A virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition

Abstract

This invention discloses a virtual power plant optimization scheduling method based on parameter-optimized variational mode decomposition. The technical solution of the invention includes the following steps: First, the whale optimization algorithm is used to solve for the number of decomposed modes and the second-order penalty factor in the variational mode decomposition, and to find the optimal parameter combination for variational mode decomposition; then, according to preset parameters, variational mode decomposition is performed on the predicted photovoltaic, wind power, and load power in the virtual power plant to obtain a series of subsequences with frequencies from low to high; the subsequences obtained by variational mode decomposition are reconstructed into high-frequency components and low-frequency components using mutual information entropy; finally, based on the frequency characteristics of the subsequences and considering the output characteristics of controllable equipment, a virtual power plant optimization scheduling model is established, and the frequency-division scheduling results of each resource in the virtual power plant are added together to obtain the optimized scheduling plan. The frequency-division optimization scheduling model proposed in this invention can perform differentiated control of battery storage and gas turbines in the virtual power plant for different frequency bands, realizing the optimized utilization of resources.

Description

Technical Field

[0001] This invention relates to the field of power systems, specifically a virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition. Technical Background

[0002] The "dual carbon" goal has driven the rapid development of new energy sources, but the imbalance between power generation and load after the integration of new energy sources poses a significant challenge to the safe and economical operation of the power grid. By involving virtual power plants (VPPs) with uncertain power sources and loads in system optimization and dispatch, the impact of the uncertainty of new energy sources and loads on the power grid can be effectively reduced, enabling new energy sources with unstable output to maintain a certain level of competitiveness in the electricity market bidding. Summary of the Invention

[0003] The purpose of this invention is to propose a virtual power plant (VPP) optimization scheduling method based on parameter-optimized variational mode decomposition. This invention processes the original power curves of wind power, photovoltaic power, and load using a variational mode decomposition algorithm optimized by the whale algorithm, and reconstructs new high- and low-frequency components using mutual information entropy. Considering the frequency characteristics of the high- and low-frequency components and the different operating characteristics of gas turbines and battery energy storage, separate high- and low-frequency optimization scheduling models for the VPP are designed.

[0004] This invention adopts the following technical solution: A virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition is established, comprising the following steps:

[0005] (1) The whale optimization algorithm is used to solve the number of decomposition modes and the second-order penalty factor in variational mode decomposition, and to find the optimal parameter combination of variational mode decomposition;

[0006] (2) Using preset parameters, variational mode decomposition is performed on the predicted photovoltaic, wind power and load power in the virtual power plant to obtain a series of subsequences with frequencies from low to high.

[0007] (3) Use mutual information entropy to reconstruct the subsequences obtained by variational mode decomposition into high-frequency components and low-frequency components;

[0008] (4) Based on the different operating characteristics of gas turbines and battery energy storage, optimize scheduling models are designed respectively. The frequency division scheduling results of each resource in the virtual power plant are added together to obtain the optimized scheduling plan.

[0009] Specifically, in step (1), the whale algorithm is used to find the optimal combination of parameters for variational mode decomposition, as described below:

[0010] The local minimum envelope entropy after Variational Mode Decomposition (VMD) is used as the fitness function in the whale optimization algorithm to search for the optimal VMD parameter combination K and α, i.e., the number of modal components and the quadratic penalty factor. Envelope entropy E p The calculation formula is:

[0011]

[0012] In the formula: a(k) is the envelope signal after Hilbert modulation; p k This is the normalized form of a(k);

[0013] The whale algorithm optimizes VMD parameters K and α in the following steps:

[0014] Step 1: Determine the fitness function. Let K and α be the solutions for individual whales, and initialize the algorithm parameters.

[0015] Step 2: Perform VMD decomposition on the signal at different positions K and α to obtain the subsequence set {u k}, calculate each subsequence u k The envelope entropy;

[0016] Step 3: Perform a global search on the system using the minimum envelope entropy value as the fitness function;

[0017] Step 4: Update the position of the individual whales. Repeat steps 3 and 4. Once the envelope entropy value is minimized or the maximum number of iterations is reached, output the optimal parameter combination K and α.

[0018] Step 5: Set the optimal combination of parameters K and α to perform VMD decomposition on the signal.

[0019] Specifically, in step (2), using preset parameters, variational mode decomposition is performed on the predicted photovoltaic, wind power, and load power in the virtual power plant to obtain a series of subsequences with frequencies from low to high, as described below:

[0020] VMD decomposes an input signal with frequency domain characteristics into a set containing multiple mode functions. It performs adaptive decomposition on the original signal. The constrained variational model for decomposing the input signal f(t) is shown in equation (2):

[0021]

[0022] Where: t is the time series of the input signal; {u k}={u1,u2,…,u k} represents the set of modal components after VMD decomposition; {ω k}={ω1,ω2,…,ωk} represents the set of center frequencies corresponding to each component; δ(t) is the impulse function; * is the convolution operator; K is the preset number of modal components after decomposition.

[0023] To make the variational problem solvable, the Lagrange operator λ and the quadratic penalty factor α are introduced into the solution process. The augmented Lagrangian function is then:

[0024]

[0025] Solving equation (3) generally employs the alternating direction method of the multiplication operator, which decomposes the input signal f(t) into K sub-components. The updated decomposition result u... k ω k The expression is:

[0026]

[0027]

[0028] In the formula: Wiener filtering for the current signal; ∧ represents the centroid of the power spectrum of the current mode function; ∧ denotes the Fourier transform.

[0029] Based on equations (4) and (5), solve the problem iteratively and update u. k and ω k And substitute it into equation (3) to update λ:

[0030]

[0031] In the formula: ν is the Lagrange multiplier update coefficient. For a given discrimination accuracy ε>0, there exists an iteration stopping condition:

[0032]

[0033] If ε satisfies the iteration stopping condition (7), the loop ends and K modal components are output to achieve adaptive decomposition of the input signal; if the iteration stopping condition is not satisfied, the result of this iteration is used as the initial value and substituted back into equations (4) and (5) to start a new round of iteration.

[0034] Specifically, in step (3), the subsequence obtained by variational mode decomposition is reconstructed into high-frequency components and low-frequency components using mutual information entropy, as described below:

[0035] Introducing the concept of energy entropy, let's assume that the energy of mode component k at time t is u. k (t), after normalization, its energy entropy is p t (u k ), can be represented as:

[0036]

[0037]

[0038] In the formula: M is the number of time periods within the scheduling cycle; E k For modal components u k The total energy during the scheduling period.

[0039] Mutual information provides a measure for analyzing the correlation between random variables; however, information itself is not measurable. Therefore, this invention utilizes the concept of entropy, using information entropy as a quantifiable indicator of information, defining the information entropy H of the modal components as:

[0040]

[0041] In the formula: H(u) k ) represents the modal component u k Information entropy;

[0042] Furthermore, the mutual information entropy I between two adjacent modal components is expressed as:

[0043] I(u k ,u k+1 )=H(u k )+H(u k+1 )-H(u k u k+1 (11)

[0044] In the formula: I(u k ,u k+1 ) for u k with u k+1 Information entropy between;

[0045] Based on the frequency, the mutual information entropy of adjacent modal components after VMD decomposition exhibits a pattern of first decreasing, then increasing, and then decreasing again. According to information theory, mutual information entropy reflects the correlation between two variables; the lower the correlation, the lower the mutual information entropy. When the two variables are independent, the mutual information entropy is at its minimum, equal to zero. Therefore, the minimum point can be chosen as the frequency boundary between high-frequency and low-frequency components to avoid aliasing of adjacent modal components at the boundary. Based on the frequency boundary point, the modal components {u} after VMD decomposition can be further classified. k Further reconstruction into low-frequency component P d and high-frequency component P g Two main components, namely:

[0046]

[0047] In the formula: i is the boundary point between high-frequency and low-frequency modal components.

[0048] Specifically, in step (4), optimized scheduling models are designed according to the different operating characteristics of gas turbines and battery energy storage. The optimized scheduling plan is obtained by adding the frequency division scheduling results of each resource in the virtual power plant as follows:

[0049] Considering time-of-use pricing and the operation and maintenance costs of each device, and with the goal of maximizing the benefits of VPP high-frequency and low-frequency scheduling respectively, the objective function for VPP optimal scheduling is constructed as follows:

[0050]

[0051] In the formula: F g F d The total revenue of the high-frequency and low-frequency scheduling models; For revenue from high- and low-frequency large power grid transactions; I b (t) represents the revenue from battery energy storage; For high and low frequency wind power operating costs; For high and low frequency photovoltaic operating costs; C m (t) represents the operating cost of the gas turbine; The revenue generated by high- and low-frequency VPP power sources is calculated. The specific expressions for each part are as follows.

[0052] (1) Electricity trading revenue

[0053] Electricity trading revenue is obtained through transactions with the main power grid under time-of-use pricing.

[0054]

[0055] In the formula: The power traded with the main power grid at time t represents high and low frequency power; a value greater than zero indicates power sales, and a value less than zero indicates power purchase. e λ is a 0-1 variable, where 0 represents electricity sales and 1 represents electricity purchases; se , λ be This refers to the time-of-use electricity sales and purchase prices.

[0056] (2) Battery energy storage revenue

[0057] I b (t)=λ l P b (t)-μ b |P b (t)| (15)

[0058] In the formula: λ l For time-of-use pricing; P b (t) represents the battery's energy output at time t, with positive for discharging and negative for charging; μ b This is the depreciation factor for battery energy storage equipment.

[0059] (3) Distributed power generation cost

[0060] The main power sources within a VPP include wind power, solar power, and gas turbines, with their respective operating costs as follows:

[0061]

[0062] In the formula: The output of the wind turbine at high and low frequency time t; Photovoltaic power output at high and low frequency time t; P m (t) represents the gas turbine output at time t; λ w , λ p For wind power and solar power operating cost coefficients; a m b m c m This represents the cost coefficient for gas turbines.

[0063] (4) Revenue from power generation

[0064] Power generation revenue describes the revenue from selling electricity generated by wind power, solar power, and gas turbines within a VPP under time-of-use pricing.

[0065]

[0066] The operation of a virtual power plant is subject to the following constraints:

[0067] (1) Power balance constraint

[0068]

[0069] In the formula: For high and low frequency loads within the VPP.

[0070] (2) Power interaction with the power grid

[0071] Considering line safety factors, there are upper and lower limits constrained for the interaction power between the VPP and the main power grid.

[0072]

[0073] In the formula: This sets the upper limit for the power that the VPP can supply to the grid.

[0074] (3) Gas turbine constraints

[0075]

[0076] In the formula: The upper and lower limits of the gas turbine output; △P m This represents the upper limit of the slope rate.

[0077] (4) Battery energy storage constraints

[0078] Battery energy storage can compensate for power imbalances during VPP operation through charging and discharging, thereby improving the stability and reliability of VPP operation. Regarding charging efficiency η... c Discharge efficiency η d Capacity E b The state of charge (SOC) expression for battery energy storage is as follows:

[0079] In the formula: SOC(t-1) and SOC(t) are the SOC of the battery at times t-1 and t, respectively; △T is the scheduling time interval.

[0080] Considering battery energy storage SOC and charge / discharge power constraints, and to ensure scheduling continuity, the constraint that the SOC is equal at the beginning and end of the battery energy storage scheduling cycle is added:

[0081]

[0082] Where: SOC max SOC min The upper and lower limits of battery energy storage SOC; This is the upper limit of battery energy storage output.

[0083] The virtual power plant operation aims at optimal economic efficiency. Considering constraints, the particle swarm optimization algorithm is used to complete day-ahead optimal scheduling and obtain the power output plans for gas turbines, wind power, photovoltaics, batteries, and tie lines.

[0084] The technical solution provided by this invention has the following beneficial effects:

[0085] By introducing virtual power plant technology, regional wind power, photovoltaics, fluctuating loads, gas turbines, and energy storage devices are integrated into a virtual power plant. The whale algorithm is used to optimize the number of VMD modal components and the penalty factor, which can reduce signal reconstruction errors. Based on the parameter optimization VMD algorithm, the original power of wind power, photovoltaics, and loads is decomposed into high-frequency and low-frequency components. Then, the battery energy storage and gas turbine in the VPP are differentiated for different frequency bands, realizing the optimized utilization of resources. Attached Figure Description

[0086] The present invention will be further described below with reference to the accompanying drawings:

[0087] Figure 1 This is a flowchart of the present invention;

[0088] Figure 2 The flowchart of the VMD algorithm optimized with whale algorithm parameters;

[0089] Figure 3 This is a graph showing the VMD decomposition results of the load curve;

[0090] Figure 4 Comparison chart of results with and without parameter optimization;

[0091] Figure 5 A comparison chart of results under different scheduling strategies; Detailed Implementation Plan

[0092] To better understand the purpose, technical solution, and technical effects of this invention, the invention will be further explained and described below in conjunction with the accompanying drawings.

[0093] This invention proposes a virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition. Figure 1 The flowchart of this invention includes the following detailed steps.

[0094] Step 1 uses the whale optimization algorithm to solve for the number of decomposition modes and the quadratic penalty factor in variational mode decomposition, and finds the optimal parameter combination for variational mode decomposition:

[0095] The local minimum envelope entropy after Variational Mode Decomposition (VMD) is used as the fitness function in the whale optimization algorithm to search for the optimal VMD parameter combination K and α, i.e., the number of modal components and the quadratic penalty factor. Envelope entropy E p The calculation formula is:

[0096]

[0097] In the formula: a(k) is the envelope signal after Hilbert modulation; p k This is the normalized form of a(k);

[0098] The whale algorithm optimizes VMD parameters K and α in the following steps:

[0099] Step 1: Determine the fitness function. Let K and α be the solutions for individual whales, and initialize the algorithm parameters.

[0100] Step 2: Perform VMD decomposition on the signal at different positions K and α to obtain the subsequence set {u k}, calculate each subsequence u k The envelope entropy;

[0101] Step 3: Perform a global search on the system using the minimum envelope entropy value as the fitness function;

[0102] Step 4: Update the position of the individual whales. Repeat steps 3 and 4. Once the envelope entropy value is minimized or the maximum number of iterations is reached, output the optimal parameter combination K and α.

[0103] Step 5: Set the optimal combination of parameters K and α to perform VMD decomposition on the signal.

[0104] Step 2 uses preset parameters to perform variational mode decomposition on the predicted photovoltaic, wind power, and load power in the virtual power plant, obtaining a series of subsequences with frequencies ranging from low to high;

[0105] VMD decomposes an input signal with frequency domain characteristics into a set containing multiple mode functions. It performs adaptive decomposition on the original signal. The constrained variational model for decomposing the input signal f(t) is shown in equation (2):

[0106]

[0107] Where: t is the time series of the input signal; {u k}={u1,u2,…,u k} represents the set of modal components after VMD decomposition; {ω k}={ω1,ω2,…,ω k} represents the set of center frequencies corresponding to each component; δ(t) is the impulse function; * is the convolution operator; K is the preset number of modal components after decomposition.

[0108] To make the variational problem solvable, the Lagrange operator λ and the quadratic penalty factor α are introduced into the solution process. The augmented Lagrangian function is then:

[0109]

[0110] Solving equation (3) generally employs the alternating direction method of the multiplication operator, which decomposes the input signal f(t) into K sub-components. The updated decomposition result u... k ω k The expression is:

[0111]

[0112]

[0113] In the formula: Wiener filtering for the current signal; ∧ represents the centroid of the power spectrum of the current mode function; ∧ denotes the Fourier transform.

[0114] Based on equations (4) and (5), solve the problem iteratively and update u. k and ω k And substitute it into equation (3) to update λ:

[0115]

[0116] In the formula: ν is the Lagrange multiplier update coefficient. For a given discrimination accuracy ε>0, there exists an iteration stopping condition:

[0117]

[0118] If ε satisfies the iteration stopping condition (7), the loop ends and K modal components are output to achieve adaptive decomposition of the input signal; if the iteration stopping condition is not satisfied, the result of this iteration is used as the initial value and substituted back into equations (4) and (5) to start a new round of iteration.

[0119] Step 3 uses mutual information entropy to reconstruct the subsequences obtained from variational mode decomposition into high-frequency and low-frequency components;

[0120] Introducing the concept of energy entropy, let's assume that the energy of mode component k at time t is u. k (t), after normalization, its energy entropy is p t (u k ), can be represented as:

[0121]

[0122]

[0123] In the formula: M is the number of time periods within the scheduling cycle; E k For modal components u k The total energy during the scheduling period.

[0124] Mutual information provides a measure for analyzing the correlation between random variables; however, information itself is not measurable. Therefore, this invention utilizes the concept of entropy, using information entropy as a quantifiable indicator of information, defining the information entropy H of the modal components as:

[0125]

[0126] In the formula: H(u) k ) represents the modal component u k Information entropy;

[0127] Furthermore, the mutual information entropy I between two adjacent modal components is expressed as:

[0128] I(u k ,u k+1 )=H(u k )+H(u k+1 )-H(u k u k+1 (11)

[0129] In the formula: I(u k ,u k+1 ) for u k with u k+1 Information entropy between;

[0130] Based on the frequency, the mutual information entropy of adjacent modal components after VMD decomposition exhibits a pattern of first decreasing, then increasing, and then decreasing again. Therefore, the minimum point can be selected as the frequency boundary between high-frequency and low-frequency components. Based on the frequency boundary point, the modal components {u} after VMD decomposition can be further classified. k Further reconstruction into low-frequency component P d and high-frequency component P g Two main components, namely:

[0131]

[0132] In the formula: i is the boundary point between high-frequency and low-frequency modal components.

[0133] In summary, the VMD algorithm flowchart after parameter optimization using the whale algorithm can be found here. Figure 2 As shown.

[0134] Step 4: Design optimized scheduling models according to the different operating characteristics of gas turbines and battery energy storage. Add the frequency-division scheduling results of each resource in the virtual power plant to obtain the optimized scheduling plan.

[0135] Considering time-of-use pricing and the operation and maintenance costs of each device, and with the goal of maximizing the benefits of VPP high-frequency and low-frequency scheduling respectively, the objective function for VPP optimal scheduling is constructed as follows:

[0136]

[0137] In the formula: F g F d The total revenue of the high-frequency and low-frequency scheduling models; For revenue from high- and low-frequency large power grid transactions; I b (t) represents the revenue from battery energy storage; For high and low frequency wind power operating costs; For high and low frequency photovoltaic operating costs; C m (t) represents the operating cost of the gas turbine; The revenue generated by high- and low-frequency VPP power sources is calculated. The specific expressions for each part are as follows.

[0138] (1) Electricity trading revenue

[0139] Electricity trading revenue is obtained through transactions with the main power grid under time-of-use pricing.

[0140]

[0141] In the formula: The power traded with the main power grid at time t represents high and low frequency power; a value greater than zero indicates power sales, and a value less than zero indicates power purchase. e λ is a 0-1 variable, where 0 represents electricity sales and 1 represents electricity purchases; se , λbe This refers to the time-of-use electricity sales and purchase prices.

[0142] (2) Battery energy storage revenue

[0143] I b (t)=λ l P b (t)-μ b |P b (t)| (15)

[0144] In the formula: λ l For time-of-use pricing; P b (t) represents the battery's energy output at time t, with positive for discharging and negative for charging; μ b This is the depreciation factor for battery energy storage equipment.

[0145] (3) Distributed power generation cost

[0146] The main power sources within a VPP include wind power, solar power, and gas turbines, with their respective operating costs as follows:

[0147]

[0148] In the formula: The output of the wind turbine at high and low frequency time t; Photovoltaic power output at high and low frequency time t; P m (t) represents the gas turbine output at time t; λ w , λ p For wind power and solar power operating cost coefficients; a m b m c m This represents the cost coefficient for gas turbines.

[0149] (4) Revenue from power generation

[0150] Power generation revenue describes the revenue from selling electricity generated by wind power, solar power, and gas turbines within a VPP under time-of-use pricing.

[0151]

[0152] The operation of a virtual power plant is subject to the following constraints:

[0153] (1) Power balance constraint

[0154]

[0155] In the formula: For high and low frequency loads within the VPP.

[0156] (2) Power interaction with the power grid

[0157] Considering line safety factors, there are upper and lower limits constrained for the interaction power between the VPP and the main power grid.

[0158]

[0159] In the formula: This sets the upper limit for the power that the VPP can supply to the grid.

[0160] (3) Gas turbine constraints

[0161]

[0162] In the formula: The upper and lower limits of the gas turbine output; △P m This represents the upper limit of the slope rate.

[0163] (4) Battery energy storage constraints

[0164] Battery energy storage can compensate for power imbalances during VPP operation through charging and discharging, thereby improving the stability and reliability of VPP operation. Regarding charging efficiency η... c Discharge efficiency η d Capacity E b The state of charge (SOC) expression for battery energy storage is as follows:

[0165] In the formula: SOC(t-1) and SOC(t) are the SOC of the battery at times t-1 and t, respectively; △T is the scheduling time interval.

[0166] Considering battery energy storage SOC and charge / discharge power constraints, and to ensure scheduling continuity, the constraint that the SOC is equal at the beginning and end of the battery energy storage scheduling cycle is added:

[0167]

[0168] Where: SOC max SOC min The upper and lower limits of battery energy storage SOC; This is the upper limit of battery energy storage output.

[0169] The virtual power plant operation aims at optimal economic efficiency. Considering constraints, the particle swarm optimization algorithm is used to complete day-ahead optimal scheduling and obtain the power output plans for gas turbines, wind power, photovoltaics, batteries, and tie lines.

[0170] To further understand this invention, an energy system in a development zone was selected as a simulation example for verification. The VPP includes a 60MW wind farm, a 15MW photovoltaic power station, a 30MW gas turbine, and 5MW / 10MW·h battery energy storage. The example considers time-of-use pricing: from 6:00 AM to 2:00 PM, the purchase and sale prices are RMB 0.68 and RMB 0.52, respectively, with a load price of RMB 0.6; from 3:00 PM to 10:00 PM, the purchase and sale prices are RMB 1.39 and RMB 1.18, respectively, with a load price of RMB 1.29; and during other times, the purchase and sale prices are RMB 0.41 and RMB 0.28, respectively, with a load price of RMB 0.35. Other VPP parameter settings are shown in Table 1.

[0171] Table 1 VPP Parameters

[0172]

[0173]

[0174] For the raw power signals of 96 uniformly sampled points of wind power, photovoltaic power and load on a certain day, this study uses the VMD algorithm with K and α values ​​optimized by the whale algorithm to decompose them, and uses mutual information entropy to obtain the boundary points i of their high and low frequency mode components. The results are shown in Table 2.

[0175] Table 2. Optimized VMD parameters and high / low frequency boundary points

[0176]

[0177] The load power decomposition is illustrated as an example. As shown in Table 2, the load power curve is decomposed into five modal components, u1 to u5. Figure 3 The five modal components of the load power curve reflect its variation trend and fluctuation at different frequencies. Among them, u1 and u2 describe the general shape of the daily load curve, u3 describes the variation trend, and u4 and u5 describe the fluctuation, characterizing the details of the random variation of the load.

[0178] To illustrate the advantages of the parameter-optimized VMD algorithm, this paper compares the load curve decomposition results with and without parameter optimization. In the unoptimized VMD algorithm, K and α were empirically selected as 4 and 2000, respectively. The comparison results are as follows: Figure 4 As shown, the sum of the modes obtained by VMD decomposition after parameter optimization almost completely coincides with the original load curve, while the decomposition result without parameter optimization differs from the original load curve when the load fluctuates greatly, and the signal reconstruction result is relatively unsatisfactory.

[0179] Figure 5To compare the power output of gas turbines and battery storage using a VMD-based two-layer scheduling strategy with that of a traditional single-layer scheduling strategy, the designed power decomposition method based on parameter optimization VMD is compared with the traditional scheduling strategy without VMD decomposition. Figure 5 This indicates that in low-frequency power optimization scheduling involving gas turbines, power changes are relatively slow with no sudden changes. The maximum output occurs around 8:00 AM, at 24MW, and the gas turbines operate below their maximum output for extended periods, which helps ensure the reliability of gas turbine power supply. In high-frequency power optimization scheduling involving battery storage, output changes frequently, especially between 2:00 PM and 4:00 PM. These rapid output changes effectively track fluctuations in renewable energy and load. Compared to traditional scheduling strategies, while meeting load demands, the battery storage charging and discharging power remains below its maximum technical output for most of the time, within 2MW, which helps extend its lifespan and reduce VPP operation and maintenance costs.

Claims

1. A virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition, characterized in that, Including the following steps: (1) The whale optimization algorithm is used to solve the number of decomposition modes and the quadratic penalty factor in variational mode decomposition, and to find the optimal parameter combination of variational mode decomposition; (2) Using preset parameters, variational mode decomposition is performed on the predicted photovoltaic, wind power and load power in the virtual power plant to obtain a series of subsequences with frequencies from low to high; (3) Use mutual information entropy to reconstruct the subsequences obtained from variational mode decomposition into high-frequency components and low-frequency components; (4) Design optimization scheduling models according to the different operating characteristics of gas turbines and battery energy storage, and add up the frequency division scheduling results of each resource in the virtual power plant to obtain the optimization scheduling plan; In step (4), an optimized scheduling model is designed according to the different operating characteristics of the gas turbine and battery energy storage. The frequency division scheduling results of each resource in the virtual power plant are added together to obtain the optimized scheduling plan: Considering time-of-use pricing and the operation and maintenance costs of each device, and with the goal of maximizing the benefits of VPP high-frequency and low-frequency scheduling respectively, the objective function for VPP optimal scheduling is constructed as follows: （1） In the formula: , The total revenue of the high-frequency and low-frequency scheduling models; , For revenue from high- and low-frequency large power grid transactions; For battery energy storage revenue; , For high and low frequency wind power operating costs; , To reduce the operating costs of high-frequency and low-frequency photovoltaic systems; For gas turbine operating costs; , The revenue generated by high- and low-frequency VPP power sources is calculated; the specific expressions for each part are as follows; (1) Electricity trading revenue Electricity trading revenue is obtained through transactions with the main power grid under time-of-use pricing: （2） In the formula: , The power traded with the power grid at time t represents high and low frequency power; a value greater than zero indicates power sales, and a value less than zero indicates power purchase. These are 0-1 variables, where 0 represents electricity sales and 1 represents electricity purchases. , For time-of-use electricity sales and electricity purchase prices; (2) Battery energy storage revenue （3） In the formula: Time-of-use pricing; The battery's energy output at time t is positive for discharging and negative for charging; This refers to the depreciation factor for battery energy storage equipment. (3) Distributed power supply cost The power sources within a VPP include wind power, solar power, and gas turbines, with their respective operating costs as follows: （4） In the formula: , The output of the wind turbine at high and low frequency time t; , Photovoltaic power output at high and low frequency time t; Let t be the output of the gas turbine; , This refers to the operating cost coefficient for wind power and solar power. , , This represents the cost coefficient for gas turbines. (4) Revenue from power generation Power generation revenue describes the revenue from selling electricity generated by wind, solar, and gas turbines within a VPP under time-of-use pricing: （5） The operation of a virtual power plant is subject to the following constraints: (1) Power balance constraint （6） In the formula: , For high and low frequency loads within the VPP; (2) Power interaction with the power grid Considering line safety factors, there are upper and lower limits constrained for the interaction power between the VPP and the main power grid: （7） In the formula: —The maximum power that a VPP can supply to the grid; (3) Gas turbine constraints （8） In the formula: , The upper and lower limits of gas turbine output; This represents the upper limit of the gradient rate. (4) Battery energy storage constraints Battery energy storage can compensate for power imbalances during VPP operation through charging and discharging, improving the stability and reliability of VPP operation and increasing charging efficiency. Discharge efficiency ,capacity The state of charge (SOC) expression for battery energy storage is as follows: （9） In the formula: , Let SOC be the state of the battery at times t-1 and t; The scheduling time interval; Considering battery energy storage SOC and charge / discharge power constraints, and to ensure scheduling continuity, the constraint that the SOC is equal at the beginning and end of the battery energy storage scheduling cycle is added: （10） In the formula: , The upper and lower limits of battery energy storage SOC; This represents the upper limit of battery energy storage output. M represents the number of time periods within the scheduling cycle; The virtual power plant operation aims at optimal economic efficiency. Considering constraints, the particle swarm optimization algorithm is used to complete day-ahead optimal scheduling and obtain the power output plans for gas turbines, wind power, photovoltaics, batteries, and tie lines.

2. The virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition according to claim 1, characterized in that, The process of optimizing variational mode decomposition parameters using the whale algorithm in step (1) is as follows: The local minimum envelope entropy after variational mode decomposition (VMD) is used as the fitness function in the whale optimization algorithm to search for the optimal combination of VMD parameters. , That is, the number of modal components and the quadratic penalty factor; the envelope entropy of the k-th subsequence. The calculation formula is: （11） In the formula: The sampling point at time t represents the envelope signal of the signal after Hilbert modulation. for The normalized form; M is the number of time periods within the scheduling cycle; The whale algorithm optimizes VMD parameters K and α in the following steps: Step 1: Determine the fitness function, let the solution for an individual whale be... , Initialize the algorithm parameters; Step 2: In different locations , The signal is decomposed using VMD to obtain a set of subsequences. Calculate each subsequence The envelope entropy; Step 3: Perform a global search on the system using the minimum envelope entropy value as the fitness function; Step 4: Update the position of the individual whales. Repeat steps 3 and 4. Once the envelope entropy value is minimized or the maximum number of iterations is reached, output the optimal parameter combination. , ; Step 5: Set optimal parameters , The combined signal is subjected to VMD decomposition.

3. The virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition according to claim 1, characterized in that, In step (2), the variational mode decomposition process is as follows: Virtual Mode Decomposition (VMD) decomposes an input signal with frequency domain characteristics into a set containing multiple mode functions, performing adaptive decomposition on the original signal. When performing decomposition, the constrained variational model is as shown in equation (12): （12） In the formula: The input signal is a time series; It is the set of modal components after VMD decomposition; This is the set of center frequencies corresponding to each component; It is a pulse function; This is the convolution operator; The preset number of modal components after decomposition; To make the variational problem solvable, the Lagrange operator is introduced in the solution process. and secondary penalty factor Then the augmented Lagrange function is: （13） By using the alternating direction method of multiplication operators to solve equation (13), the input signal can be obtained. Decomposed into Each subcomponent; where the updated decomposition result , The expression is: （14） （15） In the formula: Wiener filtering for the current signal; The centroid of the power spectrum of the current mode function; Indicates Fourier transform; Based on equations (14) and (15), iteratively solve and update. and And substitute into equation (13) to update : （16） In the formula: Let Lagrange multipliers be the update coefficients; for a given discrimination accuracy... There exists an iteration stopping condition: （17） like If the iteration stopping condition (17) is satisfied, the loop ends and the output is given. Each modal component is used to achieve adaptive decomposition of the input signal; if the iteration stopping condition is not met, the result of this iteration is used as the initial value and substituted back into equations (14) and (15) to start a new round of iteration.

4. The virtual power plant optimization scheduling method based on parameter optimization variational mode decomposition according to claim 1, characterized in that, In step (3), the subsequence obtained by variational mode decomposition is reconstructed into high-frequency components and low-frequency components using mutual information entropy: Introducing the concept of energy entropy, let's assume the energy of mode component k at time t is... After normalization, its energy entropy is , can be represented as: （18） （19） In the formula: M is the number of time periods within the scheduling cycle; E k For modal components u k The total energy during the scheduling period; Mutual information provides a measure for analyzing the correlation between random variables; however, information itself is not measurable. Therefore, using the concept of entropy, information entropy is used as a quantification of information, and the information entropy H of the modal components is defined as: (20) In the formula: Modal components Information entropy; Furthermore, the mutual information entropy I between two adjacent modal components is expressed as: （21） In the formula: for and Information entropy between; Based on the frequency, the mutual information entropy of adjacent modal components after VMD decomposition exhibits a pattern of first decreasing, then increasing, and then decreasing again. Therefore, the minimum point can be selected as the frequency boundary between high-frequency and low-frequency components. Based on this frequency boundary, the modal components after VMD decomposition can be further analyzed. Further reconstruction into low-frequency component P d and high-frequency component P g Two main components, namely: (22) In the formula: It is the dividing point between high-frequency and low-frequency modal components.

Images