] Among them, P d is the busbar load, P g is the active output of the generator, C g is the generator connection matrix, θ is the voltage phase angle of the node, P g min and P g max are the minimum and maximum active output of the generator, Y ft and Y bus are the branch admittance matrix and the node admittance matrix respectively, F min and F max are the minimum and maximum values ​​of branch power flow, P LC is the active power of load reduction, f LC is the objective function;

[0124] If f LC = 0, this state is called no load loss, and this state is inserted into A;

[0125] The query in step S31 whether the feasible domain of the state in A is included in the feasible domain corresponding to the state x is specifically:

[0126] If there is a feasible region in the load reduction model without load loss in A that is included in the feasible region of the load reduction model of the state x sampled this time, then the state x sampled this time also has no load loss;

[0127] The basis for judging whether the feasible domain is included is: if there is a discrete component of state y in the no-load loss state set A that is the same as the discrete component of state x, and the continuous components of state y are all smaller than the continuous components of state x, then it means that the feasible domain of state y is included in the feasible domain of state x.

[0128] In this embodiment, the step S33 inserts the state x into the load-free state set A in A using a hash table. The system state x is represented as a structure, which contains x d , Power two member variables; among them, x d is the discrete component of state x, and Power is the value of the continuous random variable in the component through the wind power output model function P, that is,

[0129]

[0130] Among them, x (i)is the i-th component of x, the first n components in x d The components are discrete components; this hash table is denoted as Hash.

[0131] In this embodiment, the state x is inserted into A in step S33, and the specific implementation method is as follows:

[0132] Step S331: Calculate the index value s=h(x) for state x d ), query x in the hash table Hash by the index value and delete the state whose corresponding feasible domain in A is included in the feasible domain of state x;

[0133] Step S332: In the linked list with Hash[s] as the head pointer, insert p.Power=P(x c ),px=x d ,p.next=Hash[s],Hash[s]=p;

[0134] The query operation and the delete operation are performed simultaneously, which specifically includes the following steps:

[0135] Step S331.1: Calculate the index value s=h(x d ), h(x d ) is as follows:

[0136]

[0137] Among them, prime refers to a large prime number, M i Determined by the following expression

[0138]

[0139] n d is the number of discrete random variables, n r is the number of values ​​of the rth discrete random variable;

[0140] Step S331.2: traverse the linked list with Hash[s] as the head pointer, where It is the node traversed, x c is the continuous component of state x, if Then delete It and point It to the next node;

[0141] Step S331.3: If It is empty, end; otherwise, continue with step S331.2.

[0142] In this embodiment, step S5 specifically includes the following steps:

[0143] Step S51: Calculate the power generation capacity margin for each state x using formula (2)

[0144]

[0145] Among them, P g is the sum of the available capacities of the generators in state x, L d is the system load at state x, L c =f LC is the load loss in state x when there is load loss, Ω success N k The set of states without load loss in the states, Ω faillure N k The set of load loss states in the states;

[0146] Step S52: Arrange the power generation capacity margin sequence from small to large to obtain

[0147] Step S53: If the first Elements Then let γ k =0, otherwise let

[0148] Step S54: Correct M(x) to M′(x)=M(x)-γ k , and get the sequence

[0149] Step S55: Calculate parameters using formula (3) to formula (9). The quantities calculated by formula (3) to formula (9) are all parameters in V, where Indicates the nth state x drawn in this round of sampling n The rth component of nd For state x n The discrete components of

[0150] Replace the indicator function H(x) of the reliability index in formula (4) to formula (9) with H M (x)W(x;P,Q k-1 ), and we can get V by substituting it into the calculation:

[0151]

[0152] For state x n The discrete components x nd :

[0153]

[0154] in, is the indicator function, when x nd The i-th component of takes the j-th value and is 1, otherwise it is 0. The expression is as follows:

[0155]

[0156] If the state x n The rth component of Follows the Weibull distribution:

[0157]

[0158] Where p is a fixed parameter;

[0159] If the state x n The rth component of Follows a lognormal distribution:

[0160]

[0161]

[0162] If the state x n The rth component of Obeying Rayleigh distribution:

[0163]

[0164] Step S56: Calculate parameter Q k :

[0165] Q k =αV+(1-α)Q k-1 (10).

[0166] In this embodiment, equations (4) to (9) are obtained by minimizing H(g opt (x), g(x; Q) is solved to obtain H(g opt (x), g(x; Q)) is the cross entropy, g(x; Q) is the unknown function to be solved, g opt (x) is obtained from the following formula:

[0167]

[0168] Among them, H(x) is the indicator function of a certain reliability index, l is the value of the reliability index of the system, which is a constant;

[0169] The following optimization problem needs to be solved

[0170]

[0171] Where Q refers to any parameter vector and Θ is the parameter space; the constraints here are:

[0172]

[0173] Using the Lagrange multiplier method, we can solve equations (4) to (9), where n d represents the number of discrete components of the system state x, represents the probability of the i-th discrete component in the j-th case, n i Represents the total number of cases corresponding to the i-th discrete component.

[0174] In this embodiment, the calculation formulas of LOLP and EPNS in step S10 are:

[0175] (1) Load loss probability LOLP, which indicates the probability of load loss in the system:

[0176]

[0177]

[0178] (2) Expected power shortage EPNS, which indicates the expected value of power shortage in the system:

[0179]

[0180] Among them, N sum The number of states that need to be considered for calculating the reliability index, X v The vth state of the reliability index is calculated this time; W(x v ) is the likelihood ratio W(x; P, Q) calculated in step S7 opt );ΔP is the power shortage corresponding to state x, Ω success N sum The set of states without load loss in the states, Ω failure N sum The set of load loss states in the state.

[0181] In this embodiment, the calculation formula of the variance coefficient β in step S11 is:

[0182]

[0183] in, and n 2 Determined by the following formula:

[0184]

[0185] in, It indicates the average reliability index when the sampling reaches the qth sample in this round. There are N samples in this round. m times; N sum The number of states that need to be considered for calculating the reliability index, x vThis is the vth state of the reliability index calculated this time; H(x) is the indicator function of the reliability index, using H LOLP (x) or N EPNS (x).

[0186] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto, and any changes or substitutions that are not conceived through creative work should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the reliability of a power system based on a no-load loss state set, characterized in that: The specific steps include: Step S1: Set the initial parameters of pre-sampling, including the number of samples N k , quantile parameter ρ, smoothing parameter α, original probability density or probability mass function g(x; Q0) = f(x; P), number of iterations k = 0, no load loss state set Among them, x represents a system state, Q0 represents the parameter vector of function g, and P represents the parameter vector of function f; Step S2: Iteration number k++, calculate the likelihood ratio W(x; P, Q k-1 ), let the list to be calculated U s is empty; Step S3: Using the probability density or probability mass function g(x; Q k-1 ) k Sampling times, we get N k System Status Step S4: Calculate N in step S3 k The amount of load loss in each system state; Step S5: Calculate the additional threshold parameter γ according to the calculation result of step S4 k And update the parameter Q k ; Step S6: If γ k = 0, then let the parameter vector Q of function g be opt =Q k End the pre-sampling and enter the formal sampling, otherwise return to step S2; Step S7: Set the initial parameters for formal sampling, including the number of samples N for each round of sampling m , probability density or probability mass function g(x;Q opt ), sampling round number m = 0, maximum value of variance coefficient β max , let list l LoLP and l EPNS is empty; and calculates the likelihood ratio W(x; P, Q opt ); Step S8: Sampling round number m++, let the list to be calculated U s Empty, with probability density or probability mass function g(x; Q opt ) m Sampling times, we get N m System Status Step S9: Calculate N in step S8 m The load loss amount of each system state is obtained to obtain the load loss amount list R; Step S10: Traverse N m For the qth system state, the system state extracted in the previous m-1 rounds plus the estimated reliability indicators LOLP and EPNS of the first q system states in this round are added to the list l LOLP and l EPNS middle; Step S11: Calculate the variance coefficient β. If β EPNS <β max Then end; otherwise return to step S8; Step S12: From list 1 LOLP and l EPNS The last digit of obtains the reliability indicators LOLP and EPNS.

2. A method for evaluating the reliability of a power system based on a no-load loss state set according to claim 1, characterized in that: The likelihood ratio W(x; P, Q k-1 ) and W(x;P,Q opt ) is calculated as: Where Q is Q k-1 or Q opt , f(x; P) is the original probability density or probability mass function.

3. The method for evaluating the reliability of a power system based on a load-free state set according to claim 1, characterized in that: The steps for calculating the load loss amount in step S4 and step S9 are the same, and specifically include the following steps: Step S31: traverse all sampled states x, and query whether there is a feasible domain of a state in the no-load loss state set A that is included in the feasible domain corresponding to state x; if so, mark state x as no-load loss; Otherwise, add state x to list U s middle; Step S32: s The states in the CPU are divided into core_num groups, where core_num is the number of CPU cores; each group of states corresponds to a core, and the load is calculated on this core. The calculation results on each core are calculated according to the CPU core number in U s The order in is combined together to obtain list R0; Step S33: traverse U s For all states x in , if the load loss corresponding to state x is f LC is 0, then insert state x into A; Otherwise, continue to step S34; Step S34: R0 and the load loss amounts marked as no load loss in step S31 are combined in the order of sampling to obtain a sequence R, and R is returned.

4. A method for evaluating the reliability of a power system based on a no-load loss state set according to claim 3, characterized in that: The construction of the no-load-loss state set A is specifically as follows: Each time the load loss in a certain state is calculated, the DC load reduction model is used to solve the planning problem corresponding to the state: Among them, P d is the busbar load, P g is the active output of the generator, C g is the generator connection matrix, θ is the voltage phase angle of the node, P gmin and P gmax are the minimum and maximum active output of the generator, Y ft and Y bus are the branch admittance matrix and the node admittance matrix respectively, F min and F max are the minimum and maximum values ​​of branch power flow, P LC is the active power of load reduction, f LC is the objective function; If f LC = 0, this state is called no load loss, and this state is inserted into A; The query in step S31 whether the feasible domain of the state in A is included in the feasible domain corresponding to the state x is specifically: If there is a feasible region in the load reduction model without load loss in A that is included in the feasible region of the load reduction model of the state x sampled this time, then the state x sampled this time also has no load loss; The basis for judging whether the feasible domain is included is: if there is a discrete component of state y in the no-load loss state set A that is the same as the discrete component of state x, and the continuous components of state y are all smaller than the continuous components of state x, then it means that the feasible domain of state y is included in the feasible domain of state x.

5. The method for evaluating the reliability of a power system based on a load-free state set according to claim 3, characterized in that: In step S33, the state x is inserted into the load-free state set A in A by using a hash table. The system state x is represented as a structure, which contains x d , Power two member variables; among them, x d is the discrete component of state x, and Power is the value of the continuous random variable in the component through the wind power output model function P, that is, Among them, x (i) is the i-th component of x, the first n components in x d The components are discrete components; this hash table is denoted as Hash.

6. A method for evaluating the reliability of a power system based on a no-load loss state set according to claim 5, characterized in that: In step S33, state x is inserted into A, and the specific implementation method is as follows: Step S331: Calculate the index value s=h(x) for state x d ), query x in the hash table Hash by the index value and delete the state whose corresponding feasible domain in A is included in the feasible domain of state x; Step S332: In the linked list with Hash[s] as the head pointer, insert p.Power=P(x c ), px=x d ,p.next=Hash[s],Hash[s]=p; The query operation and the delete operation are performed simultaneously, which specifically includes the following steps: Step S331.1: Calculate the index value s=h(x d ), h(x d ) is as follows: Here, prime refers to a large prime number, and M is determined by the following expression: n d is the number of discrete random variables, n r is the number of values ​​of the rth discrete random variable; Step S331.2: traverse the linked list with Hash[s] as the head pointer, where It is the node traversed, x c is the continuous component of state x, if Then delete It and point It to the next node; Step S331.3: If It is empty, end; otherwise, continue with step S331.

2.

7. A method for evaluating the reliability of a power system based on a no-load loss state set according to claim 3, characterized in that: The step S5 specifically comprises the following steps: Step S51: Calculate the power generation capacity margin for each state x using formula (2) Among them, P g is the sum of the available capacities of the generators in state x, L d is the system load at state x, L c =f LC is the load loss in state x when there is load loss, Ω success N k The set of states without load loss in the states, Ω failure N k The set of load loss states in the states; Step S52: Arrange the power generation capacity margin sequence from small to large to obtain Step S53: If the first Elements Then let γ k =0, otherwise let Step S54: Correct M(x) to M′(x)=M(x)-γ k , and get the sequence Step S55: Calculate parameters using formula (3) to formula (9). The quantities calculated by formula (3) to formula (9) are all parameters in V, where Indicates the nth state x drawn in this round of sampling n The rth component of nd For state x n The discrete components of Replace the indicator function H(x) of the reliability index in formula (4) to formula (9) with H M (x)W(x;P,Q k-1 ), and we can get V by substituting it into the calculation: For state x n The discrete components x nd : in, is the indicator function, when x nd The i-th component of takes the j-th value and is 1, otherwise it is 0. The expression is as follows: If the state x n The rth component of Follows the Weibull distribution: Where p is a fixed parameter; If the state x n The rth component of Follows a lognormal distribution: If the state x n The rth component of Obeying Rayleigh distribution: Step S56: Calculate parameter Q k : Qk=αV+(1-α)Qk-1(10).

8. A method for evaluating the reliability of a power system based on a no-load loss state set according to claim 7, characterized in that: Formulas (4) to (9) are obtained by minimizing H(g opt (x), g(x; Q)) is solved to obtain H(g opt (x), g(x; Q)) is the cross entropy, g(x; Q) is the unknown function to be solved, g opt (x) is obtained from the following formula: Among them, H(x) is the indicator function of a certain reliability index, l is the value of the reliability index of the system, which is a constant; The following optimization problem needs to be solved Where Q refers to any parameter vector and Θ is the parameter space; the constraints here are: Using the Lagrange multiplier method, we can solve equations (4) to (9), where n d represents the number of discrete components of the system state x, represents the probability of the i-th discrete component in the j-th case, n i Represents the total number of cases corresponding to the i-th discrete component.

9. The method for evaluating the reliability of a power system based on a no-load loss state set according to claim 1, characterized in that: The calculation formulas of LOLP and EPNS in step S10 are: (1) Load loss probability LOLP, which indicates the probability of load loss in the system: (2) Expected power shortage EPNS, which indicates the expected value of power shortage in the system: Among them, N sum The number of states that need to be considered for calculating the reliability index, x v The vth state of the reliability index is calculated this time; W(x v ) is the likelihood ratio W(x; P, Q) calculated in step S7 opt );ΔP is the power shortage corresponding to state x, Ω success N sum The set of states without load loss in the states, Ω failure N sum The set of load loss states in the state.

10. A method for evaluating the reliability of a power system based on a no-load loss state set according to claim 9, characterized in that: The calculation formula of the variance coefficient β in step S11 is: in, and n 2 Determined by the following formula: in, It indicates the average reliability index when the sampling reaches the qth sample in this round. There are N samples in this round. m times; N sum The number of states that need to be considered for calculating the reliability index, x v This is the vth state of the reliability index calculated this time; H(x) is the indicator function of the reliability index, using H LOLP (x) or H EPNS (x).

Images