Plant-level integrated energy system KRO-PBDP robust optimization scheduling method
By introducing relaxation factor and probability density bias prediction based on the robust optimization method, the KRO-PBDP robust optimization scheduling method solves the difficulty of traditional robust optimization methods to meet the economic and robustness requirements in complex integrated energy systems. The problem has achieved the best economic operation method of the system.
Patent Information
- Application Number
- CN202411552044.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-01
- Publication Date
- 2025-02-11
AI Technical Summary
Traditional robust optimization methods are difficult to meet the economic and robustness requirements of the system when facing complex integrated energy systems, especially in the case of high-permeability new energy access and thermoelectric coupling.
A KRO-PBDP robust optimization scheduling method for probability density bias prediction is proposed. By introducing a relaxation factor on the basis of the robust optimization method, the nonlinear constraint is converted into linear constraints, and combined with probability density bias prediction, the system is optimized objective function and constraints.
This method can better balance the objective function and conservatism in the plant-level integrated energy system, reduce the problem of constraints exceeding the limit, and achieve the optimal economic operation mode of the system.
Smart Images
Figure CN119419948A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of optimization scheduling of integrated energy systems with multi-energy complementarity, and relates to a robust optimization scheduling method for integrated energy systems, and specifically to a KRO-PBDP robust optimization prediction method for probability density bias prediction. Background Art
[0002] As the power supply moves towards a green, efficient and sustainable development path, the new power system has become a hot field. At present, the system is mainly composed of mixed power generation forms such as wind power, thermal power and solar power. The high proportion of new energy installed capacity has brought new problems to my country's power system, among which the flexible consumption of abundant wind and solar power generation has become a problem that needs to be solved urgently. In order to achieve the stable consumption of new energy, researchers have been conducting research continuously, among which the most representative are the flexible adjustment and transformation of thermal power units and cogeneration transformation. The problem of thermal power load distribution in cogeneration transformation is the main difficulty and research target. Many advanced algorithms have emerged for the research on thermal power load distribution, such as genetic algorithm, particle swarm algorithm, neural network, data twin, etc. However, as the overall installed capacity of new energy exceeds that of traditional power generation represented by coal-fired power generation, the high penetration of new energy impacts the power grid, while increasing the difficulty of the "source-grid separation" operation mode. "Source-load uncertainty" and "thermal-electric coupling" are one of the main difficulties in current related research. The proposal of "integrated energy system" integrates power generation, heating, refrigeration, energy storage and other technologies, optimizes them through various combinations, and applies the optimal planning method to them, which can well solve the above difficulties. Among the various methods for optimizing the operation of integrated energy systems, robust optimization is one of the most widely used methods. However, as the components of integrated energy systems tend to be diversified, traditional robust optimization methods can no longer meet the economic and robust requirements of operation.
[0003] Improving the robust optimization method to improve the optimization effect and system conservatism of the integrated energy system is an effective way to meet the increasingly complex integrated energy system. KRO robust optimization introduces relaxation factors on the basis of traditional robust optimization methods to convert nonlinear constraints into linear constraints. However, this method still has the problems of being too conservative and constrained beyond the limit. Summary of the invention
[0004] Based on the KRO robust optimization method, the present invention provides a KRO-PBDP robust optimization scheduling method for a plant-level integrated energy system with probability density bias prediction. This method can convert nonlinear constraints in the optimization problem into linear constraints, while taking into account the conservatism of the optimization method and the constraint overrun problem, and can solve the optimal economic operation mode of the system while connecting the plant-level integrated energy system to fluctuating power sources and multiple energy conversion equipment to achieve multi-energy complementarity.
[0005] The objective of the present invention is achieved through the following technical solutions:
[0006] A KRO-PBDP robust optimization scheduling method for a plant-level integrated energy system includes the following steps:
[0007] Step 1: Plant-level integrated energy system modeling:
[0008] Taking the plant-level integrated energy system as the optimization object, the objective function and constraint conditions of its model optimization problem are modeled, and the optimal prediction and judgment model is constructed:
[0009]
[0010] Where f(x) is the objective function, st represents the feasible domain of constraints and decision variables, A and b are the coefficient matrices composed of the coefficients of all constraints, θ and are the upper and lower bounds of the decision variables;
[0011] Step 2: Based on the KRO robust optimization method, design the KRO-PBDP robust optimization method:
[0012] Step 21: Separate variables and obtain the initial constraint interval of variables:
[0013] The design selection matrix β separates the variable parameters and fixed parameters of the matrix A(i,j) to obtain independent The selection matrix β is usually a binary matrix, in which the elements are 0 or 1, 1 in the matrix indicates that the element at that position is selected, and 0 indicates that it is not selected; A(i,j) is the state matrix of the constraint condition;
[0014] Step 22: Solve the initial constraint interval and obtain the dynamic variable column set J by selecting the matrix β. When j∈J exists is a dynamic variable; after obtaining i dynamic variables, there is [a J 1j a J 2j … a J ij ], j∈J, and its initial constraint interval is in and They are [a J 1j a J 2j … a J ij ];
[0015] Step 2 and 3 Solving the extreme interval:
[0016] In obtaining the function ψ=f(a J 1j a J 2j a J ij ) to avoid falling into the local optimum and missing The point where the value is 0 is used to construct a three-dimensional pole distribution function, thereby obtaining the value corresponding to each variable The maximum and minimum values of When getting any The optimization interval at the current moment 2 (i-1) The partial derivative operation is performed, As unknown variables, assign the upper and lower limits of the remaining i-1 variables once, and then obtain 2 (i-1) The solution of a partial differential equation, the instantaneous optimal interval of a single variable is as follows:
[0017]
[0018] Step 24: Paranoid Vector Determination of:
[0019] Design initial bias vector Guide every Towards the optimum, in order to avoid the discontinuity of the function ψ, which will cause the next step of differentiation to be interrupted, the bias vector The interval includes both positive and negative intervals, that is, it crosses the 0 point;
[0020] Step 25: Iterative solution:
[0021] After completing the design of steps 21 to 24, continue to iterate; when the iterative solution satisfies the dynamic variables The interval is And at this time, it is satisfied and At this time, the variable After the iteration is completed, the robust parameters degenerate into fixed parameters, the iterative operation is exited, and the initial assignment and the intermediate function ψ are updated again until all i dynamic variables are degraded, and the operation is completed;
[0022] Step 3: Based on the KRO-PBDP robust optimization method designed in step 2, the Cplex solver is used to solve the objective function in the prediction and judgment optimal model constructed in step 1 to obtain the KRO-PBDP robust optimization scheduling result.
[0023] Compared with the prior art, the present invention has the following advantages:
[0024] 1) Compared with the KRO robust optimization algorithm, the KRO-PBDP robust optimization algorithm reduces conservatism, that is, when performing constrained optimization operations, it can better balance the objective function and conservatism.
[0025] 2) The KRO-PBDP robust optimization algorithm can effectively solve the constraint overlimit problem that may occur when the uncertain set replaces the original constraint parameters with expected values. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 Flowchart of KRO-PBDP robust optimization method;
[0027] Figure 2 This is a simplified diagram of the plant-level integrated energy system;
[0028] Figure 3 It is the transferable load utilization diagram under KRO algorithm;
[0029] Figure 4 This is the transferable load utilization diagram under KRO-PDBP robust optimal scheduling. DETAILED DESCRIPTION
[0030] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.
[0031] The present invention provides a KRO-PBDP robust optimization scheduling method for a plant-level integrated energy system. The method takes the operation optimization of the plant-level integrated energy system as the algorithm operation target. The schematic diagram of the plant-level integrated energy system is as follows: Figure 2 As shown, we first build the optimal prediction model, which first converts the original multiple dynamic constraint variables Associated with an assumed intermediate value function ψ, then solve The maximum value interval is avoided to avoid falling into the local optimum; the real-time prediction probability value is obtained for each The deviation from the trend of the middle value ψ Finally, the paranoid vector i guide Alternative value range for Changes. Figure 1 As shown, the method comprises the following steps:
[0032] Step 1: Plant-level integrated energy system modeling.
[0033] The objective function and constraint conditions of the optimization problem of the plant-level integrated energy system are modeled, and the optimal prediction and judgment model is constructed.
[0034] In this step, the basic model of the robust optimization problem is:
[0035]
[0036] Where f(x) is the objective function, st represents the feasible domain of constraints and decision variables, A and b are the coefficient matrices composed of the coefficients of all constraints, θ and are the upper and lower bounds of the decision variables.
[0037] Step 2: Based on the KRO robust optimization method, design the KRO-PBDP robust optimization method.
[0038] Step 21: Separate the variables and obtain the initial constraint interval of the variables.
[0039] The design selection matrix β separates the variable parameters and fixed parameters of the matrix A(i,j) to obtain independent a ij J , where the selection matrix β is usually a binary matrix, in which the elements are 0 or 1, 1 in the matrix indicates that the element at that position is selected, and 0 indicates that it is not selected; A(i,j) is the state matrix of the constraint conditions.
[0040] In this step, the parameters are calculated based on the following formula: calculate:
[0041]
[0042] The coefficients in the formula are the "A(i,j) is the state matrix of the constraint conditions" mentioned above, that is, the coefficients of the unknowns in each constraint condition.
[0043] Step 22: Solve the initial constraint interval and obtain the dynamic variable column set J by selecting the matrix β. When j∈J exists is a dynamic variable; after obtaining i dynamic variables, there is [a J 1j a J 2j …a J ij ], j∈J, and its initial constraint interval is in and They are [a J 1j a J 2j … a J ij ];
[0044] Step 2 and 3 Solving the extreme value interval.
[0045] In obtaining the function ψ=f(a J 1j a J 2j … a J ij ), in order to avoid falling into the local optimum and missing The point where the value is 0 is used to construct a three-dimensional pole distribution function, thereby obtaining the value corresponding to each variable The maximum and minimum values of When getting any The optimization interval at the current moment 2 (i-1) The partial derivative operation is performed. The operation principle is to As unknown variables, assign the upper and lower limits of the remaining i-1 variables once, and then obtain 2 (i-1) The solution of a partial differential equation, the instantaneous optimal interval of a single variable is as follows:
[0046]
[0047] Step 24: Paranoid Vector of determination.
[0048] Design initial bias vector Guide every Towards the optimum, in order to avoid the discontinuity of the function ψ, which will cause the next step of differentiation to be interrupted, the bias vector must be The interval contains both positive and negative intervals, that is, it crosses the 0 point, where ψ is the intermediate value function of the constraint.
[0049] In this step, The calculation follows the following principles:
[0050]
[0051] In order to make To reflect the influence of the remaining i-1 dynamic variables on the function ψ, it is necessary to solve the 2*(i-1) functions formed by bringing them into the interval size values, and then make gradual corrections with a step size of λ. The expression is as follows:
[0052]
[0053] at this time The intervals have the following forms:
[0054]
[0055] The bias vector The interval is through each constraint The upper and lower limits of the change in ψ are obtained by normalization.
[0056] Step 25: Iterate and solve.
[0057] After completing the design of steps 2 to 5, continue to iterate; when the iterative solution satisfies the dynamic variables The interval is And at this time, it is satisfied and At this time, the variable After the iteration is completed, the robust parameters degenerate into deterministic parameters, the iterative operation is exited, and the initial assignment and the intermediate function ψ are updated again until all i dynamic variables are degraded, and the operation is completed.
[0058] Step 3: Based on the KRO-PBDP robust optimization method designed in step 2, the Cplex solver is used to solve the objective function in the prediction and judgment optimal model constructed in step 1 to obtain the KRO-PBDP robust optimization scheduling result.
[0059] Example:
[0060] Based on the actual situation of a power plant, a plant-level integrated energy system was constructed for simulation research. Data was obtained through field tests and historical operation data. The transferable load ΔP in the integrated energy system was calculated by two optimization algorithms, KRO and KRO-PBDP, to improve the operation economy of the plant-level integrated energy system. The transferable load ΔP is the thermal power distributable load obtained by decoupling after the new energy is connected to the plant-level integrated energy system. Its upper and lower limits are the sum of wind power generation and solar power generation.
[0061] When the KRO algorithm is used, the optimal operation of the target cost function can be completed by performing robust calculations on the expectation of the maximum value of the power grid constraint. The simulation results are as follows: Figure 3 As shown in the figure, when the KRO robust optimization algorithm is used, the transferred electricity is completely converted into a single plant power load. Although the economy and the utilization of the transferable load ΔP are also guaranteed, it is too conservative.
[0062] After the KRO-PDBP robust optimization scheduling strategy is used, the bias vector Re-adjust ΔP, and after the iteration result corrects the above electric heating load instruction, when the intermediate value is taken and superimposed, the above ΔP is changed into an adjustment of a certain proportion. When the positive deviation is taken, the load is transferred to the electric boiler in proportion, thereby reducing the utilization ratio of the plant power while decoupling the same amount of electricity, making the economy and conservatism of the robust optimization more balanced. The results are as follows Figure 4 shown.
Claims
1. A KRO-PBDP robust optimization scheduling method for plant-level integrated energy system, characterized in that The method comprises the following steps: Step 1: Plant-level integrated energy system modeling: Taking the plant-level integrated energy system as the optimization object, the objective function and constraint conditions of its model optimization problem are modeled, and the optimal prediction and judgment model is constructed: Where f(x) is the objective function, st represents the feasible domain of constraints and decision variables, A and b are coefficient matrices composed of the coefficients of all constraints, θ and are the upper and lower bounds of the decision variables; Step 2: Based on the KRO robust optimization method, design the KRO-PBDP robust optimization method: Step 21: Separate variables and obtain the initial constraint interval of variables: The design selection matrix β separates the variable parameters and fixed parameters of the matrix A(i,j) to obtain independent The selection matrix β is usually a binary matrix, in which the elements are 0 or 1, 1 in the matrix indicates that the element at that position is selected, and 0 indicates that it is not selected; A(i,j) is the state matrix of the constraint condition; Step 22: Solve the initial constraint interval and obtain the dynamic variable column set J by selecting the matrix β. When j∈J exists is a dynamic variable; after obtaining i dynamic variables, there is [a J 1j a J 2j … a J ij ], j∈J, and its initial constraint interval is in and They are [a J 1j a J 2j … a J ij ]; Step 2 and 3 Solving the extreme interval: In obtaining the function ψ=f(a J 1j a J 2j … a J ij ), in order to avoid falling into the local optimum and missing l i The point where the value is 0 is used to construct a three-dimensional pole distribution function, thereby obtaining the value corresponding to each variable The maximum and minimum values of When getting any The optimization interval at the current moment 2 (i-1) The partial derivative operation is performed, As unknown variables, assign the upper and lower limits of the remaining i-1 variables once, and then obtain 2 (i-1) The solution of a partial differential equation, the instantaneous optimal interval of a single variable is as follows: Step 24: Bias vector l i ∈[-1,0)∪(0,1] is determined by: Design the initial bias vector l i , guide each Towards the optimum, at this time, in order to avoid the discontinuity of the function ψ, which will cause the next step of differentiation to be interrupted, the bias vector l must be i The interval includes both positive and negative intervals, that is, it crosses the 0 point; Step 25: Iterative solution: After completing the design of steps 21 to 24, continue to iterate; when the iterative solution satisfies the dynamic variables The interval is And at this time, it is satisfied and At this time, the variable After the iteration is completed, the robust parameters degenerate into the determined parameters, the iterative operation is exited, and the initial assignment and the intermediate function ψ are updated again until all i dynamic variables are completely degraded, and the operation is completed; Step 3: Based on the KRO-PBDP robust optimization method designed in step 2, the Cplex solver is used to solve the objective function in the prediction and judgment optimal model constructed in step 1 to obtain the KRO-PBDP robust optimization scheduling result.
2. The KRO-PBDP robust optimization scheduling method for plant-level integrated energy system according to claim 1 is characterized in that In the step 21, the parameters are calculated based on the following formula: calculate:
3. The KRO-PBDP robust optimization scheduling method for plant-level integrated energy system according to claim 1 is characterized in that Step 24: i The calculation follows the following principles: In order to make l i To reflect the influence of the remaining i-1 dynamic variables on the function ψ, the 2*(i-1) functions formed by bringing them into the interval size values are solved respectively, and then gradually corrected with a step size of λ. The expression is as follows: at this time The intervals have the following forms: