A constant transmission rate and bandwidth allocation planning strategy to guarantee observability of the power grid

Abstract

This invention discloses a planning strategy for constant data transmission rate and bandwidth allocation in an ON-OFF transmission model, under the premise of ensuring smart grid observability while simultaneously satisfying relevant quality of service (QoS) indicators. Observability is a prerequisite for the normal operation of a smart grid, and the smart grid also has high real-time requirements for measurement data generated by PMUs. This invention is based on reinforcement learning algorithms, regression analysis methods, and the bisection method, and introduces effective capacity theory to measure the statistical QoS requirements corresponding to wireless fading channels, addressing strict latency constraints. The problem model is planned as follows: under the constraint of limited total bandwidth, by calculating the optimal constant transmission rate and allocating total bandwidth to the transmitters corresponding to each PMU, a bandwidth allocation scheme with the smart grid observability index as the objective function is determined.

Description

Technical Field

[0001] This invention relates to the fields of mobile communication system technology and power grid control system, and in particular to a planning strategy for constant transmission rate and bandwidth allocation to ensure power grid observability. Background Technology

[0002] In the future, the ubiquitous power Internet of Things (IoT) will connect all electricity-related physical objects and devices, forming a dual interconnection of power and information. 5G communication technology is rapidly developing, demonstrating excellent performance in bandwidth, large-scale IoT, and latency, and can be used to solve the "last mile" information interconnection problem of the ubiquitous power IoT. As the scale of the ubiquitous power IoT increases, the number of users and devices will also grow, leading to increased communication demands. However, there is still limited research on how to optimize topology and communication parameters to improve the performance indicators of the power system, and whether the optimized results can meet the ever-increasing information carrying capacity requires further investigation. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a planning strategy for constant transmission rate and bandwidth allocation under the ON-OFF transmission model, which can use the least amount of bandwidth while achieving the same observability index.

[0004] To address the aforementioned technical problems, this invention combines power grid observability, effective capacity theory, the bisection method, regression analysis, and reinforcement learning algorithms to provide the following technical solution:

[0005] Planning strategies for constant transmit rates and bandwidth allocation to ensure grid observability include the following steps:

[0006] (1) Derive the closed-form expression for effective capacity and the probability of data arriving on time;

[0007] (2) Calculate the optimal constant transmission rate of the transmitter;

[0008] (3) Define the objective function and analyze the variables to be optimized in the objective function;

[0009] (4) Generate an initial bandwidth allocation scheme;

[0010] (5) Set the relevant parameters of the reinforcement learning algorithm, including the learning rate α, discount factor γ, greedy probability ∈0 and ∈1, and positivity factor β;

[0011] (6) Initialize the Q-values ​​and reward function values ​​in the Q-table;

[0012] (7) Calculate the value corresponding to the current state s based on the initial bandwidth allocation scheme;

[0013] (8) Select the appropriate action based on the action selection strategy;

[0014] (9) Take the action selected in step (8) and calculate the next state s after taking the action. ′ The numerical values ​​and the corresponding reward function values;

[0015] (10) Update the Q value in the Q table;

[0016] (11) Set the next state s ′ The value is defined as the value corresponding to the current state s;

[0017] (12) Repeat steps (8) to (11) until the set number of iterations is reached;

[0018] (13) Output the optimal bandwidth allocation scheme.

[0019] Preferably, step (1) includes the following specific steps:

[0020] (11) Analyze the transmission model, PMU k Data transmission on the k-th channel between the base station and the ground station is represented as follows:

[0021] y k =h k x k +n k k = 1, 2, ... K

[0022] Where, x k For PMU k The emitted symbol, y k h represents the symbols received by the base station. k For PMU k The channel coefficient between the base station and n k Let N represent Gaussian noise with power spectral density N0, and K represent the number of channels, h k It follows Rayleigh block fading, and the channel coefficients between different blocks are independent. PMU k The channel gain between the base station and the base station is defined as

[0023] g k =|h k | 2 k = 1, 2, ... K

[0024] Among them, g k Follows the mean λ k The exponential distribution, λ k with PMU k This is related to the distance between the base station and the PMU. Since the PMU is unaware of the channel coefficient values, each PMU transmits data at a constant power P, therefore the real-time signal-to-noise ratio can be expressed as...

[0025]

[0026] Among them G k It follows an exponential distribution with a mean of one, ρ k This indicates the magnitude of the average signal-to-noise ratio:

[0027]

[0028] Based on Shannon capacity, PMU k The real-time channel capacity between the base station and the ground station is as follows:

[0029] C k =log2[1+ρ k G k bits / s / Hz

[0030] (12) Let PMU k The corresponding transmitter operates at a constant rate r k Data is transmitted in bits / s / Hz, and scenarios where the channel is in ON and OFF states are defined accordingly: when r k <C k The time channel is considered to be in the ON state, and is expressed as r k Data is transmitted at a rate of [r]. k >C k The channel is considered to be in the OFF state, and the effective data rate is 0.

[0031] (13) Derive the probability p that the channel is in the ON state. on ,as follows:

[0032]

[0033]

[0034] The probability that the channel is in the OFF state is p. OFF =1-p ON ;

[0035] (14) Define discrete random variable R k Its unit is bits / s / Hz, representing the random data transmission rate, and its probability distribution is: with p on The probability, R k =r k ; with p OFF The probability, R k =0;

[0036] (15) Based on the definition of effective capacity and the properties of the ON-OFF transmission model, the expression for effective capacity is as follows:

[0037]

[0038] According to R in step (14) k The probability distribution of the effective capacity can be expressed as follows:

[0039]

[0040] Therefore, the closed-form expression for the effective capacity is as follows:

[0041]

[0042] in,

[0043]

[0044] (16) Based on the effective capacity theory and queuing theory, the probability that the measurement data generated by the PMU arrives at the control center on time is calculated as follows:

[0045]

[0046] Preferably, step (2) includes the following specific steps:

[0047] (21) Analyze the constant transmission rate r of the transmitter. k Impact on data transmission rate and channel state: When r k When the value of r increases, the data transmission rate increases when the channel is in the ON state, but the probability of the channel being in the ON state decreases; conversely, when r... k When the value of r decreases, the data transmission rate when the channel is in the ON state decreases, but the probability of the channel being in the ON state increases; therefore, there exists an optimal r. k This results in a higher data transmission rate in the ON state and a higher probability that the channel is in the ON state.

[0048] (22) The magnitude of the smart grid observability index and p k Since k = 1, 2, ..., K is positively correlated, maximizing the observability index is equivalent to p. k Maximizing k = 1, 2, ..., K; furthermore, for all PMUs, as the expression f(θ) k B k ,r k As the value of ) increases, p k The value of θ will also increase accordingly; therefore, the optimization problem can be initially simplified to the expression f(θ). k B k ,r k Maximize )

[0049] (23) For a given θ k B k Calculate the optimal constant emission rate: for the expression f(θ) k B k ,r k )Calculate about r k The partial derivatives;

[0050] (24) Set the partial derivatives to zero, and then calculate the optimal r using the bisection method. k ;

[0051] (25) Repeat steps (23) to (24) to calculate several sets of θ. k B k and its corresponding optimal r k As training samples;

[0052] (26) Based on the least squares method, and combined with the training samples calculated in step (25), the optimal r is determined through multinomial regression analysis. k Fit to θ k B k The function.

[0053] Preferably, step (3) includes the following specific steps:

[0054] (31) The objective function is defined as the observability redundancy OR, observability sensitivity OS, or observability probability OP.

[0055] (32) Analyze the variables to be optimized in the objective function: the optimal r k It can be expressed as θ k B k The function; then the optimal θ k According to Perform calculations, where This represents the rate at which the PMU generates measurement data; finally, the objective function can be simplified to B. k The objective function is a function of k = 1, 2, ..., K, therefore the variable to be optimized in the objective function is the bandwidth corresponding to each PMU.

[0056] Preferably, step (7) includes the following specific steps:

[0057] (71) The value of the objective function is defined as the numerical value of the state s;

[0058] (72) The power grid has a total of K PMUs, denoted as PMU. k k = 1, 2, ..., K, therefore we have Where x i Let the vector X = {x1, x2, ..., x} N} TIf any PMU is installed on bus i, then x i =1, otherwise x i =0;

[0059] (73) with 0≤p k ≤1 indicates PMU k In D max The probability of completing a communication transmission within a given time, i.e.

[0060] Pr{d k ≤D max}=p k

[0061] Where, d k For PMU in actual situations k Communication latency, D max The maximum allowable delay;

[0062] (74) Define the diagonal probability matrix Λ P for:

[0063] Λ P =diag{P1,P2,…,P N}

[0064] Among them, if PMU k If installed on bus i, then P i =p k Otherwise P i =0; This yields the expected power grid observability vector. Represented as:

[0065]

[0066] Observability vector for:

[0067]

[0068] Where H represents the connection matrix of the power grid, which is determined by the network topology of the power grid, and the elements of this matrix H m,n Defined as: if bus m is connected to bus n, then H m,n =1, otherwise H m,n =0, m,n=1,2,…,N;Λ Q The diagonal communication constraint matrix is ​​defined as follows:

[0069] Λ Q =diag{Q1,Q2,…,Q N}

[0070] Q i It is a binary random variable, and its probability distribution is as follows:

[0071] Pr{Q i =1}=P i

[0072] Pr{Q i =0}=1-P i

[0073] (75) Based on the effective capacity theory and queuing theory, we obtain:

[0074]

[0075] (76) Assume PMU k by The constant rate generates measurement data, and the optimal θ k The corresponding effective capacity should be equal to the rate at which the measurement data is generated. Right now

[0076]

[0077] (77) Combining the initial bandwidth allocation scheme, and based on the equation in step (76), solve θ using the bisection method. k ;

[0078] (78) Based on θ obtained in step (77) k Calculate p k ;

[0079] (79) Repeat steps (77) to (78) to find θ and p for all PMUs;

[0080] (710) Calculate the value of the objective function based on the result of step (79), which is the value corresponding to the current state s.

[0081] Preferably, step (8) includes the following specific steps:

[0082] (81) Since the variable to be optimized is the bandwidth of each PMU, the content of the action is defined as the increase or decrease of the bandwidth of each PMU.

[0083] (82) Determine whether the current value of the reward function is greater than 0;

[0084] (83) Select an action based on the judgment result of (82): If the judgment result is yes, select the action from the previous round with probability β, select an action using a greedy algorithm with probability ∈1, and select an action randomly with probability 1-∈1-β; if the judgment result is no, select an action using a greedy algorithm with probability ∈0, and select an action randomly with probability 1-∈0.

[0085] The beneficial effects of this invention are as follows: Firstly, this invention combines the channel uncertainty of wireless communication with the observability of the power grid using effective capacity theory and queuing theory. It then proposes an algorithm that can simultaneously plan the constant transmission rate of the transmitter and allocate bandwidth by employing a bisection method, regression analysis, and reinforcement learning algorithms. This algorithm can ensure that the total bandwidth of the system is minimized while achieving the same observability index. By designing the action selection strategy of the reinforcement learning algorithm, the convergence speed of the objective function is significantly improved. Furthermore, considering the characteristics of the ON-OFF transmission model, by designing a constant transmission rate for the transmitter, both the probability of the channel being in the ON state and the data transmission rate when the channel is in the ON state can be taken into account, effectively meeting the needs of power grid planning. Detailed Implementation

[0086] The invention will be verified using the IEEE 14 bus power system, which is widely used as a standard test case. The channel is configured to experience independent block fading, with each PMU's transmitter transmitting data to the receiver via OFDMA. Rayleigh fading is selected as the fading characteristic of the wireless channel.

[0087] Assuming all PMUs generate measurement data at a rate of 60 kbps, the coherence time T is 0.005 s, and the maximum allowable delay for data transmission from the PMU to the control center is 10 ms, the PMU installation vector is expressed as:

[0088] X={0,1,0,1,1,1,1,1,1,0,1,0,1,0} T

[0089] The average signal-to-noise ratio values ​​for each PMU are set as follows:

[0090] PMU serial number 1 2 3 4 5 6 7 8 9 Average signal-to-noise ratio 17 11 19 15 26 30 14 28 29

[0091] The planning process is as follows, including the following steps:

[0092] (1) Cross-layer statistical delay analysis is performed using effective capacity theory. Applications related to maintaining power grid observability have high requirements for communication delay. Effective capacity theory can effectively combine communication delay and power grid observability indicators. Through effective capacity theory and queuing theory, the probability of data packets generated by PMU being transmitted to the control center within the maximum allowable delay time is calculated.

[0093] (2) Consider the optimal constant transmit rate of the transmitter in the ON-OFF transmission model. When r k When the value of r increases, the data transmission rate increases when the channel is in the ON state, but the probability of the channel being in the ON state decreases; conversely, when r... kWhen the value of r decreases, the data transmission rate when the channel is in the ON state decreases, but the probability of the channel being in the ON state increases; therefore, there exists an optimal r. k This results in a higher data transmission rate in the ON state and a higher probability that the channel is in the ON state.

[0094] (3) A bandwidth allocation scheme is generated based on reinforcement learning and the binary search method. During the optimization of the objective function, the value of the power grid observability index under the current scheme is calculated in step (1), and the optimal constant transmission rate is calculated using the result of step (2). Through reinforcement learning algorithm, the objective function is iterated repeatedly until the value of the objective function converges to the optimal value, and the optimal constant transmission rate and bandwidth allocation scheme are obtained.

[0095] In step (1), the cross-layer statistical delay analysis using effective capacity theory specifically includes the following steps:

[0096] (11) The K PMUs in the power grid are respectively denoted as PMU. k k = 1, 2, ..., K, satisfying Where x i Install vector X = {x1, x2, ..., x} for PMU N} T If any PMU is installed on bus i, then x i =1, otherwise x i =0. In a fading channel environment, PMU k With probability 0≤p k ≤1 satisfies D max Communication latency requirements.

[0097] (12) Define the diagonal probability matrix Λ P for:

[0098] Λ P =diag{P1,P2,…,P N}

[0099] Among them, if PMU k If installed on bus i, then P i =p k Otherwise P i =0; This yields the expected power grid observability vector. Represented as:

[0100]

[0101] Power grid observability vector for:

[0102]

[0103] Where H represents the connection matrix of the power grid, and the elements H in this matrix m,n Defined as follows: If bus m is connected to bus n, then H m,n =1, otherwise H m,n =0, m,n=1,2,…,N;Λ Q The diagonal communication constraint matrix is ​​defined as follows:

[0104] Λ Q =diag{Q1,Q2,…,Q N}

[0105] The elements on the diagonal of this diagonal communication constraint matrix are all binary random variables, and their probability distributions are as follows:

[0106] Pr{Q i =1}=P i

[0107] Pr{Q i =0}=1-P i

[0108] (13) Analyze the transmission model, PMU k Data transmission on the k-th channel between the base station and the ground station is represented as follows:

[0109] y k =h k x k +n k k = 1, 2, ... K

[0110] Where, x k For PMU k The emitted symbol, y k h represents the symbols received by the base station. k For PMU k The channel coefficient between the base station and n k This represents Gaussian noise with a power spectral density of N0. k Obeying Rayleigh block decay, PMU k The channel gain between the base station and the base station is defined as

[0111] g k =|h k | 2 k = 1, 2, ... K

[0112] Among them, g k Follows the mean λ k The exponential distribution, λ k with PMU kThis is related to the distance between the base station and the PMU. Since the PMU is unaware of the channel coefficient values, each PMU transmits data at a constant power P, therefore the real-time signal-to-noise ratio can be expressed as...

[0113]

[0114] Among them G k It follows an exponential distribution with a mean of one, ρ k This indicates the magnitude of the average signal-to-noise ratio:

[0115]

[0116] Based on Shannon capacity, PMU k The real-time channel capacity between the base station and the ground station is as follows:

[0117] C k =log2[1+ρ k G k bits / s / Hz

[0118] Let PMU k The corresponding transmitter operates at a constant rate r k Data is transmitted in bits / s / Hz, and scenarios where the channel is in ON and OFF states are defined accordingly: when r k <C k The time channel is considered to be in the ON state, and is expressed as r k Data is transmitted at a rate of [r]. k >C k The channel is considered to be in the OFF state, and the effective data rate is 0. Based on the above analysis, the probability p of the channel being in the ON state is derived. on ,as follows:

[0119]

[0120] The probability that the channel is in the OFF state is p. oFF =1-p ON .

[0121] (14) Define discrete random variable R k Its unit is bits / s / Hz, representing the random data transmission rate, and its probability distribution is: with p on The probability, R k =r k ; with p OFF The probability, R k =0; the expression for effective capacity is as follows:

[0122]

[0123] According to R k The probability distribution of the effective capacity can be expressed as follows:

[0124]

[0125] Therefore, the closed-form expression for the effective capacity is as follows:

[0126]

[0127] in,

[0128]

[0129] (15) Based on the theory of effective capacity and queuing theory, it can be calculated that:

[0130]

[0131] (16) Configure PMU k by The measurement data is generated at a constant rate; for reliable transmission, the effective capacity must not be less than [a certain value]. Since the effective capacity is about θ k The decreasing function of p, and k It's about θ k It is an increasing function, so given the bandwidth, the optimal θ is... k The corresponding effective capacity should be equal to the rate at which the measurement data is generated. Right now

[0132]

[0133] The optimal θ can be solved using the bisection method. k Based on the above analysis, we can calculate θ and p for all PMUs.

[0134] (17) Calculate the numerical value of the observability index. Observability redundancy (OR) provides an overall assessment of grid observability, defined as follows:

[0135]

[0136] Observability sensitivity OS is defined as the expected power grid observability vector. The smallest element in the set. The observability probability OP is defined as:

[0137]

[0138] Where t can be set to Indicates unit observability.

[0139] Step (2) involves calculating the optimal constant transmission rate of the transmitter, which includes the following steps:

[0140] (21) By analyzing the constant transmission rate r of the transmitter k Regarding the impact of data transmission rate and channel state, the conclusion is that an optimal r exists. k This results in a higher data transmission rate in the ON state and a higher probability that the channel is in the ON state.

[0141] (22) Analyze the characteristics of the system model: the numerical magnitude of the smart grid observability index and p k Since k = 1, 2, ..., K is positively correlated, maximizing the observability index is equivalent to p. k Maximizing k = 1, 2, ..., K is equivalent to solving K independent subproblems; furthermore, for all PMUs, as the expression f(θ) k B k ,r k As the value of ) increases, p k The value of θ will also increase accordingly; therefore, the optimization of the objective function can be initially simplified to the expression f(θ). k B k ,r k Maximize ).

[0142] (23) For a given θ k B k Calculate the optimal constant emission rate: for the expression f(θ) k B k ,r k )Calculate about r k The partial derivatives of .

[0143] (24) Set the partial derivatives to zero, and then calculate the optimal r using the bisection method. k .

[0144] (25) Repeat steps (23) to (24) to calculate several sets of θ. k B k and its corresponding optimal r k As training samples.

[0145] (26) Based on the least squares method, and combined with the training samples calculated in step (25), the optimal r is determined through multinomial regression analysis. k Fit to θ k B k The function.

[0146] Step (3) involves generating a bandwidth allocation scheme based on reinforcement learning algorithms and binary search, including the following steps:

[0147] (31) Define the objective function as the value of the observability index, and maximize the value of the power grid observability index through optimization algorithm.

[0148] (32) The optimal constant transmission rate of the transmitter under the ON-OFF transmission model is calculated by the bisection method, and the expression of the optimal constant transmission rate is obtained by the polynomial regression analysis method.

[0149] (33) An initial solution for bandwidth allocation is generated using an average allocation method. Then, the optimal θ for all PMUs is calculated by solving the equations using the bisection method. k For k = 1, 2, ..., K, the equation is as follows:

[0150]

[0151] Based on this, the numerical value of the objective function corresponding to the initial solution is calculated.

[0152] (34) Based on the characteristics of the model and the reinforcement learning algorithm, set the learning rate α, discount factor γ, greedy probability ∈0 and ∈1, and positivity factor β; and define the value corresponding to state s in the reinforcement learning algorithm as the value of the objective function;

[0153] (35) Initialize the Q-values ​​and reward function values ​​in the Q-table;

[0154] (36) Design action selection strategy: Since the variable to be optimized is the bandwidth of each PMU, the action is defined as the increase or decrease of the bandwidth of each PMU. First, it is determined whether the value of the current reward function is greater than 0, and then the action is selected based on the determination result: if the determination result is yes, the action of the previous round is selected with a probability of β, the action is selected by a greedy algorithm with a probability of ∈1, and the action is randomly selected with a probability of 1-∈1-β; if the determination result is no, the action is selected by a greedy algorithm with a probability of ∈0, and the action is randomly selected with a probability of 1-∈0.

[0155] (37) Design the value of the reward function: by comparing the value of state s with the next state s after taking the action. ′ The value of the objective function is used to calculate the value of the reward function. The value of the reward function and the change in the value of the objective function have a linear relationship. After taking an action, if the objective function increases, the reward function is positive; if the objective function decreases, the reward function is negative.

[0156] (38) Select the appropriate action according to the action selection strategy and take the action. At this time, the bandwidth allocation scheme will also change accordingly. Calculate the next state s based on the bandwidth allocation scheme after taking the action. ′ The value; based on the value of state s and the next state s ′ The value of the reward function r is calculated.

[0157] (39) Update the Q value in the Q table. The expression for updating this value is as follows:

[0158]

[0159] (310) Set the next state s ′ The value is defined as the value corresponding to the current state s;

[0160] (311) Repeat steps (38) to (310) until the set number of iterations is reached;

[0161] (312) Output the optimal bandwidth allocation scheme.

[0162] Although the present invention has been illustrated and described with reference to preferred embodiments, those skilled in the art should understand that various changes and modifications can be made to the present invention without departing from the scope defined by the claims.

Claims

1. A method of planning a constant transmission rate and bandwidth allocation that guarantees observability of an electrical grid, characterized in that, Includes the following steps: (1) Derive the closed-form expression for effective capacity and the probability of data arriving on time; (2) Calculate the optimal constant transmission rate of the transmitter; (3) Define the objective function and analyze the variables to be optimized in the objective function; (4) Generate an initial bandwidth allocation scheme; (5) Set the relevant parameters of the reinforcement learning algorithm, including the learning rate. Discount factor Greedy probability and Positive factors ; (6) Initialize the Q-values ​​and reward function values ​​in the Q-table; (7) Calculate the current state based on the initial bandwidth allocation scheme. The corresponding value; (8) Select the appropriate action based on the action selection strategy; (9) Take the action selected in step (8) and calculate the next state after taking the action. The numerical values ​​and the corresponding reward function values; (10) Update the Q value in the Q table; (11) Change the next state The value is defined as the current state. The corresponding value; (12) Repeat steps (8) to (11) until the set number of iterations is reached; (13) Output the optimal bandwidth allocation scheme; The specific steps of step (1) include: (11) Analyze the transmission model. The first between and base stations Data transmission across multiple channels is represented as follows: ； in, for The emitted symbols The symbols received by the base station. for Channel coefficients between the base station and the base station Indicates the power spectral density as Gaussian noise, Indicates the number of channels. It follows Rayleigh block fading, and the channel coefficients between different blocks are independent. The channel gain between the base station and the base station is defined as ； in, Follow the mean The exponential distribution, and The distance between the PMU and the base station is correlated; each PMU operates at a constant power. The real-time signal-to-noise ratio for transmitting data is expressed as follows: ； in It follows an exponential distribution with a mean of one. This indicates the magnitude of the average signal-to-noise ratio: ； According to Shannon capacity, The real-time channel capacity between the base station and the ground station is as follows: ； (12) Let The corresponding transmitter operates at a constant rate. Transmit data, and based on this, define the scenarios where the channel is in ON and OFF states respectively: When The time channel is considered to be in the ON state, and with Data transmission is performed at a rate that allows for data transfer; when The channel is considered to be in the OFF state, and the effective data rate is 0. (13) Derive the probability that the channel is in the ON state. ,as follows: ； The probability that the channel is in the OFF state is ; (14) Define discrete random variables , representing the random data transmission rate, has the following probability distribution: The probability, ;by The probability, ; (15) Based on the definition of effective capacity and the properties of the ON-OFF transmission model, the expression for effective capacity is as follows: ； According to step (14) The probability distribution, and the expected value in the effective capacity expression are as follows: ； Therefore, the closed-form expression for the effective capacity is as follows: ； in, ； (16) Based on the effective capacity theory and queuing theory, the probability that the measurement data generated by the PMU arrives at the control center on time is calculated as follows: ； The specific steps of step (2) include: (21) The optimization problem of constant emission rate is initially simplified to the expression Maximize; (22) For a given Calculate the optimal constant emission rate: for the expression Calculate about The partial derivatives; (23) Set the partial derivatives to zero, and then calculate the optimal value using the bisection method. ; (24) Repeat steps (22) to (23) to calculate several sets. and its corresponding optimal As training samples; (25) Based on the least squares method, combined with the training samples calculated in step (24), the optimal sample is determined by multinomial regression analysis. Fit to The function; The specific steps of step (3) include: (31) The objective function is defined as the observability redundancy OR, the observability sensitivity OS, or the observability probability OP. (32) Analyze the variables to be optimized in the objective function: the optimal Expressed as The function; then the optimal one. according to Perform calculations, where This represents the rate at which the PMU generates measurement data; finally, the objective function simplifies to... The objective function is a function of PMU, therefore the variable to be optimized in the objective function is the bandwidth corresponding to each PMU; The specific steps of step (7) include: (71) The numerical value of the objective function is defined as the state. The numerical value; (72) The power grid has a total of Each PMU is denoted as... Therefore, there is ,in For vectors The elements in the bus, if a PMU is installed ,but ,otherwise ; (73) with express exist The probability of completing a communication transmission within a given time, i.e. ； in, In actual circumstances Communication latency, The maximum allowable delay; (74) Define the diagonal probability matrix for: ； Among them, if Installed on the bus ,but ,otherwise This yields the expected power grid observability vector. , is represented as: ； Observability vector for: ； in The connection matrix of the power grid is determined by the network topology of the power grid, and the elements of this matrix are... Defined as: if the bus With bus If connected, then ,otherwise , ; The diagonal communication constraint matrix is ​​defined as follows: ； It is a binary random variable, and its probability distribution is as follows: ； ； (75) Based on the effective capacity theory and queuing theory, we obtain: ； (76) Assumption by The constant rate generates measurement data, optimally The corresponding effective capacity should be equal to the rate at which the measurement data is generated. ,Right now ； (77) Based on the initial bandwidth allocation scheme, and according to the equation in step (76), solve using the bisection method. ; (78) Based on the result obtained in step (77) calculate ; (79) Repeat steps (77) to (78) to find all PMUs. and ; (710) Calculate the value of the objective function based on the result of step (79), which is the current state. The corresponding numerical value.

2. The planning method for constant transmission rate and bandwidth allocation to ensure power grid observability according to claim 1, characterized in that, The specific steps of step (8) include: (81) Since the variable to be optimized is the bandwidth of each PMU, the content of the action is defined as the increase or decrease of the bandwidth of each PMU; (82) Determine whether the current value of the reward function is greater than 0; (83) Select an action based on the judgment result of (82): If the judgment result is yes, then take the action. The probability of choosing the action from the previous round, in order to The probability is used to select actions using a greedy algorithm, in order to The action is randomly selected based on probability; if the result is negative, then... The probability is used to select actions using a greedy algorithm, in order to The probability of randomly selecting an action.

Images