Power system random scheduling method based on wind-solar joint probability distribution and double-layer dynamic optimization

Abstract

The invention discloses a power system random scheduling method based on wind-solar joint probability distribution and double-layer dynamic optimization. The method comprises the following steps: firstly, constructing a wind and light output joint probability distribution model by adopting nonparametric kernel density estimation and a Frank Copula function so as to accurately describe space-time correlation; secondly, generating a wind-solar combined output scene through Monte Carlo simulation, and performing reduction by adopting a k-means + + clustering algorithm to obtain a typical scene set; secondly, establishing a double-layer optimization model, wherein the upper layer takes the minimum net load variance as a target to coordinate wind and light storage output stabilizing fluctuation; and the lower layer optimizes the output of the traditional unit by taking the minimum operation cost of the system as a target. Finally, a dynamic climbing constraint mechanism is introduced, the net load fluctuation standard deviation output by the upper layer is converted into a dynamic constraint threshold value of the climbing rate of the lower layer unit, collaborative optimization of economical efficiency and stability is achieved, and the dispatching robustness of the high-proportion renewable energy power system is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system optimal scheduling, in particular to a power system stochastic scheduling method based on wind-solar joint probability distribution and double-layer dynamic optimization. BACKGROUND

[0002] According to the Renewable Energy Development Report 2023 issued by the National Energy Administration, the installed capacity of wind and photovoltaic power in China has reached 41.2%, but the curtailment rate of wind and photovoltaic power is still as high as 5.8%. The strong randomness of wind and photovoltaic power output (the standard deviation of wind power output fluctuation is ±30% of the rated power, and photovoltaic power output presents significant "anti-peak regulation" characteristics) and the space-time coupling (the correlation coefficient of wind and photovoltaic power output can reach 0.6-0.8) lead to the intensification of system net load fluctuation, bringing unprecedented challenges to the safe and stable operation and economic scheduling of power systems.

[0003] To address this challenge, domestic and foreign scholars have conducted a large number of researches. The existing methods mainly focus on the following aspects: first, probabilistic modeling methods (such as assuming that wind power output follows Weibull distribution and photovoltaic power output follows Beta distribution) are used to describe the uncertainty of wind and photovoltaic power output, and scene sets are generated based on Monte Carlo simulation. However, this kind of method often ignores the space-time coupling characteristics between wind and light output, and the corresponding deterministic scheduling model cannot effectively cope with the volatility of wind and light, leading to frequent start-stop of generating units. According to the CPS data, the start-stop frequency of thermal power generating units increased by 27% in 2022, and the start-stop frequency of thermal power generating units remained high in 2025. Second, various double-layer optimization architectures are proposed to try to coordinate multiple optimization objectives through hierarchical decision-making. However, the coupling mechanism of upper and lower layers in such models is usually weak, and is limited to static parameter transmission, making it difficult to achieve dynamic balance between economy and stability. Third, dynamic robust optimization algorithms are applied to cope with wind and light fluctuations, but this method often faces the problem of high computational complexity, and fails to fully exploit the collaborative optimization potential of flexible resources such as energy storage and renewable energy.

[0004] The above researches have important significance in solving the uncertainty and consumption of wind and photovoltaic power, but there are still the following problems to be solved:

[0005] 1. The existing researches mostly use linear correlation coefficient or a single Copula function (such as Gaussian Copula) to describe the correlation of wind and light output, which is difficult to accurately describe the real nonlinear and asymmetric dependence relationship (for example, the "anti-peak regulation" characteristic of photovoltaic power), and the description error of Gaussian Copula for nonlinear correlation is as high as 15%-20%. This leads to a large deviation between the generated joint output scene and the actual situation, reducing the robustness of the subsequent optimal scheduling scheme.

[0006] 2. Key parameters in existing scheduling models, such as the unit ramp rate, are usually set to fixed values. This rigid constraint cannot adapt to the dynamic changes in net load caused by wind and solar fluctuations: during periods of high fluctuation, the fixed ramp capacity may be insufficient, easily leading to wind and solar curtailment or load shedding; during periods of low fluctuation, the overly conservative constraints may limit the unit's adjustment flexibility, resulting in a waste of adjustment resources and increasing system operating costs. Summary of the Invention

[0007] The purpose of this invention is to provide a stochastic dispatch method for power systems based on the combined probability distribution of wind and solar power and two-layer dynamic optimization, in order to solve the problems of severe net load fluctuations and dispatch difficulties under high-proportion renewable energy integration. This method accurately characterizes the uncertainty and correlation of wind and solar power output, and through the dynamic coupling and feedback mechanism of the upper and lower layer models, achieves synergistic optimization of system operation economy and stability, effectively reducing net load fluctuations and improving the absorption rate of wind and solar power.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] A stochastic dispatch method for power systems based on wind-solar joint probability distribution and two-level dynamic optimization is characterized by the following steps:

[0010] Step 1: Based on the predicted power data of wind power and photovoltaic power, the probability density function of wind power and photovoltaic power output is established by using the non-parametric kernel density estimation (KDE) method, which avoids prior assumptions about the power output distribution and captures its random characteristics more accurately.

[0011] Step 2: Construct a joint probability distribution model of wind and solar power output using the FrankCopula function to accurately describe the correlation between the two.

[0012] Step 3: Sample the copula function, generate a large number of wind and solar power combined output scenarios based on the Monte Carlo simulation method, and use the k-means++ clustering algorithm to reduce the number of scenarios, remove similar scenarios, and finally obtain a set of typical scenarios;

[0013] Step 4: Construct a two-layer stochastic optimization scheduling upper-layer model with the objective function of minimizing net load variance. This layer coordinates the output schemes of wind power, photovoltaic power and energy storage, and effectively smooths net load fluctuations by optimizing the coordinated operation of renewable energy and energy storage.

[0014] The objective function of the upper-level model is as follows:

[0015]

[0016] F1 is the net load variance of the receiving end power grid; T is the total number of time periods in a dispatch cycle; P net,t is the net load of the receiving end power grid in the t time period; P net,ave is the average value of the net load of the receiving end power grid in the entire dispatch cycle.

[0017] Step 5: A lower layer model of a two-layer stochastic optimization scheduling is constructed, which takes the minimization of the total expected operation cost of the system as the objective function, and optimizes the output plans of thermal power and hydropower units on the basis of the optimization results of the upper layer model, and introduces a dynamic ramping constraint mechanism based on the net load fluctuation, so as to establish a dynamic coupling relationship between the two-layer models.

[0018] The objective function of the lower layer model is as follows:

[0019]

[0020] In the formula, F2 is the total expected operation cost of the system, S is the total number of typical scenarios, p s is the probability of scenario s, C 1s is the operation cost of the thermal power unit in scenario s, C 2s is the penalty cost of abandoned wind and light in scenario s, C 3s is the system spinning reserve cost in scenario s, C 4s is the ramping cost of the traditional unit in the adjacent time period in scenario s, and C5 is the environmental benefit of the wind, light and water grid-connected consumption and the energy storage system in scenario s.

[0021] Step 6: The standard deviation of the net load fluctuation output by the upper layer model after optimization is mapped to the dynamic constraint condition of the unit ramping rate of the lower layer model. The core idea of the dynamic constraint mechanism is that when the standard deviation of the net load fluctuation exceeds a preset threshold, the upper limit of the ramping rate constraint of the unit is appropriately relaxed, the output of the unit is allowed to be adjusted more quickly to cope with the fluctuation, and the loss of load is avoided; when the standard deviation of the net load fluctuation is lower than the preset threshold, the upper limit of the ramping rate constraint of the unit is correspondingly tightened, unnecessary frequent adjustment of the unit is suppressed, and the operation wear and tear and the ramping cost are reduced. The two-layer optimization problem is iteratively solved to form the final optimal scheduling scheme of the power system. The dynamic ramping constraint mechanism is as follows:

[0022] The upper and lower limit adjustment rules of the dynamic ramping rate are as follows:

[0023]

[0024]

[0025] In the formula, UR i and DR iare the original up and down ramp rates of thermal power unit i, respectively; σ is the standard deviation of net load fluctuation; σ0 is the preset threshold of the standard deviation of net load fluctuation; γ and κ are adjustment coefficients (γ≥0, κ≥0) that control the relaxation and tightening amplitudes, respectively; and are the adjusted dynamic ramp rate limits.

[0026] The unit ramp rate constraint condition is as follows:

[0027]

[0028] In the formula: is the start-stop state variable of thermal power unit i at time period t, indicates that unit i is in the start state, indicates that unit i is in the stop state. and are the actual power generation of thermal power unit i at time periods t and t-1, respectively.

[0029] The boundary protection condition is as follows:

[0030]

[0031] In the formula: UR i,min / DR i,min and UR i,max / DR i,max are the technical upper and lower limits of the unit ramp rate, respectively.

[0032] Specifically, in step 1, the kernel density estimation function is as follows:

[0033]

[0034] In the formula: is the estimated value of the probability density function at the calculation point x; n is the sample quantity; h is the bandwidth; K(·) is the kernel function.

[0035] Specifically, in step 2, the Frank Copula joint distribution function is expressed as:

[0036]

[0037] In the formula: θ is a parameter describing the correlation of wind and light output; μ and ν are the cumulative distribution function (CDF) values of wind power and photovoltaic output, respectively.

[0038] Specifically, in step 4, the calculation formula of the net load and its average value is as follows:

[0039]

[0040] In the formula, PD,t P0is the initial load of the receiving end power grid at time t; P w,t Pfis the output power of the wind farm at time t; λ w,t λfrepresents the wind curtailment rate at time t, 0≤λ w,t ≤1; P pv,t Pv is the output power of the photovoltaic power station at time t; λ pv,t λvrepresents the light curtailment rate at time t, 0≤λ pv,t ≤1; P c,t Psis the charging power of the energy storage system at time t; η c ηsis the charging efficiency of the energy storage system; P d,t Pdis the discharging power of the energy storage system at time t; η d ηdis the discharging efficiency of the energy storage system.

[0041] Further, the upper model in step 4 further adds system constraint conditions, which are respectively system wind power and photovoltaic output constraints, wind and light curtailment rate constraints, energy storage system charging and discharging power constraints, and energy storage system charge proportion constraints.

[0042] The wind power and photovoltaic output constraints are as follows:

[0043]

[0044] In the formula, P w,max Pfmaxis the maximum output of the wind farm at time t; P pv,max Pvmaxis the maximum output of the photovoltaic power station at time t;

[0045] The wind and light curtailment rate constraints are as follows:

[0046]

[0047] In the formula: k w,max kffis the maximum wind curtailment rate allowed by the wind farm; k pv,max kvis the maximum light curtailment rate allowed by the photovoltaic power station;

[0048] The energy storage system charging and discharging power constraints are as follows:

[0049]

[0050] In the formula: P c,max Psis the maximum charging power of the energy storage system; P d,max Pdis the maximum discharging power of the energy storage system;

[0051] The energy storage system charge proportion constraint is as follows:

[0052]

[0053] In the formula: Q t Qsis the remaining electric quantity of the energy storage system at time t; Q smaxQ represents the maximum amount of electricity that the energy storage system can store; soc,t Q represents the charge ratio of the energy storage system during time period t; soc,min Q represents the lower limit of the charge ratio for energy storage systems. soc,max This represents the upper limit of the charge ratio for energy storage systems.

[0054] Furthermore, in step 5, system constraints are added to the lower-level model, namely system power balance constraints, thermal power unit output constraints, hydropower output constraints, and system spinning reserve constraints, wherein:

[0055] The system power balance constraints are as follows:

[0056]

[0057] Where: N G This represents the total number of thermal power units. P represents the output of thermal power unit i during time period t in scenario s; h,t The hydropower grid-connected power during time period t under scenario s; For the grid-connected wind power during time period t under scenario s; P represents the grid-connected photovoltaic power during time period t under scenario s; d,t,s P c,t,s These represent the energy storage discharge power and charging power during time period t in scenario s, respectively.

[0058] The output constraints of thermal power units are as follows:

[0059]

[0060] In the formula: and These are the lower and upper limits of the output of thermal power unit i, respectively.

[0061] The constraints of the hydropower unit are as follows:

[0062] 0≤P h,t,s ≤P h,max

[0063] In the formula: P h,max This is the upper limit of the output of the hydropower unit.

[0064] The system spin-off standby constraints are as follows:

[0065]

[0066] Where: μ w1 μ w,2 These are the positive and negative spinning reserve capacity coefficients, respectively, to address wind power forecasting errors; μ pv1 μ pv2 These are the positive and negative rotating reserve capacity coefficients, respectively, to address photovoltaic forecasting errors.

[0067] The main design idea of the present application is as follows:

[0068] 1. Non-parametric kernel density estimation is used to model wind power and photovoltaic power output respectively, avoiding the limitations of assuming a specific distribution and better fitting the statistical characteristics of actual data.

[0069] 2. The nonlinear and asymmetric spatiotemporal correlation between wind and light output is captured based on the FrankCopula function, and a joint probability distribution model of wind and light output is constructed.

[0070] 3. The wind and light joint output scenarios are generated through Monte Carlo simulation, and the k-means++ clustering algorithm is used for scenario reduction to obtain a representative set of typical scenarios, which maximizes the statistical characteristics of the original scenario set while significantly reducing the complexity of the optimization model, achieving a balance between computational efficiency and accuracy.

[0071] 4. A double-layer optimization scheduling model is constructed, in which the upper model takes system operation stability as the primary goal, coordinates the output of wind, light, and energy storage, actively manages the net load waveform, and minimizes its fluctuations to create a "friendly" load environment for the lower economic dispatch. The lower model takes system economy as the primary goal, considers the operation cost, standby cost, and environmental benefits of thermal power and hydropower on the basis of the upper optimization results, minimizes the expected cost of all typical scenarios, and develops a robust and economical unit commitment and scheduling plan.

[0072] The net load fluctuation standard deviation output by the upper model is used as a feedback signal to dynamically adjust the ramp rate constraint of the units in the lower model. This breaks the rigid pattern of fixed ramp rate constraints in traditional scheduling, realizes intelligent and adaptive adjustment of "fluctuation large, relax constraint to ensure safety, fluctuation small, tighten constraint to reduce cost", and is the key to improving system stability and economy.

[0073] Therefore, compared with the prior art, the present application has the following beneficial effects:

[0074] (1) In view of the problem that the existing method is difficult to accurately depict the nonlinear correlation of wind and light, the present application constructs a joint probability distribution model through the frankcopula function and generates a representative set of typical scenarios, so that the scheduling strategy can cover various fluctuation modes, significantly improving the robustness of the decision scheme.

[0075] (2) In view of the problem that high fluctuation periods are prone to cause wind and light curtailment, the present application optimizes wind, light, and storage in the upper model and considers wind and light curtailment constraints, which can suppress fluctuations while maximizing the utilization rate of renewable energy, effectively reducing the phenomenon of wind and light curtailment.

[0076] (3) In order to solve the problem of waste of adjustment resources caused by fixed climbing constraints, the application introduces a dynamic climbing constraint mechanism, and adaptively adjusts the upper limit of the climbing rate of the unit according to the net load fluctuation level, effectively reduces the operation cost and equipment wear of the system under the premise of ensuring the safety of the system. BRIEF DESCRIPTION OF DRAWINGS

[0077] Figure 1 The flowchart of the application is shown. DETAILED DESCRIPTION

[0078] The application discloses a power system stochastic scheduling method based on wind-solar joint probability distribution and double-layer dynamic optimization. The scheduling method first adopts non-parametric kernel density estimation to respectively establish probability density functions of wind power and photovoltaic output, and constructs a joint probability distribution model of wind-solar output through a Frank Copula function to accurately describe the correlation; a large number of wind-solar joint output scenarios are generated through Monte Carlo simulation, and scenario reduction is performed through k-means++ clustering to obtain a typical scenario set. On this basis, a double-layer optimization model is constructed: the upper layer takes minimization of net load variance as an objective, coordinates the output of wind power, photovoltaic and energy storage to suppress fluctuation; the lower layer takes minimization of total expected operation cost of the system as an objective, optimizes the output plan of thermal power and hydropower units, and introduces a dynamic climbing constraint mechanism based on the standard deviation of net load fluctuation to adaptively adjust the upper and lower limits of the climbing rate constraint of the unit according to the fluctuation level, and realize the collaborative optimization of economy and stability. The method effectively solves the problem of system scheduling difficulty under high proportion of renewable energy access, and improves the renewable energy consumption capacity and system operation efficiency.

[0079] As shown in Figure 1 , the implementation process of the application is as follows:

[0080] Step 1: Based on the predicted power data of wind power and photovoltaic, the non-parametric kernel density estimation method (Kernel Density Estimation, KDE) is used to respectively establish the probability density functions of wind power and photovoltaic output, avoiding the prior assumption of the form of output distribution, and more accurately capturing the random characteristics thereof;

[0081] Kernel density estimation function:

[0082]

[0083] In the formula: is the probability density function estimation value at the calculation point x; n is the sample number; h is the bandwidth; K(·) is the kernel function.

[0084] Step 2: A Frank Copula function is used to construct a joint probability distribution model of wind-solar output to accurately describe the correlation between the two;

[0085] Frank Copula joint distribution function:

[0086]

[0087] where θ is the parameter describing the correlation between wind and solar power; μ and v are the cumulative distribution function (CDF) values of wind and solar power, respectively.

[0088] Step 3: Sample the copula function, generate a large number of wind-solar joint output scenarios based on the Monte Carlo simulation method, and use the k-means++ clustering algorithm to reduce the scenarios, eliminate similar scenarios, and finally obtain a set of typical scenario set;

[0089] Step 4: Build a two-level stochastic optimization scheduling model for the power system:

[0090] 4.1, the upper model takes the minimum variance of net load as the objective function, coordinates the output scheme of wind power, photovoltaic and energy storage, optimizes the coordinated operation of renewable energy and energy storage, and effectively suppresses the fluctuation of net load. The model is as follows:

[0091]

[0092] where F1 is the variance of the receiving end power grid net load; T is the total number of time periods in a scheduling period; P net,t is the net load of the receiving end power grid at time t; P net,ave is the average value of the net load of the receiving end power grid in the entire scheduling period.

[0093] The calculation formula of net load and its average value

[0094]

[0095] where P D,t is the initial load of the receiving end power grid at time t; P w,t is the output power of the wind farm at time t; λ w,t represents the wind curtailment rate at time t, 0≤λ w,t ≤1; P pv,t is the output power of the photovoltaic power station at time t; λ pv,t represents the light curtailment rate at time t, 0≤λ pv,t ≤1; P c,t is the charging power of the energy storage system at time t; η c is the charging efficiency of the energy storage system; P d,t is the discharging power of the energy storage system at time t; η d is the discharging efficiency of the energy storage system.

[0096] System constraint conditions:

[0097] (1) Wind and solar power output constraints

[0098]

[0099] Pmax,t w,max is the maximum output of the wind farm in the t period; P pv,max is the maximum output of the photovoltaic power station in the t period;

[0100] wind and light abandonment rate constraint

[0101]

[0102] k w,max is the maximum wind abandonment rate allowed by the wind farm; k pv,max is the maximum light abandonment rate allowed by the photovoltaic power station;

[0103] energy storage system charging and discharging power constraint

[0104]

[0105] Pmax,charge c,max is the maximum charging power of the energy storage system; P d,max is the maximum discharging power of the energy storage system;

[0106] energy storage system state of charge constraint

[0107]

[0108] Q t is the remaining power of the energy storage system in the t period; Q smax is the maximum storage power of the energy storage system; Q soc,t is the state of charge of the energy storage system in the t period; Q soc,min is the lower limit of the state of charge of the energy storage system; Q soc,max is the upper limit of the state of charge of the energy storage system.

[0109] 4.2, the lower layer model takes the minimum total expected operation cost of the system as the objective function, and optimizes the output plan of thermal power and hydropower units based on the optimization results of the upper layer model, and introduces a dynamic climbing constraint mechanism based on net load fluctuation, so as to establish a dynamic coupling relationship between the two-layer models. The model is as follows:

[0110]

[0111] F2 is the total expected operation cost of the system, S is the total number of typical scenarios, p s is the probability of scenario s, C 1s is the operation cost of thermal power unit in scenario s, C 2s is the wind and light abandonment penalty cost in scenario s, C 3sTo account for the system's rotational backup cost in scenario s, C 4s C5 represents the ramp-up cost of traditional generator units in adjacent time periods under scenario s, while C5 represents the environmental benefits of wind, solar, and hydropower grid connection and energy storage systems under scenario s.

[0112] System constraints:

[0113] (1) System power balance constraints

[0114]

[0115] Where: N G This represents the total number of thermal power units. P represents the output of thermal power unit i during time period t in scenario s; h,t The hydropower grid-connected power during time period t under scenario s; For the grid-connected wind power during time period t under scenario s; P represents the grid-connected photovoltaic power during time period t under scenario s; d,t,s P c,t,s These represent the energy storage discharge power and charging power during time period t in scenario s, respectively.

[0116] Thermal power unit output constraints

[0117]

[0118] In the formula: and These are the lower and upper limits of the output of thermal power unit i, respectively.

[0119] Hydropower unit constraints

[0120] 0≤P h,t,s ≤P h,max

[0121] In the formula: P h,max This is the upper limit of the output of the hydropower unit.

[0122] System rotational spare constraints

[0123]

[0124] Where: μ w1 μ w,2 These are the positive and negative spinning reserve capacity coefficients, respectively, to address wind power forecasting errors; μ pv1 μ pv2 These are the positive and negative rotating reserve capacity coefficients, respectively, to address photovoltaic forecasting errors.

[0125] Step 5: Map the net load fluctuation standard deviation output by the upper model optimization to the dynamic constraint condition of the unit ramp rate of the lower model. The core idea of the dynamic constraint mechanism is: when the net load fluctuation standard deviation exceeds the preset threshold, appropriately relax the upper limit of the unit ramp rate constraint to allow the unit to adjust the output more quickly to respond to fluctuations and avoid loss of load; when the net load fluctuation standard deviation is below the preset threshold, the upper limit of the unit ramp rate constraint is tightened accordingly to suppress unnecessary frequent adjustment of the unit and reduce operating wear and tear and ramping cost. Iterative solution of the bi-level optimization problem forms the final optimal power system scheduling scheme. The dynamic ramping constraint mechanism is as follows:

[0126] (1) Dynamic ramping rate upper and lower limit adjustment rule

[0127]

[0128]

[0129] wherein: UR i and DR i are the original upper and lower ramping rates of the thermal power unit i; σ is the net load fluctuation standard deviation; σ0 is the preset threshold of the net load fluctuation standard deviation; γ and κ are adjustment coefficients (γ≥0, κ≥0) for controlling the amplitude of ramping rate constraint relaxation and tightening, respectively; and are the adjusted dynamic ramping rate limits.

[0130] (2) Unit ramping rate constraint condition

[0131]

[0132] wherein: is the start-stop state variable of the thermal power unit i at time period t, indicates that the unit i is in the start state, indicates that the unit i is in the stop state; and are the actual power generation of the thermal power unit i at time periods t and t-1, respectively.

[0133] Boundary protection condition

[0134]

[0135] wherein: UR i,min / DR i,min and UR i,max / DR i,max are the technical upper and lower limits of the unit ramping rate, respectively.

Claims

1. A power system stochastic scheduling method based on wind-solar joint probability distribution and double-layer dynamic optimization, characterized in that, The method comprises the following steps: Step 1: Based on the predicted power data of wind power and photovoltaic power, a non-parametric kernel density estimation method is used to construct the probability density functions of wind power and photovoltaic power output respectively; Step 2: A Frank Copula function is used to establish a joint probability distribution model of wind and light output to accurately describe the correlation between them; Step 3: The Copula function is sampled, and a large number of wind and light joint output scenarios are generated based on Monte Carlo simulation, and the k-means++ clustering algorithm is used for scenario reduction to eliminate similar scenarios, and finally a set of representative typical scenario set is obtained; Step 4: A double-layer stochastic optimization scheduling upper model is constructed with the minimum variance of net load as the objective function, which coordinates the output scheme of wind power, photovoltaic power and energy storage, optimizes the coordinated operation of renewable energy and energy storage, and effectively suppresses the net load fluctuation; the scheduling model is as follows: In the formula, F1 is the net load variance of the receiving end power grid, T is the number of time periods in a dispatch cycle, P net,t is the net load of the receiving end power grid in the t time period, P net,ave is the average value of the net load of the receiving end power grid in the entire dispatch cycle; Step 5: A double-layer stochastic optimization scheduling lower model is constructed with the minimum total expected operation cost of the system as the objective function, which optimizes the output plan of thermal power and hydropower units based on the optimization results of the upper model, and introduces a dynamic ramping constraint mechanism based on net load fluctuation to establish a dynamic coupling relationship between the two models; the scheduling model is as follows: where F2 is the total expected operation cost of the system, S is the total number of typical scenarios, p s is the probability of scenario s occurring, C 1s is the operation cost of thermal power units under scenario s, C 2s is the penalty cost of curtailment of wind and solar power under scenario s, C 3s is the system spinning reserve cost under scenario s, C 4s is the ramping cost of thermal power units and hydro power units in adjacent time periods under scenario s, and C5 is the environmental benefit of wind-solar-hydro grid integration and energy storage systems under scenario s. Step 6: The standard deviation of net load fluctuation output by the upper model after optimization is mapped to the dynamic constraint condition of unit ramping rate of the lower model, and the core idea of the dynamic constraint mechanism is: when the standard deviation of net load fluctuation exceeds the preset threshold, the upper limit of the ramping rate constraint of the unit is appropriately relaxed to allow the unit to adjust the output more quickly to respond to fluctuations and avoid loss of load; when the standard deviation of net load fluctuation is lower than the preset threshold, the upper limit of the ramping rate constraint of the unit is tightened accordingly to inhibit unnecessary frequent adjustment of the unit, reduce operation wear and tear and ramping cost, and iteratively solve the double-layer optimization problem to form the final optimal scheduling scheme of the power system; the constraint condition is as follows: The dynamic ramping rate upper and lower limit adjustment rule is as follows: UR i and DR i are the original up and down ramping rates of the thermal power unit i, σ is the standard deviation of the net load fluctuation, σ0 is the preset threshold of the standard deviation of the net load fluctuation, γ and κ are adjustment coefficients (γ≥0, κ≥0) that control the relaxation and tightening amplitudes, respectively, and are the adjusted dynamic ramping rate limits; The unit ramping rate constraint condition is as follows: In the formula: is the start-stop state variable of the thermal power generating unit i at the time period t, represents that the unit i is in the start-up state, represents that the unit i is in the shut-down state, and are the actual power generation of the thermal power generating unit i at the time periods t and t-1, respectively. The boundary protection condition is as follows: where UR i,min / DR i,min and UR i,max / DR i,max are the technical upper and lower limits of the unit ramp rate, respectively. 2.The power system stochastic dispatching method based on wind-solar combined probability distribution and double-layer dynamic optimization according to claim 1, characterized in that, In the step 1, the kernel density estimation function is as follows: In the formula: For the estimation of the probability density function at point x, n is the sample number, h is the bandwidth, and K(·) is the kernel function. 3.The power system stochastic dispatching method based on wind-solar combined probability distribution and double-layer dynamic optimization according to claim 1, characterized in that, In the step 2, the Frank Copula joint distribution function is expressed as: In the formula, θ is a parameter describing the correlation between wind and light output, μ and ν are the cumulative distribution function values of wind and photovoltaic output respectively. 4.The power system stochastic dispatching method based on wind-solar combined probability distribution and double-layer dynamic optimization according to claim 1, characterized in that, In the step 4, the calculation formula of net load and its average value is as follows: P net,t = P D,t -(1-λ w,t )P w,t -(1-λ w,t )P pv,t +η c P c,t -η d P d,t where P D,t is the initial load of the receiving power grid at time t, P w,t is the output power of the wind farm at time t, λ w,t represents the wind curtailment rate at time t, 0≤λ w,t ≤1; P pv,t is the output power of the photovoltaic power station at time t, λ pv,t represents the light curtailment rate at time t, 0≤λ pv,t ≤1, P c,t is the charging power of the energy storage system at time t, η c is the charging efficiency of the energy storage system, P d,t is the discharging power of the energy storage system at time t, η d is the discharging efficiency of the energy storage system.

5. The power system stochastic dispatching method based on wind-solar combined probability distribution and double-layer dynamic optimization according to claim 1, characterized in that, In the step 4, the upper model also adds system constraint conditions, which are system wind and photovoltaic output constraints, wind and light rejection rate constraints, energy storage system charge and discharge power constraints and energy storage system charge proportion constraints, wherein: The wind and photovoltaic output constraints are as follows: 0 < P w,t ≤ P w,max 0 < P pv,t ≤ P pv,max In the formula, P w,max is the maximum output of the wind farm in the t period, P pv,max is the maximum output of the photovoltaic power station in the t period; The wind and light rejection rate constraints are as follows: In the formula: k w,max is the maximum wind curtailment rate allowed for the wind farm, k pv,max is the maximum light curtailment rate allowed for the photovoltaic power station; The energy storage system charge and discharge power constraints are as follows: 0 < P c,t ≤ P c,max 0 < P d,t ≤ P d,max P c,t P d,t =0 where: P c,max Pmax is the maximum charge power of the energy storage system, d,max Pmin is the maximum discharge power of the energy storage system. The energy storage system charge proportion constraints are as follows: Q soc,min ≤Q soc,t ≤Q soc,max wherein: Q t Q is the remaining energy of the energy storage system for the time period t smax Q is the maximum storage energy of the energy storage system soc,t Q is the state of charge of the energy storage system for the time period t soc,min Q is the lower limit of the state of charge of the energy storage system soc,max Q is the upper limit of the state of charge of the energy storage system 6.The method of claim 1, wherein, In the step 5, the lower model also adds system constraint conditions, which are system power balance constraints, thermal power unit output constraints, hydropower output constraints and system spinning reserve constraints, wherein: The system power balance constraints are as follows: In the formula: N G is the total number of thermal power units, is the output of thermal power unit i in time period t under scenario s, P h,t is the grid-connected power of hydropower in time period t under scenario s, is the grid-connected power of wind power in time period t under scenario s, is the grid-connected power of photovoltaic power in time period t under scenario s, P d,t,s , P c,t,s are the discharging power and charging power of energy storage in time period t under scenario s, respectively; The thermal power unit output constraints are as follows: In the formulae: and are the lower and upper limits of the output of the thermal power unit i, respectively. The hydropower unit constraints are as follows: 0 < P h,t,s ≤ P h,max In the formula: P h,max is the upper limit of the output of the hydroelectric generating set.

7. The system spinning reserve constraints are as follows: wherein: μ w1 , μ w,2 are positive and negative spinning reserve capacity coefficients for wind power prediction errors, respectively, μ pv1 , μ pv2 are positive and negative spinning reserve capacity coefficients for photovoltaic prediction errors, respectively.

Images