A Power System Look-Ahead Dispatch Method Based on Migration Cost Function Approximation

Abstract

This invention discloses a power system forward scheduling method based on migration-type cost function approximation, comprising the following steps: S1. Constructing a power system forward scheduling model based on a cost function, with the minimization of power system scheduling cost as the objective function, and determining the constraints of each parameter of the state variables and decision variables in the cost function; S2. Based on the cost function approximation theory and considering the influence of random information, transforming the objective function, and obtaining an approximate cost function by introducing approximate parameters to estimate the influence of random information on the cost function; S3. Establishing a mapping relationship between the approximate parameters and random information, using a Long Short-Term Memory (LSTM) neural network for fitting, identifying the parameters of the LSTM, and obtaining the optimal parameters of the approximate parameters; S4. Inputting the optimal parameters of the approximate parameters into the power system forward scheduling model, the power system forward scheduling model outputs a real-time scheduling optimization scheme, and scheduling the power system according to the real-time scheduling optimization scheme. This improves the accuracy of forward scheduling.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and more specifically, to a power system forward dispatching method based on migration-type cost function approximation. Background Technology

[0002] With the implementation of the "dual-carbon" strategy and the construction of new power systems, renewable energy has developed rapidly. However, the strong volatility, randomness, and intermittency of renewable energy pose significant challenges to the optimal scheduling of power systems.

[0003] Existing economic dispatch methods for power systems can be divided into static dispatch methods and dynamic dispatch methods. Short-sighted strategies, a commonly used static dispatch method, optimize only the operating cost at the current moment without considering the operating cost of the entire operating period. Therefore, it has a fast solution speed, but suffers from low solution quality. Look-ahead dispatch, a dynamic dispatch method, has been proven to solve the above problems. Model predictive control (MMC) is the most commonly used look-ahead dispatch method. It uses predictive information from several future time periods to solve the optimization problem, executing only current decisions. Because look-ahead dispatch considers the impact of current decisions on the future, it can obtain better solutions than static dispatch methods. However, MMC can only utilize information from several future time periods and cannot fully consider the random information of the entire dispatch period, which to some extent affects its dispatch decision accuracy.

[0004] The cost function approximation method is a powerful stochastic optimization approach that has been widely applied in areas such as flight scheduling and supply chain management. It can utilize information from several future time periods for decision-making and introduce parameters into the objective function or constraints to estimate the impact of future stochasticity. This effectively addresses the problem that model predictive control methods cannot fully consider stochastic information throughout the entire scheduling period, and in principle, it is naturally compatible with forward-looking scheduling in power systems. However, the application of the cost function approximation method in forward-looking scheduling of power systems has not been reported. Furthermore, the computation of the cost function approximation method requires numerous iterative updates, which may increase the computational burden. Therefore, in most cases, it cannot meet the requirements of real-time decision-making, and how to solve this problem still requires further research. Summary of the Invention

[0005] The purpose of this invention is to solve the problem of how to quickly and accurately calculate the operating costs of multiple time periods in the economic dispatch of power systems. It provides a forward-looking dispatch method for power systems based on migration-type cost function approximation, and applies the cost function approximation method to power system dispatch, providing a basis for the accuracy of real-time economic dispatch schemes for power systems.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A forward-looking scheduling method for power systems based on migration-type cost function approximation includes the following steps:

[0008] S1. Construct a forward-looking power system dispatch model based on a cost function, with the goal of minimizing the dispatch cost of the power system, and determine the constraints on the parameters of the state variables and decision variables in the cost function;

[0009] S2. Based on the cost function approximation theory and considering the influence of random information, the objective function is transformed, and the influence of random information on the cost function is estimated by introducing approximate parameters to obtain an approximate cost function;

[0010] S3. Establish the mapping relationship between approximate parameters and random information, use a Long Short-Term Memory Neural Network (LSTM) for fitting, identify the parameters of the LSTM, and obtain the optimal parameters of the approximate parameters;

[0011] S4. Input the optimal parameters of the approximate parameters into the forward-looking scheduling model of the power system. The forward-looking scheduling model of the power system outputs a real-time scheduling optimization scheme. The power system is then scheduled according to the real-time scheduling optimization scheme.

[0012] Furthermore, the objective function is:

[0013]

[0014] In the formula, C t (S t ,x t Let x be the cost function for time period t. t Let S be the decision variable for time period t. t Let C be the state variable for time period t. τ (S τ ,x τ Let x be the cost function for time period τ. τ Let S be the decision variable for time period τ. τ Let Ω be the state variable for time interval τ, where Ω = {t + Δt, t + 2Δt, ..., t + HΔt};

[0015] Wherein, the cost function C for time period t t (S t ,x t The formula for calculating ) is:

[0016]

[0017] In the formula, C g，t For the cost of generator g in time period t, C ESS,t For the charging and discharging cost of energy storage during time period t, Cpun,t Penalty costs for reducing renewable energy and load during period t;

[0018] State variables in time period t E represents the generator power during time period t-1. t P represents the remaining energy stored during time period t. t NG P represents the predicted value of renewable energy in time period t. t load This represents the predicted load value for time period t;

[0019] The decision variable for time period t is x t =(P t g ;P t bat ;P t NG,cur ;P t load,cur ), P t g Let P be the generator power during time period t. t bat P represents the energy storage charging and discharging power during time period t. t NG,cur Let P be the amount of renewable energy reduction during period t. t load,cur The load reduction amount for period t;

[0020] C g，t With P t g Regarding, C ESS,t With P t bat Regarding, C pun,t With P t NG,cur and P t load,cur related.

[0021] Furthermore, to ensure the safety of the power system, the parameters of the state variables and decision variables must satisfy the following constraints:

[0022]

[0023]

[0024]

[0025]

[0026]

[0027] In the formula, Let be the generator power of node i during time period t; Let be the predicted value of renewable energy for node i in time period t; The energy storage charging power of node i during time period t; Let be the energy storage discharge power of node i during time period t; Let be the predicted load of node i during time period t; Let be the amount of renewable energy reduction at node i during time period t; Let be the load reduction amount of node i during time period t; For a set of nodes; and These are sets of nodes that have generators and renewable energy sources, respectively; G ij and B ij e represents the real and imaginary parts of the node admittance matrix in the i-th row and j-th column; i,t and f i,t e represents the real and imaginary parts of the voltage phasor at node i during time period t; j,t and f j,t It represents the real and imaginary parts of the voltage phasor at node j during time period t; Let be the reactive load of node i during time period t; Let be the reactive power of the generator at node i during time period t.

[0028] Furthermore, based on the cost function approximation theory, the objective function of the power system look-ahead scheduling model is transformed into:

[0029]

[0030] In the formula, C t (S t ,x t C represents the cost function for the current time period t; τ (S τ ,x τ ,ω) represents the cost function for the subsequent τ time interval. Due to the presence of random information ω, C τ (S τ ,x τ ,ω) is unknown.

[0031] Furthermore, the influence of random information is estimated by introducing an approximate parameter β, which includes β... NG and β load After introducing the approximate parameter β, we get:

[0032]

[0033]

[0034]

[0035]

[0036] in, It is the approximate cost function after introducing approximate parameters, β NG β is an approximate parameter for the predicted value of renewable energy. load These are approximate parameters for the load forecast.

[0037] Furthermore, in order for the approximate parameter β to be accurately estimated... The influence of random information necessitates minimizing the difference between the approximate cost function and the actual cost function, resulting in:

[0038]

[0039] in, Let F be the gradient of F with respect to β; J be the Jacobian matrix; α be the iteration step size; after a certain number of iterations, the optimal parameter β of the cost function approximation optimization method can be obtained. * .

[0040] Furthermore, the optimal parameter β is solved using the stochastic gradient descent method. * The process of updating the approximate parameter β is as follows:

[0041]

[0042]

[0043] in, α is the gradient of F with respect to β; J is the Jacobian matrix; α is the iteration step size;

[0044] After iterative updates, the optimal parameter β of the cost function approximation optimization method can be obtained. * .

[0045] Furthermore, to improve the computational efficiency of iterative updates of the approximate parameter β, transfer learning is used to directly obtain the approximate parameter β based on the random information ω. A transfer function is then used to map the relationship between the approximate parameter β and the random information ω. The mapping relationship is as follows:

[0046] β=L(ω)

[0047] In the formula, L(ω) is the constructed transfer function.

[0048] Furthermore, a Long Short-Term Memory (LSTM) neural network was used for fitting, and... This indicates that the parameters of the LSTM model are identified, so that... By getting as close as possible to L(ω), we obtain:

[0049]

[0050] Where p represents the parameter vector of the LSTM model, p * Let G(ω) represent the optimal value of p. The error value of L(ω), This represents the transfer function fitted by the LSTM.

[0051] Furthermore, the optimal value of the parameter vector p of the LSTM is solved using the adaptive moment estimation method, and the update equation is:

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] In the formula, p c p represents the parameter vector for the c-th iteration; c-1 This represents the parameter vector for the (c-1)th iteration; This represents the gradient of the parameter vector p; G represents the error value; κ represents the learning rate; s c This represents the gradient of the exponential moving average in the c-th iteration. r represents the gradient of the modified exponential moving average in the c-th iteration; c R represents the biased quadratic primitive moment estimate of the c-th iteration; c-1 This represents the biased quadratic primitive moment estimate for the (c-1)th iteration; Let ρ1, ρ2, and ρ3 represent the corrected bias quadratic primitive moment estimate for the c-th iteration; It is a constant.

[0058] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0059] 1. This invention proposes for the first time a method based on transfer learning and cost function approximation to a look-ahead scheduling model. In addition to using prediction information from future time periods for decision-making, the method can also estimate the impact of randomness in other time periods through the cost function, thereby improving the accuracy of look-ahead scheduling.

[0060] 2. This invention proposes for the first time a power system dispatching method based on transfer learning cost function approximation. Unlike the cost function approximation method that iteratively updates the approximation parameters, the proposed method uses transfer learning technology to avoid iterative updates of the cost function approximation method, which can significantly reduce the computation time.

[0061] 3. This invention proposes for the first time a forward-looking scheduling method for power systems that combines transfer learning with cost function approximation. It develops forward-looking scheduling by utilizing cost function approximation methods, which can make full use of future forecast information in scheduling decisions. Furthermore, by introducing transfer learning technology, it can improve the solution efficiency.

[0062] 4. By improving the traditional forward-looking dispatch model and method for power systems, this invention can better handle the randomness of renewable energy, which is conducive to reducing the total operating cost of the power system while ensuring power balance, and promoting the optimal consumption of renewable energy, thereby bringing good social and economic benefits. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating the forward scheduling method for power systems based on migration-type cost function approximation according to the present invention. Detailed Implementation

[0064] The following description, in conjunction with the accompanying drawings and specific embodiments, further illustrates the power system forward scheduling method based on migration-type cost function approximation of the present invention.

[0065] Please see Figure 1 This invention discloses a power system forward scheduling method based on migration cost function approximation, comprising the following steps:

[0066] S1. Construct a forward-looking dispatch model for the power system based on a cost function, with the goal of minimizing the dispatch cost of the power system, and determine the constraints on the parameters of the state variables and decision variables in the cost function.

[0067] S2. Based on the cost function approximation theory, an approximate parameter is introduced to estimate the impact of random information on power system dispatch, and a mapping relationship between the approximate parameter and the random information is established.

[0068] S3. By using the cost function approximation strategy of transfer learning, the optimal parameters of the cost function approximation method are obtained. The optimal parameters are then input into the power system look-ahead scheduling model for optimization to obtain the optimal real-time scheduling scheme of the power system.

[0069] S4. Calculate the optimal real-time dispatch scheme for the power system based on the cost function approximation method, and dispatch the power system according to the approximate optimal strategy.

[0070] This invention applies the cost function approximation method to power system optimal scheduling, forming a novel look-ahead scheduling method. Furthermore, it utilizes transfer learning technology to improve the cost function approximation method, avoiding the iterative update process of the cost function approximation method and thus accelerating the decision-making speed.

[0071] The power system look-ahead scheduling problem can be described as a Markov decision process (MDP), with the objective function being:

[0072]

[0073] In the formula, C t (S t ,x t Let x be the cost function for time period t. t Let S be the decision variable for time period t. t Let C be the state variable for time period t. τ (S τ ,x τ Let x be the cost function for time period τ. τ Let S be the decision variable for time period τ. τ Let Ω be the state variable for time interval τ, where Ω = {t + Δt, t + 2Δt, ..., t + HΔt};

[0074] Wherein, the cost function C for time period t t (S t ,x t The formula for calculating ) is:

[0075]

[0076] In the formula, C g，t For the cost of generator g in time period t, C ESS,t For the charging and discharging cost of energy storage during time period t, C pun,t Penalty costs for reducing renewable energy and load during period t;

[0077] State variables in time period t E represents the generator power during time period t-1. t P represents the remaining energy stored during time period t. t NG P represents the predicted value of renewable energy in time period t. t load This represents the predicted load value for time period t;

[0078] The decision variable for time period t is x t =(P t g ;P t bat ;P t NG,cur ;P t load,cur ), P t g Let P be the generator power during time period t. t bat P represents the energy storage charging and discharging power during time period t.t NG,cur Let P be the amount of renewable energy reduction during period t. t load,cur The load reduction amount for period t;

[0079] C g，t With P t g Regarding, C ESS,t With P t bat Regarding, C pun,t With P t NG,cur and P t load,cur related.

[0080] To ensure the safety of the power system, the parameters of the state variables must satisfy the following constraints:

[0081]

[0082]

[0083]

[0084]

[0085] in, Let be the generator power of node i during time period t; Let be the predicted value of renewable energy for node i in time period t; The energy storage charging power of node i during time period t; Let be the energy storage discharge power of node i during time period t; Let be the predicted load of node i during time period t; Let be the amount of renewable energy reduction at node i during time period t; Let be the load reduction amount of node i during time period t; For a set of nodes; and These are sets of nodes that have generators and renewable energy sources, respectively; G ij and B ij e represents the real and imaginary parts of the node admittance matrix in the i-th row and j-th column; i,t and f i,t e represents the real and imaginary parts of the voltage phasor at node i during time period t; j,t and f j,t It represents the real and imaginary parts of the voltage phasor at node j during time period t; Let be the reactive load of node i during time period t; Let be the reactive power of the generator at node i during time period t.

[0086] In the aforementioned look-ahead scheduling model, the decision at each time step only considers the predicted information for the next few time periods, failing to adequately account for the randomness throughout the entire scheduling cycle. To address this issue, this invention proposes a look-ahead scheduling method that approximates the cost function. Based on the cost function approximation theory, the objective function can be transformed into:

[0087]

[0088] In the formula, C t (S t ,x t C represents the cost function for the current time period; τ (S τ ,x τ ,ω) represents the cost function for the subsequent τ time period. Since there is random information ω, this term is unknown.

[0089] By forecasting the value P of renewable energy and load at time t t NG P t load Before introducing approximate parameters β (including β), NG and β load This can be used to estimate the impact of random information in other time periods.

[0090] At this point, formulas (3), (5), (6), and (7) can be redefined as:

[0091]

[0092]

[0093]

[0094]

[0095]

[0096] in, It is the approximate cost function after introducing approximate parameters, β NG β is an approximate parameter for the predicted value of renewable energy. load These are approximate parameters for the load forecast.

[0097] In a stochastic environment, the accuracy of the solution to the above optimization problem depends on whether the approximate parameter β can be accurately estimated. The influence of random information. Therefore, we need to minimize the difference between the approximate cost function and the actual cost function, that is:

[0098]

[0099] Considering the characteristics of the above equation, the stochastic gradient descent method is used for solving it. The detailed description of updating β is as follows:

[0100]

[0101]

[0102] in, Let F be the gradient of F with respect to β; J be the Jacobian matrix; and α be the iteration step size. After a certain number of iterations, the optimal parameter β of the cost function approximation optimization method can be obtained. * .

[0103] The approximate parameter β in the above method (including β) NG and β load The need for iterative updates inevitably reduces computational efficiency. To address this issue, this invention utilizes transfer learning techniques to directly obtain β from random information ω. Specifically, a transfer function maps the relationship between parameter β and random information ω, with the mapping relationship being:

[0104] β=L(ω) (15)

[0105] In the formula, L(ω) is the constructed transfer function.

[0106] Considering the correlation of random information in adjacent time intervals, a Long Short-Term Memory (LSTM) neural network method, which is adept at handling such problems, was used to build a fitting model, and... To represent this. In order to identify the parameters of the LSTM model, so that... By getting as close as possible to L(ω), we obtain:

[0107]

[0108] Where p represents the parameter vector of the LSTM, p * Let G(ω) represent the optimal value of p. The error value of L(ω), This represents the transfer function fitted by the LSTM.

[0109] Solving the above equation using the adaptive moment estimation method, the update equation can be described as:

[0110]

[0111]

[0112]

[0113]

[0114]

[0115] In the formula, p c p represents the parameter vector for the c-th iteration; c-1 This represents the parameter vector for the (c-1)th iteration; This represents the gradient of the parameter vector p; G represents the error value; κ represents the learning rate; s c This represents the gradient of the exponential moving average in the c-th iteration. r represents the gradient of the modified exponential moving average in the c-th iteration; c R represents the biased quadratic primitive moment estimate of the c-th iteration; c-1 This represents the biased quadratic primitive moment estimate for the (c-1)th iteration; Let ρ1, ρ2, and ρ3 represent the corrected bias quadratic primitive moment estimate for the c-th iteration; It is a constant.

[0116] In summary, the present invention has the following advantages and beneficial effects:

[0117] 1. This invention proposes for the first time a method based on transfer learning and cost function approximation to a look-ahead scheduling model. In addition to using prediction information from future time periods for decision-making, the method can also estimate the impact of randomness in other time periods through the cost function, thereby improving the accuracy of look-ahead scheduling.

[0118] 2. This invention proposes for the first time a power system dispatching method based on transfer learning cost function approximation. Unlike the cost function approximation method that iteratively updates the approximation parameters, the proposed method uses transfer learning technology to avoid iterative updates of the cost function approximation method, which can significantly reduce the computation time.

[0119] 3. Theoretical Innovation. This invention proposes for the first time a forward-looking scheduling method for power systems that combines transfer learning with cost function approximation. It develops forward-looking scheduling by utilizing cost function approximation methods, which can fully utilize future forecast information in scheduling decisions. Furthermore, the introduction of transfer learning technology can improve solution efficiency.

[0120] 4. Socioeconomic benefits. This invention, through improvements to the traditional forward-looking dispatch model and method for power systems, can better handle the stochasticity of renewable energy, which is conducive to reducing the total operating cost of the power system while ensuring power balance, and promotes the optimal consumption of renewable energy, thereby bringing good socioeconomic benefits.

[0121] The above description is a detailed description of the preferred embodiments of the present invention. However, the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modifications made under the technical spirit disclosed in the present invention should fall within the patent scope covered by the present invention.

Claims

1. A power system forward scheduling method based on a migratory cost function approximation, characterized in that, Includes the following steps: S1. Construct a forward-looking power system dispatch model based on a cost function, with the goal of minimizing the dispatch cost of the power system, and determine the constraints on the parameters of the state variables and decision variables in the cost function; S2. Based on the cost function approximation theory and considering the influence of random information, the objective function is transformed, and the influence of random information on the cost function is estimated by introducing approximate parameters to obtain an approximate cost function; S3. Establish the mapping relationship between approximate parameters and random information, use a Long Short-Term Memory Neural Network (LSTM) for fitting, identify the parameters of the LSTM, and obtain the optimal parameters of the approximate parameters; S4. Input the optimal parameters of the approximate parameters into the power system forward scheduling model. The power system forward scheduling model outputs a real-time scheduling optimization scheme. The power system is then scheduled according to the real-time scheduling optimization scheme. The objective function is: ； wherein is a cost function for the period t, is a decision variable for the period t, is a state variable for the period t, is a cost function for the period τ, is a decision variable for the period τ, is a state variable for the period τ, Ω = {t t, t +2 t,…, t +H t}; The cost function of the t period is The calculation formula is: ； wherein, Cg(t) is the cost of the generator g at time period t, Cst(t) is the charge-discharge cost of the energy storage at time period t, Cp(t) is the curtailment penalty cost of the renewable energy and the load at time period t; State variables in time period t , The generator power during time period t-1 The remaining electricity stored during time period t. This represents the predicted value of renewable energy in time period t. This represents the predicted load value for time period t; The decision variables for time period t are , Let t be the generator power during time period t. The energy storage charging and discharging power during time period t. Let t represent the amount of renewable energy reduction during period t. The load reduction amount for period t; and related, and related, and and related.

2. The power system look-ahead scheduling method based on migration cost function approximation according to claim 1, characterized in that, To ensure the safety of the power system, the parameters of the state variables and decision variables must satisfy the following constraints: ； ； ； ； ； ； In the formula, Let be the generator power of node i during time period t; Let be the predicted value of renewable energy for node i in time period t; The energy storage charging power of node i during time period t; Let be the energy storage discharge power of node i during time period t; Let be the predicted load of node i during time period t; Let be the amount of renewable energy reduction at node i during time period t; Let be the load reduction amount of node i during time period t; For a set of nodes; and These are sets of nodes that have generators and renewable energy sources, respectively; G ij and B ij e represents the real and imaginary parts of the node admittance matrix in the i-th row and j-th column; i,t and f i,t e represents the real and imaginary parts of the voltage phasor at node i at time t; j,t and f j,t It represents the real and imaginary parts of the voltage phasor at node j during time period t; Let be the reactive load of node i during time period t; Let be the reactive power of the generator at node i during time period t.

3. The power system forward scheduling method based on migration cost function approximation according to claim 2, characterized in that, Based on the cost function approximation theory, the objective function of the power system forward scheduling model is transformed into: ； In the formula, This represents the cost function for the current time period t. This represents the cost function for the subsequent τ time intervals, due to the presence of random information ω. It is unknown.

4. The power system look-ahead scheduling method based on migration cost function approximation according to claim 3, characterized in that, By introducing an approximation parameter β to estimate the influence of the random information, the approximation parameter β includes β NG and β load . After introducing the approximation parameter β, we get: ； ； ； ； ； in, It is the approximate cost function after introducing approximate parameters, β NG β is an approximate parameter for the predicted value of renewable energy. load These are approximate parameters for the load forecast.

5. The power system look-ahead scheduling method based on migration cost function approximation according to claim 4, characterized in that, In order to accurately estimate the approximate parameter β The influence of random information necessitates minimizing the difference between the approximate cost function and the actual cost function, resulting in: ； in, Let F be the gradient of F with respect to β; J be the Jacobian matrix; α be the iteration step size; after a certain number of iterations, the optimal parameters of the cost function approximation optimization method can be obtained. .

6. The power system look-ahead scheduling method based on migration cost function approximation according to claim 5, characterized in that, The optimal parameters are solved using the stochastic gradient descent method. The process of updating the approximate parameter β is as follows: ； ； in, α is the gradient of F with respect to β; J is the Jacobian matrix; α is the iteration step size; After iterative updates, the optimal parameters of the cost function approximation optimization method can be obtained. .

7. The power system forward scheduling method based on migration cost function approximation according to claim 6, characterized in that, To improve the computational efficiency of iterative updates of the approximate parameter β, transfer learning is used to directly obtain the approximate parameter β based on the random information ω. A transfer function is then used to map the relationship between the approximate parameter β and the random information ω. The mapping relationship is as follows: ； In the formula, The constructed transition function.

8. The power system look-ahead scheduling method based on migration cost function approximation according to claim 7, characterized in that, The Long Short-Term Memory (LSTM) neural network was used for fitting, and... This indicates that the parameters of the LSTM model are identified, so that... as close as possible ,get: ； Where p represents the parameter vector of the LSTM model, This represents the optimal value of p. express and The error value, This represents the transfer function fitted by the LSTM.

9. The power system look-ahead scheduling method based on migration cost function approximation according to claim 8, characterized in that, The optimal value of the parameter vector p of the LSTM is obtained using the adaptive moment estimation method, and the update equation is: ； ； ； ； ； In the formula, This represents the parameter vector for the c-th iteration; This represents the parameter vector for the (c-1)th iteration; This indicates that the gradient of the parameter vector p is calculated. Indicates the error value; Indicates the learning rate; This represents the gradient of the exponential moving average in the c-th iteration. This represents the gradient of the modified exponential moving average in the c-th iteration. This represents the biased quadratic primitive moment estimate for the c-th iteration; This represents the biased quadratic primitive moment estimate for the (c-1)th iteration; Let ρ1, ρ2, and ρ3 represent the corrected bias quadratic primitive moment estimate for the c-th iteration; It is a constant.

Images