Power capacity evaluation method based on dynamic confidence interval and double-layer optimization
By creating a source-load joint scenario model and double-layer robust optimization, the unilateral uncertainty and static evaluation problems in the capacity abundance assessment of power system are solved, and the accuracy of the evaluation results is improved and the balance between market economy and safety is achieved, and reliable decision-making support is provided.
Patent Information
- Application Number
- CN202510371214.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-27
- Publication Date
- 2025-07-04
AI Technical Summary
The existing power system capacity abundance evaluation technology has problems such as insufficient unilateral uncertainty modeling, static evaluation and source-charge coupling modeling, resulting in biased and unreliable evaluation results.
Using a method based on dynamic confidence intervals and double-layer optimization, a source-load joint scenario model is created, and a wind power output and demand response model is integrated by improving the conditional generation adversarial network and dual discriminator structure, combined with a double-layer robust optimization model, the risk probability index and capacity shortage expectations are optimized and evaluated.
It significantly improves the coverage and accuracy of the evaluation results in the time and space dimensions, realizes the balance between economics and operational safety of the power market, and provides a scientific decision-making basis for the stable operation and optimization scheduling of the power system.
Smart Images

Figure CN120257820A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power system dispatching automation, and particularly to a power capacity evaluation method based on dynamic confidence intervals and bi-level optimization focusing on the capacity adequacy evaluation link of the power spot market, aiming to construct an evaluation method that comprehensively considers the uncertainty factors on both the power supply side (such as renewable energy sources like wind power and photovoltaic power) and the load side (demand response, flexible loads, etc.), and provides a solid technical support for the risk quantification analysis and safe and stable operation decision-making of cross-regional power transactions. Background Art
[0002] Under the background of the global energy transition, the proportion of renewable energy in the power system continues to rise, which makes the impact of the uncertainty on both the power supply side and the load side (i.e., the power supply side - load side) on the capacity adequacy (CA) of the power spot market become more prominent. However, there are many limitations in the existing evaluation technologies:
[0003] (1) Limitations in unilateral uncertainty modeling: Most traditional evaluation methods only focus on the uncertainty of the power supply side. For example, the inherent volatility of wind power output is considered, but the randomness characteristics presented by demand response on the load side are ignored. This one-sided modeling method cannot comprehensively reflect the complex uncertainty factors faced in the actual operation of the power system, resulting in biased evaluation results.
[0004] (2) Deficiencies of static evaluation models: Most relevant studies only use deterministic scenarios or a single confidence interval setting when conducting capacity adequacy evaluations. This static evaluation method is difficult to dynamically track the risks brought by real-time market changes and cannot provide timely and accurate basis for the real-time dispatching and decision-making of the power system.
[0005] (3) Defects in power supply - load coupling modeling: Existing power supply - load joint scenario generation methods often rely on historical data binning or simple Monte Carlo sampling techniques. These methods have obvious deficiencies in capturing the complex spatio-temporal correlations between the power supply and the load, and cannot accurately describe the interaction relationship between the power supply and the load in the power system at different time and space scales, thus affecting the reliability of the evaluation results. Summary of the Invention
[0006] The object of the present invention is to solve the above problems existing in the prior art, and to provide a power capacity evaluation method based on dynamic confidence intervals and double-layer optimization. A source-load joint scenario model is created, which is used to evaluate both the power supply side and the load side, and can significantly improve the coverage and accuracy of the evaluation results in the time and space dimensions. At the same time, with the help of the dynamic confidence interval adjustment mechanism and the double-layer optimization strategy, an effective balance between the economic efficiency and operation safety of the power market is achieved, providing a scientific and reliable decision-making basis for the stable operation and optimal dispatching of the power system.
[0007] The above technical object of the present invention is mainly solved by the following technical solutions:
[0008] A power capacity evaluation method based on dynamic confidence intervals and double-layer optimization, characterized in that:
[0009] Based on the historical data and predicted values of wind power output, the prediction error of wind power output is obtained;
[0010] Based on the prediction error, the influencing parameters that cause the error are determined;
[0011] Based on the prediction error and the influencing parameters, a power supply side model is created;
[0012] Based on the historical load data, demand response execution records, and external influencing factors, the demand response sensitivity coefficient and time-varying adjustment parameters are obtained;
[0013] Based on the demand response sensitivity coefficient and the time-varying adjustment parameters, a load side model is created;
[0014] Based on the power supply side model and the load side model, a source-load joint scenario model is created;
[0015] The source-load joint scenario model is optimized through a double-layer robust optimization model;
[0016] Based on the optimized source-load joint scenario, the evaluation risk probability index and the expected capacity shortage are determined.
[0017] Creating a source-load joint scenario model, which is used to evaluate both the power supply side and the load side, can significantly improve the coverage and accuracy of the evaluation results in the time and space dimensions. At the same time, with the help of the dynamic confidence interval adjustment mechanism and the double-layer optimization strategy, an effective balance between the economic efficiency and operation safety of the power market is achieved, providing a scientific and reliable decision-making basis for the stable operation and optimal dispatching of the power system.
[0018] As a further improvement and supplement to the above technical solution, the present invention adopts the following technical measures:
[0019] Preferably, when creating the power supply side model,
[0020] First, collect historical data and forecast values of wind power output;
[0021] Clean historical data and forecast values, remove outliers, and perform normalization;
[0022] By ΔP w =P w,实际 -P w,预测 To calculate the prediction error,
[0023] Among them, P w,实际 is the historical data of power output, P w,预测 is the predicted value of wind power output, ΔP w is the prediction error of wind power output;
[0024] Determine the probability density function of the power supply side as
[0025]
[0026] Where K represents the number of sub-distributions in the mixed distribution; α k is the mixing weight of the kth sub-distribution; and satisfies That is, the sum of the weights of all sub-distributions is 1, and each weight is positive; represents the kth sub-distribution, μ k and are the mean and variance of the kth sub-distribution respectively;
[0027] Create a power-side model based on the power-side probability density function.
[0028] Preferably, α is determined by the EM algorithm. k , μ k and σ k The value of
[0029] By alternately executing the expectation step and the maximization step, the optimal parameter value is gradually approached;
[0030] In the expectation step, the posterior probability of each data point belonging to each subdistribution is calculated based on the current parameter estimates;
[0031] In the maximization step, the posterior probability is used to update the parameter value, and this process is repeated for multiple iterations until the parameters converge to obtain the parameters that best fit the historical data.
[0032] Preferably, when creating a load side model,
[0033] Based on the demand response sensitivity coefficient and time-varying adjustment parameters, the load side probability density function is determined as follows:
[0034]
[0035] Among them, ξ is the demand response sensitivity coefficient, γ and β are time-varying adjustment parameters, γ is used to adjust the mean shift of the load deviation, and β is used to adjust the fluctuation range of the load deviation;
[0036] Based on the probability density function of the load side, a load side model is created.
[0037] Preferably, when creating the source-load joint scenario model,
[0038] an improved conditional generative adversarial network is adopted. The improved conditional generative adversarial network includes a generator and a discriminator, and the discriminator includes a first discriminator and a second discriminator;
[0039] The first discriminator constrains the spatio-temporal correlation of the power output in the source-load joint scenario model, and is used to make the generated power scenario approximate the real power scenario;
[0040] The second discriminator constrains the spatio-temporal correlation of the load demand in the source-load joint scenario model, and is used to make the generated power scenario approximate the real power scenario;
[0041] The generator generates a joint scenario set:
[0042]
[0043] Among them, represents the power output under scenario s, represents the load demand under scenario s, and N is the total number of generated scenarios.
[0044] Preferably, a dynamic confidence interval is set in the source-load joint scenario model. The formula adopted by the dynamic confidence interval is:
[0045]
[0046] The confidence level α is adjusted by the risk preference parameter ρ,
[0047] where ρ is the risk preference parameter, α is the confidence level, and ρ hist is the historical risk preference data set, and max(ρ hist ) is the maximum value in the historical risk preference data set.
[0048] Preferably, the two-layer robust optimization model is constructed. The two-layer robust optimization model includes an upper layer model and a lower layer model,
[0049] The upper layer model is modeled based on the upper layer objective function. The upper layer objective function is:
[0050]
[0051] Among them, T represents the optimized time interval, and C g is the power generation cost coefficient of the conventional unit, and P g,t is the wind power output of the conventional wind turbine at time t; C w is the accommodation cost coefficient of the wind power output, and P w,t is the wind power output at time t.
[0052] Preferably, the constraint conditions that the upper-layer model needs to satisfy are as follows:
[0053] Restrict the wind power output of the wind turbine: For each time t, the wind power output of the conventional unit needs to be within the allowable upper and lower limits corresponding to each conventional wind turbine, that is, P g,min ≤P g,t ≤P g,max , where P g,t is the wind power output at time t, p g,min is the lower limit of the wind power output at time t; p g,max is the upper limit of the wind power output at time t;
[0054] Constrain the power balance of the wind turbine: The total power generation of the wind turbines in the wind power system at each time should be equal to the load demand, that is, P g,t +P w,t =L t , where L t is the system load at time t;
[0055] Constrain the accommodation of the wind power output: In the formula, The upper limit of the prediction error of the wind power output.
[0056] Preferably, the lower-layer model is modeled based on the lower-layer objective function, and a dynamic confidence interval is embedded in the lower-layer model. The lower-layer objective function is:
[0057]
[0058] Among them, P l s is the tie-line power flow under scenario s, and P l max is the upper limit of the tie-line capacity, and max(0, P ls -P lmax ) is the overlimit value when the tie-line power flow exceeds the capacity limit. If the tie-line power flow does not exceed the capacity limit, then max(0, P ls -P lmax ) takes the value of 0.
[0059] Preferably, when determining the evaluation risk probability index, the tie-line power flow over-limit probability function is adopted:
[0060]
[0061] where I(·) is the indicator function, N is the total number of scenarios. When the condition in the parentheses holds, the value of the indicator function is 1, otherwise it is 0. That is, for each scenario s, if the tie-line power flow P l s exceeds the capacity limit P l max , then I(P ls >P lmax ) = 1, otherwise it is 0;
[0062] When determining the expected value of capacity shortage, the expected value function of capacity shortage is adopted:
[0063]
[0064] For each scenario s, max(0, P l s -P l max ) represents the part where the tie-line power flow exceeds the capacity limit, and if it does not exceed, it is 0.
[0065] The beneficial effects of the present invention are as follows:
[0066] 1. The improved conditional generative adversarial network C-GAN with a double discriminator structure organically integrates the wind power output asymmetric hybrid distribution model (power source side model) and the demand response deviation model (load side model), and generates a source-load joint scenario model. Through the source-load joint scenario model, the spatio-temporal correlation on the power source side and the load side is significantly improved, and it can more realistically reflect the source-load relationship in the actual operation of the power system, providing more reliable basic data for subsequent evaluation;
[0067] 2. The dynamic confidence interval adaptively adjusted based on historical risk preference data can effectively cover the typical operation scenarios of the power system, avoiding the conservative error that may be caused by the traditional fixed confidence interval in the evaluation process, making the evaluation result more in line with the actual operation situation, and providing more valuable reference information for decision-makers;
[0068] 3. By constructing a two-layer robust optimization model, while reducing the risk of tie-line power flow over-limit, the market clearing cost can be controlled within the historical mean fluctuation range, achieving the Pareto optimality between risk and cost, and providing a strong guarantee for the economic and safe operation of the power system. Description of the Drawings
[0069] Figure 1It is a schematic flow chart of a method of the present invention.
[0070] Figure 2 It is a schematic diagram of the cooperation between the source-load joint scenario model involved in the present invention and the dynamic confidence interval adjustment architecture.
[0071] Figure 3 The invention discloses a flowchart of a double-layer robust optimization method. DETAILED DESCRIPTION
[0072] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.
[0073] Example: Figures 1 - 3 As shown in the figure, a power capacity assessment method based on dynamic confidence interval and double-level optimization is proposed:
[0074] 1. Create a power supply side model:
[0075] During the operation of the power system, the volatility and uncertainty of wind power output pose a major challenge to the stable operation and capacity adequacy assessment of the power system. The power supply side model can quantify these uncertainties and accurately assess the risks of the power system under different operating conditions, so that corresponding measures can be taken in advance to ensure the stable operation of the system. Therefore, the core of the power supply side model is to model the probability distribution of wind power output forecast errors. The specific process is as follows:
[0076] Based on the historical data and predicted value of wind power output, the prediction error of wind power output is obtained;
[0077] Based on the prediction error, determine the influencing parameters that produce the error;
[0078] Based on prediction errors and influencing parameters.
[0079] Specifically, when creating the power supply side model,
[0080] First, collect historical data and forecast values of wind power output;
[0081] Clean historical data and forecast values, remove outliers, and perform normalization;
[0082] By ΔP w =P w,实际 -P w,预测 To calculate the prediction error;
[0083] Among them, P w,实际 is the historical data of power output, P w,预测 is the predicted value of wind power output, ΔP w is the prediction error of wind power output;
[0084] The probability density function of the power supply side is determined as follows:
[0085]
[0086] where K represents the number of sub - distributions in the mixture distribution; α k is the mixing weight of the k - th sub - distribution; and it satisfies That is, the sum of the weights of all sub - distributions is 1, and each weight is a positive value; represents the k - th sub - distribution, μ k and are the mean and variance of the k - th sub - distribution respectively;
[0087] Create a power supply side model based on the power supply side probability density function.
[0088] Moreover, the values of α k , μ k and σ k are determined through the EM algorithm. The specific steps are as follows:
[0089] Gradually approximate the optimal parameter values by alternately executing the expectation step (E - step) and the maximization step (M - step);
[0090] In the E - step, calculate the posterior probability that each data point belongs to each sub - distribution according to the current parameter estimation values;
[0091] In the M - step, update the parameter values using the posterior probability. Repeat this process. After multiple iterations, until the parameters converge, the parameters that best fit the historical data are obtained.
[0092] II. Create a load side model:
[0093] In actual situations, on the load side, the demand response deviation is an important factor affecting the operation of the power system. Demand response is that users adjust their electricity consumption behaviors according to power market price signals or incentive mechanisms, but there are inevitably certain deviations in this adjustment process. Therefore, the core of the load side model is to perform probabilistic modeling on the demand response deviation. The specific process is as follows:
[0094] Based on historical load data, demand response execution records, and external influencing factors, obtain the demand response sensitivity coefficient and the time - varying adjustment parameter;
[0095] Based on the demand response sensitivity coefficient and the time - varying adjustment parameter.
[0096] Specifically, when creating the load side model,
[0097] Based on the demand response sensitivity coefficient and the time - varying adjustment parameter, determine the probability density function of the load side as follows:
[0098]
[0099] Among them, ξ is the demand response sensitivity coefficient, which represents the sensitivity of users to price or incentive signals. The relationship between user response behavior and price signals can be fitted by linear regression. γ and β are time-varying adjustment parameters. γ is used to adjust the mean shift of load deviation, and β is used to adjust the fluctuation range of load deviation;
[0100] Based on the probability density function of the load side, a load side model is created.
[0101] In the probability density function of the load side, each parameter can be adjusted according to different factors such as time, season, user type, etc. For example, on holidays, when the sensitivity of users to electricity prices decreases, γ increases; in extreme weather conditions, when the load uncertainty increases, β also increases.
[0102] III. Creating a source-load joint scenario model based on the power source side model and the load side model:
[0103] Optimizing the source-load joint scenario model through a two-layer robust optimization model;
[0104] Determining the evaluation risk probability index and the expected capacity shortage based on the optimized source-load joint scenario.
[0105] In the power system, the uncertainties on the power source side and the load side interact with each other. Considering the uncertainties on the power source side or the load side alone cannot fully reflect this complex interaction. The source-load joint scenario (generally speaking, the source-load joint scenario is equivalent to a database) can combine the uncertainties of both.
[0106] When creating the source-load joint scenario model:
[0107] The traditional conditional generative adversarial network (GAN) consists of a generator and a discriminator. The task of the generator is to generate data similar to the real data distribution, while the discriminator is responsible for distinguishing between the generated data and the real data. In the application scenario involved in this technical solution, in order to effectively constrain the correlation between the power source and load scenarios in the time and space dimensions, an improved conditional generative adversarial network (that is, the improved conditional generative adversarial network (GAN), Conditional Generative Adversarial Network, C-GAN) is adopted. The improved conditional generative adversarial network includes a generator and a discriminator, and the discriminator includes a first discriminator and a second discriminator (that is to say, a dual discriminator structure is adopted);
[0108] The first discriminator constrains the spatio-temporal correlation of the power output in the source-load joint scenario model, and is used to make the generated power scenario approximate the real power scenario;
[0109] The second discriminator constrains the spatio-temporal correlation of the load demand in the source-load joint scenario model, and is used to generate a power scenario that approximates the real power scenario;
[0110] Next, the generator generates a joint scenario set:
[0111]
[0112] Among them, represents the power output under scenario s, represents the load demand under scenario s, and N is the total number of generated scenarios.
[0113] In order to more accurately describe the characteristics of the uncertainty of the source-load (i.e., the power supply side and the load side), and to meet the diverse needs of different decision-makers' risk preferences, a dynamic confidence interval is set in the source-load joint scenario model. The formula adopted by the dynamic confidence interval is:
[0114]
[0115] The confidence level α is adjusted by the risk preference parameter ρ,
[0116] where ρ is the risk preference parameter, α is the confidence level, and ρ hist is the historical risk preference data set, and max(ρ hist ) is the maximum value in the historical risk preference data set.
[0117] In the source-load joint scenario model, the dynamic confidence interval can be used to reflect the change range of the uncertainty on the power supply side and the load side. And with the passage of time and the acquisition of new information, this dynamic confidence interval can be dynamically adjusted to more accurately describe the characteristics of the source-load uncertainty.
[0118] That is to say, in the process of evaluating the adequacy of the electricity spot market capacity, the dynamic confidence interval can help evaluate the risk level of the adequacy of the electricity spot market capacity. A wider confidence interval indicates a greater degree of dispersion of the data, meaning a higher uncertainty of the adequacy of the capacity, so the risk is higher; a narrower confidence interval indicates that the data is relatively concentrated, indicating that the adequacy of the capacity is relatively certain, so the risk is lower. Different decision-makers have different requirements for the evaluation results due to the differences in their own risk preferences.
[0119] Therefore, the dynamic confidence interval can flexibly adjust the confidence interval according to different risk preferences, and then quantify the adequacy of the capacity under different risk thresholds.
[0120] For example, if the decision maker is more conservative, worried about the high risk of capacity adequacy, and hopes to cover various possible situations more comprehensively to reduce the evaluation risk, a wider confidence interval will be selected. Because a wider confidence interval means greater uncertainty in capacity adequacy, but it can more likely avoid evaluation errors caused by missing some extreme situations. On the contrary, if the decision maker is more inclined to take risks, willing to bear the uncertainty risk that may exist in the capacity adequacy evaluation, and expects to make decisions under relatively more definite and concentrated evaluation results, a narrower confidence interval will be chosen.
[0121] IV. Construct a two - layer robust optimization model, which includes an upper - layer model and a lower - layer model;
[0122] Traditional capacity adequacy evaluation often uses a single - layer optimization model, which is difficult to balance the economy (minimizing operating costs) and security (minimizing the risk of tie - line power flow limit) of the electricity spot market. Therefore, in this technical solution, by constructing a two - layer robust optimization model, the upper - layer model is used for market clearing and the lower - layer model is used for risk assessment to achieve two - way feedback. And the lower - layer model embeds an adjustment mechanism for dynamic confidence intervals to quantify the capacity thresholds under different risk preferences.
[0123] Next, an elaboration on the upper - layer model is as follows:
[0124] The upper - layer model takes minimizing the operating cost as the upper - layer objective. The operating cost of the power system mainly includes the power generation cost of conventional units and the cost of wind power accommodation, etc.
[0125] The upper - layer model is modeled based on the upper - layer objective function, and the upper - layer objective function is:
[0126]
[0127] where \(T\) represents the optimized time interval, \(C\) g is the power generation cost coefficient of the conventional unit, \(P\) g,t is the wind power output of the conventional wind turbine at time \(t\); \(C\) w is the accommodation cost coefficient of the wind power output, \(P\) w,t is the wind power output at time \(t\).
[0128] Furthermore, the constraint conditions that the upper - layer model needs to satisfy:
[0129] Limit the wind power output of the wind turbine: For each time \(t\), the wind power output of the conventional unit needs to be within the allowable upper and lower limits corresponding to each conventional wind turbine, that is, \(P\) g,min ≤\(P\) g,t ≤\(P\) g,max where \(P\) g,t is the wind power output at time \(t\), \(P\) g,minis the lower limit of wind power output at time t; P g,max is the upper limit of wind power output at time t;
[0130] Constrain the power balance of wind turbines: The total power generation of wind turbines in the wind power system at each moment should be equal to the load demand, that is, P g,t +P w,t =L t , where L t is the system load at time t;
[0131] Constrain the accommodation of wind power output: In the formula, is the upper limit of the prediction error of wind power output.
[0132] Next, the lower-layer model is elaborated:
[0133] The lower-layer model aims to minimize the risk of tie-line power flow exceeding the limit as the lower-layer objective. As a key channel connecting different regional power systems, if the power flow of the tie-line exceeds the capacity limit, it may lead to the instability of system operation.
[0134] The lower-layer model is modeled based on the lower-layer objective function, and a dynamic confidence interval is embedded in the lower-layer model. The lower-layer objective function is:
[0135]
[0136] Among them, P l s is the tie-line power flow under scenario s, P l max is the capacity limit of the tie-line, max(0, P ls -P lmax ) is the over-limit value when the tie-line power flow exceeds the capacity limit. If the tie-line power flow does not exceed the capacity limit, then max(0, P ls -P lmax ) takes a value of 0. Minimizing the sum of squares of the parts where the tie-line power flow exceeds the capacity limit in all scenarios can effectively reduce the risk of tie-line power flow exceeding the limit.
[0137] To sum up, the optimization process of the two-layer robust optimization model is simply as follows: The upper-layer model transfers the unit output plan and the wind power accommodation plan to the lower-layer model, and the lower-layer model feeds back the risk value of tie-line power flow exceeding the limit and the dynamic capacity constraint to the source-load joint scenario.
[0138] V. When determining the evaluation risk probability index, the tie-line power flow over-limit probability function is adopted:
[0139]
[0140] Among them, I(·) is an indicator function, and N is the total number of scenarios. When the condition inside the parentheses holds, the value of the indicator function is 1, otherwise it is 0. That is, for each scenario s, if the tie-line power flow P l s exceeds the capacity limit P l max , then I(P ls >P lmax ) = 1, otherwise it is 0;
[0141] The Risk Probability Index (RPI) is used to measure the probability of tie-line power flow over-limit. Summing up the values of the indicator functions for all scenarios and dividing by the total number of scenarios N gives the probability of tie-line power flow over-limit.
[0142] When determining the expected capacity deficit, the expected capacity deficit function is adopted:
[0143]
[0144] For each scenario s, max(0, P l s -P l max ) represents the part by which the tie-line power flow exceeds the capacity limit, and is 0 if not exceeded.
[0145] The expected capacity deficit is used to measure the average deficit size of tie-line power flow over-limit. Summing up the capacity deficit values for all scenarios and dividing by the total number of scenarios N gives the expected capacity deficit.
[0146] With the help of these two indicators, namely the risk probability index and the expected capacity deficit, decision-makers can more intuitively and accurately understand the capacity adequacy risk status of the power system under different scenarios, and thus make more reasonable and scientific decisions.
[0147] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. In the above embodiments, various changes and modifications can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A power capacity assessment method based on dynamic confidence interval and double-level optimization, characterized in that: Based on the historical data and predicted value of wind power output, the prediction error of wind power output is obtained; Based on the prediction error, determine the influencing parameters that produce the error; Create a power supply side model based on the prediction error and influencing parameters; Based on historical load data, demand response execution records, and external influencing factors, the demand response sensitivity coefficient and time-varying adjustment parameters are obtained; Create a load side model based on the demand response sensitivity coefficient and time-varying regulation parameters; Create a source-load joint scenario model based on the power side model and the load side model; Optimizing the source-load joint scenario model through a two-layer robust optimization model; An assessment risk probability index and capacity deficit expectation are determined based on the optimized source-load joint scenario.
2. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 1, wherein: When creating the power side model, First, collect historical data and forecast values of wind power output; Clean historical data and forecast values, remove outliers, and perform normalization; Through ΔP w = P w,实际 - P w,预测 to calculate the prediction error Among them, P w,实际 is the historical data of the electrical output, P w,预测 is the predicted value of the wind power output, and ΔP w is the prediction error of the wind power output; Determine the probability density function of the power supply side as where K represents the number of sub - distributions in the mixture distribution; α k is the mixing weight of the k - th sub - distribution; and it satisfies that is, the sum of the weights of all sub - distributions is 1, and each weight is a positive value; represents the k - th sub - distribution, μ k and are the mean and variance of the k - th sub - distribution respectively; Create a power-side model based on the power-side probability density function.
3. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 2, characterized in that: Determine α, μ, and σ values through the EM algorithm k , μ k , and σ k . By alternately executing the expectation step and the maximization step, the optimal parameter value is gradually approached; In the expectation step, the posterior probability of each data point belonging to each subdistribution is calculated based on the current parameter estimates; In the maximization step, the posterior probability is used to update the parameter value, and this process is repeated for multiple iterations until the parameters converge to obtain the parameters that best fit the historical data.
4. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 1, wherein: When creating a load side model, Based on the demand response sensitivity coefficient and time-varying adjustment parameters, the load side probability density function is determined as follows: Among them, ξ is the demand response sensitivity coefficient, γ and β are time-varying adjustment parameters, γ is used to adjust the mean shift of the load deviation, and β is used to adjust the fluctuation range of the load deviation; Create a load side model based on the load side probability density function.
5. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to any one of claims 1-4, characterized in that: When creating a source-load joint scenario model, An improved conditional generative adversarial network is used, wherein the improved conditional generative adversarial network includes a generator and a discriminator, wherein the discriminator includes a first discriminator and a second discriminator; The first discriminator constrains the spatiotemporal correlation of the power output in the source-load joint scenario model, so that the generated power scenario is close to the real power scenario; The second discriminator constrains the temporal and spatial correlation of load demand in the source-load joint scenario model, so that the generated power scenario is close to the real power scenario; The generator generates a set of joint scenes: Among them, represents the power output under scenario s, represents the load demand under scenario s, and N is the total number of generated scenarios.
6. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 5, wherein: A dynamic confidence interval is set in the source-load joint scenario model. The formula used in the dynamic confidence interval is: The confidence level α is adjusted by the risk preference parameter ρ, Among them, ρ is the risk preference parameter, α is the confidence level, and ρ hist is the historical risk preference data set, and max(ρ hist ) is the maximum value in the historical risk preference data set.
7. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 5, wherein: Constructing the two-layer robust optimization model, wherein the two-layer robust optimization model includes an upper layer model and a lower layer model, The upper layer model is modeled based on the upper layer objective function, and the upper layer objective function is: Among them, T represents the optimized time interval, and C g is the power generation cost coefficient of the conventional unit, and P g,t is the wind power output of the conventional wind turbine at time t; C w is the accommodation cost coefficient of the wind power output, and P w,t is the wind power output at time t.
8. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 7, wherein The upper model needs to satisfy the following constraints: Limit the wind power output of wind turbines: For each moment t, the wind power output of conventional units needs to be within the allowable upper and lower limits corresponding to each conventional wind turbine, that is, P g,min ≤P g,t ≤P g,max , where P g,t is the wind power output at moment t, and P g,min is the lower limit of the wind power output at moment t; P g,max is the upper limit of the wind power output at moment t; Constrain the power balance of wind turbines: the total power generation of wind turbines in the wind power system at each moment should be equal to the load demand, that is, P g,t +P w,t =L t , where L t is the system load at time t; Restrict the accommodation of wind power output: In the formula, Upper limit of the prediction error of wind power output.
9. The power capacity evaluation method based on dynamic confidence interval and double-layer optimization according to claim 7, characterized in that The lower layer model is modeled based on the lower layer objective function, and a dynamic confidence interval is embedded in the lower layer model. The lower layer objective function is: Among them, is the tie-line power flow under scenario s, is the upper limit of the tie-line capacity, and max(0, P ls - P lmax ) is the over-limit value when the tie-line power flow exceeds the capacity limit. If the tie-line power flow does not exceed the capacity limit, then max(0, P ls - P lmax ) takes the value of 0.
10. The power capacity assessment method based on dynamic confidence interval and double-layer optimization according to claim 9, characterized in that: When determining the evaluation risk probability index, the tie line power flow exceeding limit probability function is used: Among them, I(·) is an indicator function, and N is the total number of scenarios. When the condition inside the parentheses holds, the value of the indicator function is 1, otherwise it is 0. That is, for each scenario s, if the tie-line power flow exceeds the capacity limit then I(P ls >P lmax ) = 1, otherwise it is 0; When determining the expected capacity deficit, the expected capacity deficit function is used: For each scenario s, represents the part of the tie-line power flow that exceeds the capacity limit, and is 0 if it does not exceed.
Citation Information
Cited By
High-permeability new energy power distribution network collaborative optimization method, system and device and storage medium
CN121076935A
A high-permeability new energy power distribution network collaborative optimization method, system, device and storage medium
CN121076935B
Virtual power plant collaborative optimization method for dynamically adjusting reserve capacity based on risk preference
CN121584782A
