Energy storage coordination optimal control method based on photovoltaic power generation and electric vehicle

The Cholesky decomposition and random dual dynamic programming algorithm SDDP optimized the photovoltaic power generation and electric vehicle energy storage system, which solved the problem of coordinated control under the uncertainty of photovoltaic power generation and load demand, and improved the system resource utilization and economic benefits.

CN120377217APending Publication Date: 2025-07-25STATE GRID HUBEI ELECTRIC POWER RES INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510268283.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-07
Publication Date
2025-07-25

AI Technical Summary

Technical Problem

The failure of the prior art to effectively coordinate photovoltaic power generation and electric vehicle energy storage systems leads to waste of resources and uncertainty inability to meet load demand, especially when the correlation between photovoltaic power generation and load demand changes is not considered.

Method used

The Cholesky decomposition method is used to generate correlation scenarios between photovoltaic power generation and load demand, establish a system coordinated optimization model, and use the random dual dynamic programming algorithm SDDP to solve the model to minimize costs and formulate a system control strategy.

Benefits of technology

The optimal coordinated control under the uncertainty of photovoltaic power generation and load demand is achieved, the resource utilization rate and economic benefits are improved, and the grid operation risks are reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120377217A_ABST
    Figure CN120377217A_ABST
Patent Text Reader

Abstract

An energy storage coordination optimal control method based on photovoltaic power generation and an electric vehicle comprises the following steps: 1) using a Cholesky decomposition method to consider a scene generation process of correlation between a load demand and solar energy production, and establishing a model of solar power generation and load demand uncertainty; 2) establishing a system coordination optimization model by taking cost minimization as an objective function based on the model of solar power generation and load demand uncertainty and the optimal operation of energy storage equipment integrating photovoltaic and electric vehicles in the system; and 3) applying a random dual dynamic programming algorithm SDDP to solve the system coordination optimization model, and solving the optimal coordination control of the photovoltaic energy storage and the electric vehicle energy storage under the condition of considering the uncertainty of the photovoltaic power generation and the load demand.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of electrical engineering, and more specifically, relates to a method for optimal coordinated control of energy storage based on photovoltaic power generation and electric vehicles. Background Art

[0002] With the rapid development of technology and the increasing goals of implementing sustainable standards in the power and energy systems of various countries, renewable energy is becoming more and more popular around the world. There are more and more renewable energy sources with lower costs and environmental friendliness, and their penetration in the power grid is increasing day by day. Among all renewable energy sources, solar energy is considered to be one of the important technologies for new energy power generation because of its easy access and low cost.

[0003] With the development of photovoltaic power generation, energy storage technology has also been fully developed. Energy storage systems can be used in combination with solar panels to store the excess power generated by photovoltaic power generation and release the power when the photovoltaic power generation is insufficient, so as to improve the utilization rate and stability of the system. Similar to photovoltaic power generation and energy storage systems, electric vehicles have also received increasing attention in reducing carbon dioxide emissions. However, when integrating electric vehicles into the power grid, the uncoordinated charging and discharging brought by the bidirectional power flow ability of electric vehicles will significantly increase the peak load, which may affect the safety of power grid operation and distribution infrastructure. When coordinating their control, existing research has considered the coordinated control of electric vehicles and photovoltaic panel energy storage, but has not considered the uncertainty of photovoltaic power generation and load distribution, as well as the correlation between the uncertain changes in load demand and the changes in solar photovoltaic power generation, resulting in a series of problems such as unreasonable energy storage configuration causing waste of resources and inability to meet the load demand under different photovoltaic output scenarios.

[0004] Therefore, how to coordinate the control of the energy storage of photovoltaic power generation systems and the energy storage of electric vehicles, while considering the randomness of photovoltaic power generation and load demand and the correlation between the two, to improve the rational utilization of system resources and maximize economic benefits, has important research significance. Summary of the Invention

[0005] Aiming at the above defects or improvement requirements of the prior art, the present invention aims to provide a method for optimal coordinated control of energy storage based on photovoltaic power generation and electric vehicles, aiming to solve the optimal coordinated control of photovoltaic energy storage and electric vehicle energy storage under the condition of considering the uncertainty of photovoltaic power generation and load demand.

[0006] To achieve the above object, the present invention discloses a method for optimal coordinated control of energy storage based on photovoltaic power generation and electric vehicles, including the following steps:

[0007] 1) The scenario generation process using the Cholesky decomposition method to consider the correlation between load demand and solar power production, and establish a model for the uncertainty of solar power generation and load demand;

[0008] 2) Based on the models of the uncertainty of solar power generation and load demand and the optimal operation of energy storage devices integrating photovoltaic and electric vehicles in the system, with the cost minimization as the objective function, establish a system coordination optimization model;

[0009] 3) Apply the Stochastic Dual Dynamic Programming algorithm SDDP to solve the system coordination optimization model.

[0010] Step 1 is specifically as follows:

[0011] a. Based on the historical and predicted data of photovoltaic power generation and load demand during a 24-hour period, determine the matrix X that follows an independent distribution in the normal distribution N[0,1] according to the series T and the number n of uncertain parameters. R is the correlation matrix (n×n) between load and photovoltaic power generation. The Cholesky decomposition of R is a lower triangular matrix, satisfying:

[0012] R = LL'

[0013] b. Transmit the correlation structure through Cholesky decomposition

[0014] Define the matrix Y as Y = LX. Y is a draw starting from N[0,Σ], where Σ = R. Multiply the sampling value y of the i-th column of Y i,t by σ i , and then sum the average value u i , to obtain the sampling values of photovoltaic power generation and load demand which follows the normal distribution Ν t.i [μ i ,σ i 2 , where i = 1, 2 in a specific time period;

[0015] Generate scenarios in an independent manner according to steps a and b, representing random parameters considering the correlation effects between scenarios, while taking into account the correlation effects between scenarios.

[0016] The objective function of the system coordination optimization model is:

[0017]

[0018] Among them, C t is the time-of-use electricity price, P g,t is the power demand of the power grid at time t. k is a penalty factor to ensure that the energy storage system of the electric vehicle can reach the set SOC value before leaving the residential system is the electricity amount when the electric vehicle leaves the residence, C dIt is the penalty factor to avoid the discharge of electric vehicles during off-peak hours. It is the discharge power of electric vehicles during off-peak hours.

[0019] The constraint conditions of the system coordination optimization model include power balance constraint, charge balance constraint, and charging and discharging operation constraints based on PV and load:

[0020] The power balance constraint is:

[0021]

[0022] The charge balance constraint is:

[0023]

[0024] The charging and discharging operation constraints based on PV and load are:

[0025]

[0026] k ≥ 0

[0027] P g,t ≥ 0

[0028]

[0029] Among them, P def,t is the abandoned power of photovoltaic power generation, and are the instantaneous charging and discharging power of photovoltaic energy storage, and is the instantaneous charging and discharging power of the electric vehicle battery, and are the maximum charging power and maximum discharging power of photovoltaic energy storage, and are the minimum charging power and minimum discharging power of photovoltaic energy storage, and are the maximum charging power and maximum discharging power of electric vehicles, and are the minimum charging power and minimum discharging power of electric vehicles, and are the state of charge of photovoltaic energy storage and electric vehicles at time t, respectively, and are the upper and lower limits of the state of charge of photovoltaic energy storage, and are the upper and lower limits of the state of charge of electric vehicles, Q PV and Q PEV are the total capacities of photovoltaic energy storage and electric vehicle batteries, η PV and η PEVThey are the charging efficiencies of photovoltaic energy storage and electric vehicles respectively.

[0030] The stochastic dual dynamic programming algorithm SDDP constructs an approximation of the future cost function using piecewise linear functions, which is represented by the Benders decomposition method. It constructs a piecewise linear approximation of the future cost function using the dual variables derived from the transition constraints and adds them iteratively as the algorithm progresses. When the stopping condition is reached, the process stops.

[0031] Step 3) is specifically as follows:

[0032] a. The objective function of the system coordination optimization model aims to minimize the total cost including the current cost and the expected future cost:

[0033]

[0034] Among them, the decision variable at a specific stage t is regarded as the vector x t , which includes the electricity purchased from the grid, the charging and discharging power, and the SOC level of the energy storage system; the parameter b t represents the stochastic photovoltaic power generation and load at stage t.

[0035] b. The structural constraints of the system coordination optimization model are expressed as:

[0036] A t x t =B t x t-1 +b t :π t

[0037] Subsequently, according to the Benders decomposition scheme, a piecewise linear approximation of the future cost function is constructed using the dual variables (represented by π t ) derived from the transition constraints;

[0038] c. The simple boundaries of the decision variables of the system coordination optimization model are:

[0039] x t ≥0

[0040] The realization of the stochastic parameter b t+1 affects the state of the system at stage t. Therefore, the expected cost function at stage t + 1 for the given decision x t at stage t is realized.

[0041] Step 3) also includes defining the control strategy of the system.

[0042] The control strategy of the system is: during the forward pass of the system, the simplex method is used to solve the system coordination optimization model at each time stage;

[0043] During the solution process, the Benders decomposition accumulated in the previous iteration is used as an additional constraint, so that the result can be closer to the future cost;

[0044] In the last stage of the forward pass, the total expected cost of the system is estimated and used as the upper bound of the problem;

[0045] Considering the current and future expected costs, the lower bound of the sampling problem is calculated by solving the first-stage problem in the forward pass;

[0046] If the lower bound cost reaches the stopping criterion, the solution process is stopped; otherwise, the iteration process continues until the desired convergence level is reached;

[0047] In each iteration, independent sampling is performed in the scenario tree to obtain a new forward path;

[0048] To reach the desired convergence level, the algorithm performs a backward pass. In the backward pass, the algorithm calculates the new Benders decomposition of the previous stage to improve the approximation of the future cost function at each stage.

[0049] Generally speaking, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:

[0050] The fault diagnosis model of distribution transformer migration learning considering the evolution of multivariate factors established by the present invention migrates the fault information of the auxiliary distribution transformer to the target distribution transformer through the way of transfer learning, and preferably solves the problem brought by the small amount of single distribution transformer fault data to the distribution transformer fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a system composed of a photovoltaic energy storage and an electric vehicle according to an embodiment of the present invention;

[0052] Figure 2 is a summer solar energy and load generation profile according to an embodiment of the present invention;

[0053] Figure 3 is a visualization diagram of solving a problem by the SDDP algorithm according to an embodiment of the present invention;

[0054] Figure 4 is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments of the present invention.

[0056] As Figure 4 , a coordinated optimal control method for energy storage based on photovoltaic power generation and electric vehicles includes the following steps:

[0057] Step 1), considering the uncertainties of solar power generation and electricity demand, a scenario generation process that takes into account the correlation between load demand and solar production using Cholesky decomposition;

[0058] Step 2), based on the uncertainties of solar power generation and electricity demand and the optimal operation of energy storage devices integrated with photovoltaic and electric vehicles in the system, establish a system coordinated optimization model with cost minimization as the objective function;

[0059] Step 3), apply the Stochastic Dual Dynamic Programming algorithm SDDP to solve the model and define the control strategy of the system;

[0060] Specifically, in Step 1), future uncertainties are simulated by developing a scenario tree to represent possible events of random parameters:

[0061] a. Considering the historical and forecast data of photovoltaic power generation and household electricity demand during a 24-hour period, independently sample from the normal distribution Ν[μ,σ 2 to generate scenarios of solar production and electricity demand;

[0062] b. Transmit the correlation structure through Cholesky decomposition. Let the series be T and n be the number of uncertain parameters (in this example, n = 2). Let X be a matrix (T×n) independently distributed in the normal distribution n[0,1], and let R be the correlation matrix between load and photovoltaic power generation. The Cholesky decomposition of R is a lower triangular matrix L that satisfies

[0063] R = LL'

[0064] Define the matrix Y as

[0065] Y = LX

[0066] The original sampling comes from Ν[0,1], σ = 1, and the covariance matrix ∑ = R. The sampling of solar photovoltaic power generation and electricity demand is given by Ν t.i [μ i ,σ i 2 , where i = 1, 2 at a specific time stage t. We can multiply the sampling value of the i-th column of Y by σ i , and then sum the average value u i .

[0067]

[0068] Among them, y i,t is an element of the i-th sampling in the matrix Y, is from Ν t.i [μ i ,σ i2 Elements of sampling. Scenarios are generated according to this process to jointly represent random parameters, taking into account the correlation effects between them.

[0069] Such as Figure 1 , taking the system composed of photovoltaic energy storage and electric vehicles as an example. In this embodiment, the predicted solar power generation profile and the typical summer load profile on a summer day are as shown in Figure 2 , and the correlation coefficient between them is -0.15.

[0070] Specifically, the objective function of the system coordination optimization model in step 2) is:

[0071]

[0072] Where C t is the time-of-use electricity price, P g,t is the power demand of the power grid at time t, k is a penalty factor to ensure that the energy storage system of the electric vehicle can reach the set SOC value before leaving the residential system is the electricity amount when the electric vehicle leaves the residence, C d is a penalty factor to avoid the electric vehicle discharging during off-peak hours, is the discharging power of the electric vehicle during off-peak periods.

[0073] Specifically, in this embodiment, the time-of-use electricity price varies according to off-peak, partial-peak, and peak periods, as shown in Table 1.

[0074] Table 1

[0075]

[0076] In addition, in this example, the electric vehicle is considered to be plugged into the system at 18:00 with an SOC of 40%, and it is assumed that it will leave the system at 07:00 the next day, with an SOC of 80%. In this algorithm, we assume that the charging mode of the electric vehicle is deterministic.

[0077] Specifically, in step 2), the optimal coordinated control method based on photovoltaic power generation and energy storage of electric vehicles is characterized in that the constraint conditions include power balance constraint, charge balance constraint, and charging and discharging operation constraints based on PV and load:

[0078]

[0079] k ≥ 0

[0080] P g,t ≥ 0

[0081]

[0082] Among them, P def,t is the abandoned power of photovoltaic power generation, and are the instantaneous charge and discharge power of photovoltaic energy storage, and is the instantaneous charge and discharge power of the electric vehicle battery, and are the maximum charge power and maximum discharge power of photovoltaic energy storage, and are the minimum charge power and minimum discharge power of photovoltaic energy storage, and are the maximum charge power and maximum discharge power of the electric vehicle, and are the minimum charge power and minimum discharge power of the electric vehicle, and are the state of charge of photovoltaic energy storage and electric vehicle at time t respectively, and are the upper and lower limits of the state of charge of photovoltaic energy storage, and are the upper and lower limits of the state of charge of the electric vehicle, Q PV and Q PEV are the total capacities of photovoltaic energy storage and electric vehicle battery, η PV and η PEV are the charging efficiencies of photovoltaic energy storage and electric vehicle respectively.

[0083] Specifically, in this embodiment, and are 3KW, and are 0KW, and are 20KW, and are 0KW, and The initial values of are 20% and 40% respectively, and is 20%, and is 80%, Q PV and Q PEV are 4KWh and 85KWh respectively, η PV and η PEV is 92%.

[0084] Specifically, in step 3), the SDDP algorithm constructs an approximation of the future cost function using piecewise linear functions, which is represented by the Benders decomposition method and added iteratively as the algorithm progresses. When the stopping condition is reached, the process stops.

[0085] The general T-stage stochastic linear programming is formulated as follows:

[0086] a. The objective function of the model aims to minimize the total cost including the current cost and the expected future cost:

[0087]

[0088] where the decision variable at a specific stage t is regarded as the vector x t , which includes the electricity purchased from the grid, the charging and discharging power, and the SOC level of the energy storage system; the parameter b t represents the stochastic photovoltaic power generation and load at stage t.

[0089] b. The structural constraints of the model are expressed as:

[0090] A t x t = B t x t-1 + b t : π t

[0091] Subsequently, according to Benders' decomposition scheme, a piecewise linear approximation of the future cost function is constructed using the dual variable (represented by π t ) derived from the transition constraints.

[0092] c. The simple bounds of the decision variables of the model are:

[0093] x t ≥ 0

[0094] The realization of the stochastic parameter b t+1 affects the state of the system at stage t. Therefore, the expected cost function at stage t + 1 for a given decision x t at stage t is realized.

[0095] Specifically, in this embodiment, the SDDP algorithm simplifies the problem-solving process to a three-stage problem-solving process, as Figure 3As shown. In the forward pass, the simplex method is used to solve the model at each time stage. During the solution process, the Benders cuts accumulated in the previous iteration are used as additional constraints to better estimate future costs and improve the decision-making process. At the last stage of the forward pass, the total expected cost is estimated and used as the upper bound of the problem. Considering the current and future expected costs, the lower bound of the sampling problem is calculated by solving the first-stage problem in the forward pass. If the lower bound cost reaches the stopping criterion, i.e., within the 95% confidence interval of the upper bound cost, the SDDP process stops. Otherwise, the iterative process continues until the desired convergence level is reached. In each iteration, new forward paths are sampled independently in the scenario tree. When the desired convergence level is reached, the algorithm performs a backward pass as shown in Figure 3 (b). In the backward pass, the algorithm calculates new Benders cuts for the previous stage to improve the approximation of the future cost function at each stage.

[0096] It is easy for those skilled in the art to understand that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. A coordinated optimal control method for energy storage based on photovoltaic power generation and electric vehicles, characterized in that It includes the following steps: 1) Use the Cholesky decomposition method to establish a model of the uncertainty of solar power generation and load demand in the scenario generation process considering the correlation between load demand and solar power production; 2) Based on the model of the uncertainty of solar power generation and load demand and the optimal operation of the energy storage devices integrating PV and electric vehicles in the system, establish a system coordinated optimization model with the cost minimization as the objective function; 3) Apply the Stochastic Dual Dynamic Programming Algorithm SDDP to solve the system coordinated optimization model and...

2. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 1, wherein Step 1 is specifically as follows: a. Based on the historical and predicted data of photovoltaic power generation and load demand during a 24-hour period, determine the matrix X independently distributed in the normal distribution N[0,1] according to the series T and the number n of uncertain parameters. R is the correlation matrix (n×n) between the load and photovoltaic power generation. The Cholesky decomposition of R is a lower triangular matrix, satisfying: R = LL' b. Transfer the correlation structure through Cholesky decomposition Define the matrix Y as Y = LX, where Y is a draw starting from N[0, Σ], with Σ = R. Let the sampled value of the i-th column of Y be y i,t multiply by σ i , and then sum the average value u i , to obtain the sampled values for photovoltaic power generation and load demand which follows a normal distribution Ν t.i [μ i , σ i 2 , where i = 1, 2 in a specific time period.

3. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 1, wherein, The objective function of the system coordinated optimization model is: Among them, C t is the time-of-use electricity price, P g,t is the power demand of the power grid at time t, and k is the penalty factor to ensure that the energy storage system of the electric vehicle can reach the set SOC value before leaving the residential system is the electricity quantity when the electric vehicle leaves the residence, C d is the penalty factor to avoid the discharge of the electric vehicle during off-peak hours, and is the discharge power of the electric vehicle during off-peak periods.

4. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 1, characterized in that, The constraint conditions of the system coordinated optimization model include power balance constraint, charge balance constraint, and charging and discharging operation constraints based on PV and load: The power balance constraint is: The charge balance constraint is: The charging and discharging operation constraints based on PV and load are: Among them, P def,t is the amount of abandoned power of photovoltaic power generation, and are the instantaneous charge and discharge powers of photovoltaic energy storage, and are the instantaneous charge and discharge powers of the electric vehicle battery, and are the maximum charge power and maximum discharge power of photovoltaic energy storage, and are the minimum charge power and minimum discharge power of photovoltaic energy storage, and are the maximum charge power and maximum discharge power of the electric vehicle, and are the minimum charge power and minimum discharge power of the electric vehicle, and are the state of charge of photovoltaic energy storage and electric vehicle at time t respectively, and are the upper and lower limits of the state of charge of photovoltaic energy storage, and are the upper and lower limits of the state of charge of the electric vehicle, Q PV and Q PEV are the total capacities of photovoltaic energy storage and electric vehicle battery, η PV and η PEV are the charging efficiencies of photovoltaic energy storage and electric vehicle respectively.

5. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 1, wherein The Stochastic Dual Dynamic Programming Algorithm SDDP constructs an approximation of the future cost function using piecewise linear functions, which is represented by the Benders decomposition method. The piecewise linear approximation of the future cost function is constructed using the dual variables derived from the transition constraints and added iteratively as the algorithm progresses. When the stopping condition is reached, the process stops.

6. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 1, wherein Step 3) is specifically as follows: a. The objective function of the system coordinated optimization model aims to minimize the total cost including the current cost and the expected future cost: Among them, the decision variable at a specific stage t is regarded as a vector x t , which includes the electricity purchased from the power grid, the charging and discharging power, and the SOC level of the energy storage system; the parameter b t represents the stochastic photovoltaic power generation and load at stage t; b. The structural constraints of the system coordinated optimization model are expressed as: A t x t = B t x t-1 + b t : π t Subsequently, according to the Benders' decomposition scheme, a piecewise linear approximation of the future cost function is constructed using the dual variables (denoted by π t ) derived from the transition constraints; c. The simple boundaries of the decision variables of the system coordinated optimization model are: x t ≥0 Random parameter b t+1 The implementation affects the state of the system at stage t. Therefore, the expected cost function at stage t+1 for a given decision x at stage t is implemented t .

7. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 1, characterized in that Step 3) also includes defining the control strategy of the system.

8. The energy storage coordination optimal control method based on photovoltaic power generation and electric vehicles according to claim 7, characterized in that, The control strategy of the system is: during the forward pass of the system, use the simplex method to solve the system coordinated optimization model at each time stage; During the solution process, use the Benders decomposition accumulated in the previous iteration as an additional constraint to make the result closer to the future cost; In the last stage of the forward pass, estimate the total expected cost of the system and use it as the upper bound of the problem; Under the consideration of the current and future expected costs, calculate the lower bound of the sampling problem by solving the first-stage problem in the forward pass process; If the lower bound cost reaches the stopping criterion, stop the solution process; otherwise, the iteration process will continue until the desired convergence level is reached; Each iteration performs independent sampling in the scenario tree to obtain a new forward path; To reach the desired convergence level, the algorithm performs a backward pass. In the backward pass, the algorithm calculates the new Benders decomposition of the previous stage to improve the approximation of the future cost function at each stage.