Multi-risk management method based on conditional value-at-risk model in worst scene

Abstract

The invention discloses a multi-risk management method based on a conditional value-at-risk model in a worst scene, and the method comprises the steps: setting an operation mode of a virtual power plant in a single power market, and analyzing the bidding profit and constraint of the virtual power plant; a typical distribution scene of uncertain variables which can bring risks is constructed by adopting a Monte Carlo simulation method and a fast preceding elimination technology based on probability distance, and then a conditional value-at-risk model under the worst scene is constructed and the maximization of the conditional value-at-risk under the worst scene is taken as a target. And converting a max-min problem into a semi-definite programming problem by adopting a duality theory, and solving to obtain an optimal bidding strategy of the virtual power plant after the management risk. The method can be used as a powerful tool for assisting decision making, theoretical support is provided for a virtual power plant to make a scheme for improving random variable prediction precision, market risks caused by prediction errors are minimized, and a transaction strategy is closer to an optimal strategy in an accurate situation.

Description

Technical Field

[0001] The present invention relates to a technology in the field of virtual power plant control, specifically a multiple risk management method based on a conditional risk value model under the worst scenario. Background Art

[0002] When conducting power transactions, virtual power plants need to face a series of uncertainties, such as the fluctuation of electricity prices in the day-ahead market, the output fluctuation of new energy distributed power sources (such as wind and solar units) due to natural environmental factors, and the accuracy of load forecasting. Relying solely on traditional risk management methods can only manage one of the risks, and it is impossible to assess its losses to virtual power plants, resulting in a waste of adjustable load regulation capabilities. Summary of the invention

[0003] In view of the fact that the prior art does not take into account the uncertainty factors of load forecasting errors and electricity price fluctuations, is difficult to cope with complex and changeable actual situations, and has low application efficiency under extreme conditions, the present invention proposes a multiple risk management method based on a conditional risk value model under the worst scenario. After analyzing the bidding profits and constraints of virtual power plants in the day-ahead main energy market, the Monte Carlo simulation method and the fast predecessor elimination technology based on probability distance are used to construct a typical distribution scenario of uncertain variables that may bring risks. Finally, a conditional risk value model under the worst scenario is constructed.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a multiple risk management method based on a conditional risk value model under the worst scenario, comprising:

[0006] Step 1) Set the operation mode of the virtual power plant in the single electricity market and analyze its bidding profits and constraints.

[0007] Step 2) Monte Carlo simulation and fast previous generation elimination technology based on probability distance are used to construct typical distribution scenarios of uncertain variables that may bring risks.

[0008] Step 3) Construct a conditional risk value model under the worst scenario and take maximizing the conditional risk value under the worst scenario as the goal. Use the duality theory to transform the max-min problem into a semi-definite programming problem for solution, and obtain the optimal bidding strategy for the virtual power plant after managing the risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a flow chart of the present invention;

[0010] Figure 2 It is the prediction curve of distributed photovoltaic output and distributed wind power output;

[0011] Figure 3 is the power load forecast curve;

[0012] Figure 4 Generate graphs for 5,000 scenarios for power load forecasting;

[0013] Figure 5 Generate graphs for 25 typical scenarios for power load forecasting;

[0014] Figure 6 This is a strategy diagram for the virtual power plant to declare electricity purchase amount in different scenarios. DETAILED DESCRIPTION

[0015] like Figure 1 As shown, this embodiment relates to a multiple risk management method based on a conditional risk value model under the worst scenario, including:

[0016] Step 1: Set the operation mode of the virtual power plant in the single electricity market and analyze its bidding profit and constraints, including:

[0017] 1.1 Considering the operation mode of a virtual power plant in a single electricity market, its revenue comes from two core market businesses: purchasing electricity on behalf of users and organizing flexible and adjustable loads to participate in demand response. In the electricity purchase and sales business, the price at which the virtual power plant sells electricity to users is based on the contract electricity price, which is a fixed electricity sales price, while the price at which it purchases electricity from the power grid follows the time-of-use electricity price mechanism, which is an agency electricity purchase price. In demand response services, in order to optimize the balance of electricity supply and demand and achieve the goal of peak shaving and valley filling, system operators will implement certain economic incentives for virtual power plants, that is, compensate for the contribution of virtual power plants in demand response services. In order to encourage users to participate more actively in demand response, virtual power plants also implement compensation strategies for the response volume of users who will participate in demand response and set certain compensation prices. Based on the above-mentioned setting of the virtual power plant operation mode, the present invention further explores the profit sources and related constraints that virtual power plants need to consider when bidding on the day before.

[0018] 1.2 When a virtual power plant participates in the day-ahead main energy market bidding, its net profit is calculated based on total revenue and total cost, specifically: F LA =R LA -C LA , where: F LA is the net profit of the virtual power plant participating in the day-ahead main energy market, R LA is the total revenue of the virtual power plant participating in the day-ahead main energy market, C LA is the total cost of the virtual power plant participating in the day-ahead main energy market, where: the total revenue of the virtual power plant participating in the day-ahead main energy market R LA =R u +R idr , R uis the revenue from the virtual power plant selling electricity to users, R idr Benefits of virtual power plants participating in demand response.

[0019] The total cost C of the virtual power plant participating in the day-ahead primary energy market is LA =C E +C udr +C PV +C W +C ES , where: C E is the proxy power purchase cost of the virtual power plant, C udr is the compensation cost of the virtual power plant to users participating in demand response, C PV is the penalty cost of abandoned light for distributed photovoltaics, C W is the penalty cost of wind power abandonment, C ES is the charging and discharging cost of the energy storage device.

[0020] The revenue from the virtual power plant selling electricity to users Where: e u is the contract price of electricity sold by the virtual power plant to users, P k,t is the actual power of user load k at time t after implementing demand response, N u The number of users who entrust virtual power plants to purchase electricity on their behalf.

[0021] The benefits of virtual power plants participating in demand response Where: e idr is the demand response compensation price paid by the system operator to the virtual power plant, ΔL idr,t is the demand response amount provided by the virtual power plant to the system operator at time t.

[0022] The proxy power purchase cost of the virtual power plant Where: e E,t is the time-of-use electricity price of the main energy market agent purchasing electricity at time t, L t is the bid amount of the virtual power plant for the main energy on the day before at time t.

[0023] The compensation cost of the virtual power plant to users participating in demand response Where: e udr The compensation price provided by the virtual power plant to users participating in demand response; ΔP udr,k,t is the demand response of user load k at time t, N udr is the number of users participating in demand response.

[0024] The penalty cost of abandoned light for distributed photovoltaics Where: γ PV is the light abandonment penalty coefficient, is the actual output power of distributed photovoltaic k at time t, is the predicted output power of distributed photovoltaic k at time t, N PV is the number of distributed photovoltaics.

[0025] The penalty cost of wind power abandonment of distributed wind power Where: γ W is the wind abandonment penalty coefficient, is the actual output power of distributed wind power k at time t, is the predicted output power of distributed wind power k at time t, N W is the amount of distributed wind power.

[0026] The charging and discharging cost of the energy storage device Where: e ES is the charging and discharging cost price of the energy storage device, P ch,k,t and P dis,k,t are the charging and discharging power of energy storage device k at time t, N ES is the number of energy storage devices.

[0027] 1.3 Construct constraints for virtual power plants to participate in bidding in the day-ahead main energy market, including:

[0028] Power balance constraints: Energy storage device constraints: Distributed power constraints: User load constraints: Where: P ch,k,t and P dis,k,t are the charging and discharging power of energy storage device k at time t, and are the charge and discharge states of the energy storage device k at time t, and and All are 0-1 variables. and are the maximum charge and discharge power of the energy storage device, η ch and η dis are the charging and discharging efficiency of the energy storage device, E b is the battery capacity of the energy storage device, SOC es,k,t is the state of charge of energy storage device k at time t, SOC max and SOC min are the upper and lower limits of the state of charge of the energy storage device, and are respectively the upper and lower operating limits of user load k at time t.

[0029] Step 2) Use Monte Carlo simulation and rapid previous generation elimination technology based on probability distance to construct typical distribution scenarios of uncertain variables that may bring risks, including:

[0030] 2.1 Obtain historical data and calculate the standard deviation s of the historical data as an important indicator to measure data volatility. Then multiply the calculated standard deviation s by the standard normal distribution of generated random numbers to simulate fluctuations under different probabilities. Then superimpose the fluctuation value on the predicted value of the original random variable y and use the Monte Carlo method to generate m scenarios with equal probability, i.e. S pre (i)=D+sσ(i),i=1,2,...,m, where: S pre (i) is the i-th predicted scene generated, D is the original scene, and σ(i) is the generated random number.

[0031] 2.2 Manhattan distance is used to measure the differences between the m generated scenes, that is, Where: d ij (i,j=1,2,...,m) is the Manhattan distance between the generated i-th scene and the j-th scene, ε is the total number of data in the scene, X iu and X ju are the u-th data of the i-th scene and the j-th scene respectively. Then calculate the Manhattan distance between the m scenes and get the corresponding distance matrix

[0032] 2.3 Calculate the sum of the Manhattan probability distances between each scene and the remaining scenes to obtain the probability distance matrix Where: γ i (i=1,2,...,m) is the probability of the i-th scene generated.

[0033] 2.4 According to the probability distance matrix, select the scene with the smallest sum of probability distances with the remaining scenes as the scene to be cut, and merge the two scenes into a new scene by taking the average value, that is, S pre (k) is the newly generated scene, γ k The corresponding generation probability, S pre (g) and S pre (h) is the cut scene, γ g and γ h are their corresponding generation probabilities respectively.

[0034] 2.5 Repeat steps 2.2) to 2.4) to continuously reduce the scenes, and finally obtain N reduced scenes.

[0035] Step 3) Construct a conditional risk value model under the worst scenario and conduct risk assessment management, including:

[0036] 3.1 When x and y are the decision variables and random variables in the decision process, respectively, and the decision variable satisfies x∈X, X is the decision space, and the continuous probability density function of the random variable y is denoted by p(y), then for a given decision variable x, the loss function f(x,y) caused by y does not exceed the cumulative distribution function of the threshold α, Set the confidence level β∈(0,1), then for a given decision variable x, the calculation expression of its risk value VaR is V VaR-β (x) = min{α∈R:φ(x,α)≥β}, and obtain its corresponding conditional risk value CvaR, Constructor Helper Function Where: [f(x,y)-α] + is [f(x,y)-α] + =max{f(x,y)-α,0}, then the conditional value at risk V CVaR-β (x) = minF β (x,α).

[0037] 3.2 The probability density function of the random variable y is p(y)∈P, where P is the distribution set of some known partial information. For any fixed value x∈X, the conditional value at risk under the worst scenario is Mix the random variables and construct their distribution set using the linear combination of several known typical distributions of the random variable y Where: N d is the total number of typical distributions of the random variable y, is the d-th typical distribution of the random variable y, γ d is the corresponding weight coefficient; construct the profit function R(x,y) = -f(x,y), assume that the number of random variables in the profit function is W, and ignore the correlation between the random variables, that is, the random variables will not affect each other. When each random variable obeys N ω typical distributions, then the total number of typical distributions The conditional value at risk under the worst-case scenario under the construction of the mixed distribution set is

[0038] 3.3 Set the objective function to maximize the conditional risk value maxV under the worst scenario WCVaR , the duality theory is used to transform the max-min problem into a semidefinite programming problem for solution, and the optimal bidding strategy of the virtual power plant after risk management is obtained.

[0039] After specific experiments, the distributed photovoltaic, distributed wind power output and load forecast data are set as shown in the figure. The day-ahead main energy market electricity price forecast is also carried out, and the specific electricity price parameters are shown in Table 1.

[0040] Table 1 Electricity price parameters

[0041] In this embodiment, the confidence level in the WCVaR model is set to 0.99, and it is assumed that the day-ahead electricity price forecast error, wind and solar output forecast error, and load forecast error all conform to the normal distribution. The Monte Carlo method is first used to generate 5,000 sets of scenarios, and then the fast predecessor elimination technology based on probability distance is used to reduce them to 25 scenarios. Taking the power load forecast random variable as an example, the specific generation process of the corresponding scenario is shown in the figure.

[0042] According to the embodiment provided, the bidding scenarios of the virtual power plant in the day-ahead main energy market are divided into:

[0043] Scenario 1: The virtual power plant only acts as an agent for users to purchase electricity.

[0044] Scenario 2: The virtual power plant not only purchases electricity on behalf of users, but also organizes flexible and adjustable loads to participate in demand response.

[0045] The bidding strategy of the virtual power plant in the day-ahead main energy market under the corresponding scenario is shown in the figure.

[0046] The WE and WCVaR values ​​of the virtual power plant before and after participating in demand response are shown in Table 2

[0047] Table 2 WE and WCVaR of virtual power plants and their corresponding ratios under different scenarios

[0048] In order to further explore the sensitivity of the constructed WCVaR model, this embodiment reduces the standard deviation of the day-ahead market electricity price forecast error, distributed photovoltaic output forecast error, distributed wind power output forecast error and load forecast error by 1% respectively, and obtains the changes of WCVaR value and WE value in scenario 2 as shown in Table 3.

[0049] Table 3 Changes in WCVaR and WE values ​​of virtual power plants under different prediction errors

[0050] Based on this example, a relevant analysis was conducted on the constructed WCVaR model under multiple risks to verify its feasibility and effectiveness as a decision-making auxiliary tool.

[0051] Compared with the prior art, by comparing the WE and WCVaR and corresponding ratios of virtual power plants under different scenarios, the present invention shows that by participating in demand response, the worst expected profit of the virtual power plant under the worst scenario is improved, but the risk is increased; by comparing the changes in WCVaR and WE values ​​of virtual power plants under different prediction errors, it can be seen that improving the load forecasting accuracy has the most significant impact on risks and benefits, while, in comparison, the increase in benefits and the reduction in risks brought about by improving the accuracy of wind power forecasting and photovoltaic forecasting will be smaller.

[0052] In summary, the present invention can serve as a powerful tool to assist decision-making and provide theoretical support for virtual power plants to formulate plans to improve the prediction accuracy of random variables, thereby minimizing the market risks caused by prediction errors and making trading strategies closer to the optimal strategies under precise circumstances.

[0053] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A multiple risk management method based on a worst-case scenario conditional value-at-risk model, characterized in that: By setting the operation mode of the virtual power plant in a single electricity market and analyzing its bidding profit and constraints, the Monte Carlo simulation method and the fast previous elimination technology based on probability distance are used to construct a typical distribution scenario of uncertain variables that may bring risks, and then the conditional risk value model under the worst scenario is constructed with the goal of maximizing the conditional risk value under the worst scenario. The duality theory is used to transform the max-min problem into a semi-definite programming problem for solution, and the optimal bidding strategy of the virtual power plant after risk management is obtained. The present invention can serve as a powerful tool to assist decision-making and provide theoretical support for virtual power plants to formulate plans to improve the prediction accuracy of random variables, thereby minimizing the market risks caused by prediction errors and making trading strategies closer to the optimal strategies under precise scenarios.

2. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 1 is characterized in that: The above-mentioned setting of the operation mode of the virtual power plant in the single electricity market and the analysis of its bidding profits and constraints include: 1.1 Considering the operation mode of a virtual power plant in a single electricity market, its revenue comes from two core market businesses: purchasing electricity on behalf of users and organizing flexible and adjustable loads to participate in demand response. In the electricity purchase and sales business, the price at which the virtual power plant sells electricity to users is based on the contract electricity price, which is a fixed electricity sales price, while the price at which the virtual power plant purchases electricity from the power grid follows the time-of-use electricity price mechanism, which is an agent electricity purchase price. In demand response services, in order to optimize the balance of electricity supply and demand and achieve the goal of peak shaving and valley filling, the system operator will implement certain economic incentives for the virtual power plant, that is, compensate the virtual power plant for its contribution to demand response services. In order to encourage users to participate more actively in demand response, the virtual power plant also implements a compensation strategy for the response volume of users who will participate in demand response and sets a certain compensation price. Based on the setting of the above virtual power plant operation mode, the present invention further explores the profit sources and related constraints that the virtual power plant needs to consider when bidding on the day before; 1.2 When a virtual power plant participates in the day-ahead main energy market bidding, its net profit is calculated based on total revenue and total cost, specifically: F LA =R LA -C LA , where: F LA is the net profit of the virtual power plant participating in the day-ahead main energy market, R LA is the total revenue of the virtual power plant participating in the day-ahead main energy market, C LA is the total cost of the virtual power plant participating in the day-ahead main energy market, where: the total revenue of the virtual power plant participating in the day-ahead main energy market R LA =R u +R idr , R u is the revenue from the virtual power plant selling electricity to users, R idr the benefits of participating in demand response for virtual power plants; 1.3 Construct constraints for virtual power plants when participating in bidding in the day-ahead main energy market.

3. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 2 is characterized in that: The total cost C of the virtual power plant participating in the day-ahead primary energy market is LA =C E +C udr +C PV +C W +C ES ,in: C E is the proxy power purchase cost of the virtual power plant, C udr is the compensation cost of the virtual power plant to users participating in demand response, C PV is the penalty cost of abandoned light for distributed photovoltaics, C W is the penalty cost of wind power abandonment, C ES is the charging and discharging cost of the energy storage device; The revenue from the virtual power plant selling electricity to users Where: e u is the contract price of electricity sold by the virtual power plant to users, P k,t is the actual power of user load k at time t after implementing demand response, N u The number of users who entrust virtual power plants to purchase electricity on their behalf; The benefits of virtual power plants participating in demand response Where: e idr is the demand response compensation price paid by the system operator to the virtual plant, ΔL idr,t is the demand response amount provided by the virtual power plant to the system operator at time t; The proxy power purchase cost of the virtual power plant Where: e E,t is the time-of-use electricity price of the main energy market agent purchasing electricity at time t, L t is the bid amount of the virtual power plant for the main energy on the day before at time t; The compensation cost of the virtual power plant to users participating in demand response Where: e udr The compensation price provided by the virtual power plant to users participating in demand response; ΔP udr,k,t is the demand response of user load k at time t, N udr The number of users participating in demand response; The penalty cost of abandoned light for distributed photovoltaic Where: γ PV is the light abandonment penalty coefficient, is the actual output power of distributed photovoltaic k at time t, is the predicted output power of distributed photovoltaic k at time t, N PV is the number of distributed photovoltaics; The penalty cost of wind power abandonment of distributed wind power Where: γ W is the wind abandonment penalty coefficient, is the actual output power of distributed wind power k at time t, is the predicted output power of distributed wind power k at time t, N W is the amount of distributed wind power; The charging and discharging cost of the energy storage device Where: e ES is the charging and discharging cost price of the energy storage device, P ch,k,t and P dis,k,t are the charging and discharging power of energy storage device k at time t, N ES is the number of energy storage devices.

4. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 2 is characterized in that: The constraints include: power balance constraints: Energy storage device constraints: Distributed power constraints: User load constraints: Where: P ch,k,t and P dis,k,t are the charging and discharging power of energy storage device k at time t, and are the charge and discharge states of the energy storage device k at time t, and and All are 0-1 variables. and are the maximum charge and discharge power of the energy storage device, η ch and η dis are the charging and discharging efficiency of the energy storage device, E b is the battery capacity of the energy storage device, SOC es,k,t is the state of charge of energy storage device k at time t, SOC max and SOC min are the upper and lower limits of the state of charge of the energy storage device, and are respectively the upper and lower operating limits of user load k at time t.

5. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 1 is characterized in that: The typical distribution scenario is constructed in the following ways, including: 2.1 Obtain historical data and calculate the standard deviation s of the historical data as an important indicator to measure data volatility. Then multiply the calculated standard deviation s with the standard normal distribution of generated random numbers to simulate fluctuations under different probabilities, and superimpose the fluctuation value on the predicted value of the original random variable y. Use the Monte Carlo method to generate m scenarios with equal probability. 2.2 The difference between the m scenes generated using Manhattan distance measurement; 2.3 Calculate the sum of the Manhattan probability distances between each scene and the remaining scenes to obtain the probability distance matrix Where: γ i (i=1,2,...,m) is the probability of the i-th scene generated; 2.4 According to the probability distance matrix, select the scene with the smallest sum of probability distances with the remaining scenes as the scene to be cut, and merge the two scenes into a new scene by taking the average value, that is, S pre (k) is the newly generated scene, γ k The corresponding generation probability, S pre (g) and S pre (h) is the cut scene, γ g and γ h are their corresponding generation probabilities respectively; 2.5 Repeat steps 2.2) to 2.4) to continuously reduce the scenes, and finally obtain N reduced scenes.

6. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 5 is characterized in that: The m scenarios with equal probability refer to: S pre (i)=D+sσ(i),i=1,2,...,m, where: S pre (i) is the i-th predicted scene generated, D is the original scene, and σ(i) is the generated random number.

7. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 5 is characterized in that: The differences mentioned are: Where: d ij (i,j=1,2,...,m) is the Manhattan distance between the generated i-th scene and the j-th scene, ε is the total number of data in the scene, X iu and X ju The u-th data of the i-th scene and the j-th scene are respectively used, and then the Manhattan distance between the m scenes is calculated to obtain the corresponding distance matrix 8. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 1 is characterized in that: The risk assessment management mentioned above refers to: calculating the conditional risk value under the worst scenario based on the conditional risk value, setting the objective function to maximize the conditional risk value maxV under the worst scenario WCVaR , the duality theory is used to transform the max-min problem into a semidefinite programming problem for solution, and the optimal bidding strategy of the virtual power plant after risk management is obtained.

9. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 8 is characterized in that: The conditional risk value is obtained in the following way: when x and y are the decision variable and random variable in the decision process respectively, and the decision variable satisfies x∈X, X is the decision space, and the continuous probability density function of the random variable y is denoted by p(y), then for a given decision variable x, the loss function f(x,y) caused by y does not exceed the cumulative distribution function of the threshold α, Set the confidence level β∈(0,1), then for a given decision variable x, the calculation expression of its risk value VaR is V VaR-β (x) = min{α∈R:φ(x,α)≥β}, and obtain its corresponding conditional risk value CvaR, Constructor Helper Function Where: [f(x,y)-α] + is [f(x,y)-α] + =max{f(x,y)-α,0}, then the conditional value at risk V CVaR-β (x) = minF β (x,α).

10. The multiple risk management method based on the worst-case scenario conditional value-at-risk model according to claim 8, characterized in that: The conditional risk value under the worst scenario is obtained by the following method: the probability density function of the random variable y p(y)∈P, P is the distribution set of some known partial information, then for any fixed value x∈X, the conditional risk value under the worst scenario is Mix the random variables and construct their distribution set using the linear combination of several known typical distributions of the random variable y Where: N d is the total number of typical distributions of the random variable y, is the d-th typical distribution of the random variable y, γ d is the corresponding weight coefficient; construct the profit function R(x,y) = -f(x,y), assume that the number of random variables in the profit function is W, and ignore the correlation between the random variables, that is, the random variables will not affect each other. When each random variable obeys N ω typical distributions, then the total number of typical distributions The conditional value at risk under the worst-case scenario under the construction of the mixed distribution set is

Images