Integrated energy system optimization scheduling method based on two-stage probability flow solution algorithm

Abstract

The invention discloses an integrated energy system optimization scheduling method based on a two-stage probability flow solving algorithm, and the method comprises the steps: constructing a multi-target optimization model of an integrated energy system, and determining the constraint condition of the optimization model; a probability flow is introduced, an NSGA algorithm is improved, and a two-stage probability flow solving algorithm is obtained; solving the multi-target optimization model by adopting the two-stage probability flow solving algorithm to obtain an optimal solution, namely an optimal scheduling scheme; and performing operation scheduling on the integrated energy system according to the optimal scheduling scheme. According to the solving algorithm, the probability flow is introduced, it can be ensured that the global optimal solution is searched, and the overall solving efficiency and optimization precision are improved. According to the optimal scheduling scheme obtained according to the two-stage probability flow solving algorithm, operation scheduling is carried out on the integrated energy system, the system operation cost is effectively reduced, the environmental benefits are optimized, meanwhile, dependence on external electric power is remarkably reduced, and double improvement of economical efficiency and environmental benefits is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of optimal scheduling of energy systems, and particularly relates to an optimal scheduling method for an integrated energy system based on a two-stage probability flow solution algorithm. Background Art

[0002] At present, the energy system is accelerating its transformation and upgrading towards high efficiency and greenness. The integrated energy system realizes the complementary and collaborative optimal operation of multiple energy forms through electric energy, heat energy, cold energy and renewable energy. However, there are still many challenges in achieving the collaborative optimization of multiple objectives such as economy, carbon emissions and electricity purchase volume in the integrated energy system.

[0003] At present, the integrated energy system involves the collaborative operation of multiple energy forms and multiple devices, and factors such as energy supply-demand balance, equipment capacity limitation, energy storage state constraint and energy flow constraint are coupled with each other, increasing the complexity of system operation. In addition, there is a natural trade-off relationship between the economy of system operation and carbon emissions. How to reduce carbon emissions to the greatest extent while reducing the system operation cost is also one of the key tasks for realizing the green and low-carbon transformation of the energy system.

[0004] Existing algorithms for solving the multi-objective optimization problem of integrated energy systems, such as particle swarm algorithm and NSGA-III, are prone to falling into local optimum and have poor optimization ability.

[0005] Therefore, it is necessary to study and improve the solution algorithm for the multi-objective optimization problem of integrated energy systems. Summary of the Invention

[0006] The object of the present invention is to solve the above problems, and provide an optimal scheduling method for an integrated energy system based on a two-stage probability flow solution algorithm. By combining a two-stage optimization strategy based on probability flow guidance, the non-dominated sorting genetic algorithm is improved to solve the problems of insufficient search of the solution space, difficulty in balancing convergence and diversity, and uneven distribution of objective weights in the current multi-objective collaborative optimization of economy, carbon emissions, electricity purchase volume and renewable energy penetration rate in the integrated energy system.

[0007] In order to achieve the above object, the technical solution provided by the present invention is as follows: An optimal scheduling method for an integrated energy system based on a two-stage probability flow solution algorithm, comprising the following steps: Step 1: Construct a multi-objective optimization model of the integrated energy system, and the optimization objectives of the multi-objective optimization model include minimizing the operation cost of the integrated energy system, minimizing the total carbon emissions and minimizing the electricity purchase volume of the integrated energy system; Step 2: Determine the constraint conditions of the multi-objective optimization model; Step 3: Introduce the probability flow to improve the non-dominated sorting genetic algorithm, and obtain a two-stage probability flow solution algorithm, where the two-stage probability flow solution algorithm includes an exploration stage and a convergence stage; Step 4: Use the two-stage probability flow solution algorithm to solve the multi-objective optimization model to obtain the optimal solution, that is, the optimal scheduling plan; Step 5: Perform operation scheduling on the integrated energy system according to the optimal scheduling plan obtained in Step 4.

[0008] Preferably, the integrated energy system includes a gas turbine, a gas boiler, a energy storage battery, an absorption chiller, an electric chiller, an electric boiler, and a heat storage tank.

[0009] The multi-objective optimization model of the integrated energy system is: ; (1) Among them, is the objective function of the multi-objective optimization of the integrated energy system, is the annual operating cost of the integrated energy system, is the total annual carbon emissions, is the electricity purchase volume of the integrated energy system.

[0010] The annual operating cost of the integrated energy system The expression is: ; (2) In the formula, is the initial purchase cost of the integrated energy system equipment, is the replacement cost after the equipment reaches the end of its life, is the daily operation and maintenance cost of all equipment, is the gas usage cost of the gas turbine and the gas boiler, represents the grid electricity purchase cost.

[0011] The total annual carbon emissions of the integrated energy system The expression is: ; (3) In the formula, is the carbon emission factor generated by the combustion of natural gas per unit energy, is the thermal input power of the gas turbine, is the thermal output power of the gas boiler, is the thermal efficiency of the gas boiler, is the carbon emission factor generated by the grid electricity per unit energy, is the electricity volume purchased by the energy system from the grid; t represents the time period serial number.

[0012] The electricity purchase volume of the integrated energy system The expression is: ; (4) In the formula, is the electricity quantity purchased by the energy system from the power grid.

[0013] Preferably, in step 2, the constraint conditions of the multi-objective optimization model include the constraints on the power generation power and heat production power of the gas turbine, the constraint on the heat production power of the gas boiler, the constraints on the charging power and discharging power of the energy storage battery, and the constraint on the output cooling power of the absorption chiller.

[0014] Furthermore, in step 2, the constraint conditions of the multi-objective optimization model also include the power balance of electric energy, heat energy, and cooling energy in the integrated energy system.

[0015] Preferably, in step 3, the two-stage probability flow solution algorithm specifically includes: 1) Exploration stage: Using probability flow to introduce randomness to conduct heuristic search on the solution set of the multi-objective optimization problem, and heuristically explore the solution space by adjusting the probability weights of the solution set flow; Adopting an adaptive threshold optimization strategy to optimize the misclassification problem between non-dominated solutions and dominated solutions, which can effectively balance the convergence and diversity of solutions, and accelerate the search process of identifying and promoting potential optimal solutions.

[0016] 2) Convergence stage: Introducing a selection mechanism with the idea of normal distribution probability to select the optimal solution set flow, effectively balancing the weights between optimization objectives, avoiding selecting extreme individuals in some objectives, and more comprehensively exploring the solution space.

[0017] Compared with the prior art, the beneficial effects of the present invention include: 1) The present invention constructs a multi-objective optimization model for the integrated energy system with the minimum operating cost, the minimum total carbon emissions, and the minimum electricity purchase quantity, and uses a two-stage probability flow solution algorithm improved from the non-dominated sorting genetic algorithm to solve the multi-objective optimization model of the integrated energy system, ensuring the search for the global optimal solution, and improving the overall solution efficiency and optimization accuracy; performing operation scheduling on the integrated energy system according to the optimal scheduling plan obtained by the two-stage probability flow solution algorithm, effectively reducing the system operating cost, reducing the carbon emissions, optimizing the environmental benefits, and at the same time significantly reducing the dependence on external power, achieving a double improvement in economic and environmental benefits.

[0018] 2) The two-stage probability flow solution algorithm of the present invention can dynamically adjust the search strategy according to the current solution during the optimization process compared with the existing solution algorithms. In the initial stage, it emphasizes exploration diversity. By introducing more randomness and breadth, it quickly traverses the solution space to find potential global optimal solutions in the solution space. In the later stage, it emphasizes convergence. It adaptively adjusts the search step size and direction according to the quality of the solution, can avoid over-convergence and improve accuracy, fully utilizes the complementary advantages of global and local information, and improves the solution efficiency and optimization accuracy.

[0019] 3) By introducing probability flow, the two-stage probability flow solution algorithm of the present invention can dynamically evaluate the quality of the solution set and preferentially allocate higher search probabilities to high-quality solution regions according to the evaluation results. 4) By introducing the normal distribution probability method, the two-stage probability flow solution algorithm of the present invention can dynamically weaken the influence of extreme solutions deviating from the mainstream through the natural attenuation in the probability density function, automatically match lower selection weights for extreme values, improve the quality of the non-dominated solution set, and there is no need for manual adjustment of the selection pressure.

[0020] 5) The two-stage probability flow solution algorithm of the present invention dynamically adjusts the exploration direction of the solution set through probability weights, retains diversity while strengthening the aggregation ability for high-quality solutions, and realizes the unity of the global distribution breadth and the local convergence depth.

[0021] 6) The two-stage probability flow solution algorithm of the present invention uses the threshold optimization method to dynamically and flexibly adjust the threshold, maximize the distribution between solution set flows, avoid the drawbacks of over-aggregation or dispersion of the solution set in traditional algorithms, and precisely control the selection pressure of the solution set. By dynamically setting a reasonable threshold, it can effectively eliminate low-quality solutions, greatly reduce redundant calculations and ineffective search processes, significantly improve the algorithm operation efficiency, and is especially suitable for complex optimization scenarios with a large number of objectives. Through the hard shielding of extreme solutions and noise interference by the threshold, it ensures that the search direction always advances around the reasonable solution space, enhances the anti-interference ability of the algorithm in a dynamic environment and the stability of the search process. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention will be further described below in conjunction with the drawings and embodiments.

[0023] Figure 1 It is a schematic diagram of the integrated energy system according to the embodiment of the present invention.

[0024] Figure 2 It is a schematic flow chart of the two-stage probability flow solution algorithm according to the embodiment of the present invention.

[0025] Figure 3 It is a schematic diagram of distinguishing probability flow based on the Poisson cumulative distribution function according to the embodiment of the present invention.

[0026] Figure 4 This is the iterative optimization curve graph of three different solution algorithms in the embodiments of the present invention. Specific implementation manners

[0027] Figure 1 The shown integrated energy system of the embodiment mainly includes an electric network, a thermal network, and a cold network; the integrated energy system operates in coordination through an electric chiller, an absorption chiller, a gas turbine, a gas boiler, an energy storage battery, a heat storage tank, etc., to achieve the integrated scheduling and efficient utilization of energy. The energy storage battery of the embodiment is a storage battery.

[0028] The main equipment and functions in this energy storage system include: Gas turbine and gas boiler: Use natural gas for combustion power generation and simultaneously generate heat energy. The gas turbine has the characteristics of efficient energy conversion and can convert natural gas into electric energy and heat energy. The gas boiler uses natural gas for combustion to generate heat energy for heating and refrigeration.

[0029] Electric chiller and absorption chiller: Use electric energy and heat energy for refrigeration to achieve the efficient production and utilization of cold energy. The electric chiller is mainly driven by electricity, while the absorption chiller uses heat energy for refrigeration.

[0030] Storage battery and heat storage tank: Used to store energy when the energy is sufficient and release energy during the peak energy demand period to ensure the stable operation of the system.

[0031] An integrated energy system optimal scheduling method based on a two-stage probability flow solution algorithm includes: Step 1: Construct a multi-objective optimization model of the integrated energy system, and the optimization objectives of the multi-objective optimization model include minimizing the operating cost of the integrated energy system, minimizing the total carbon emissions, and minimizing the electricity purchase amount of the integrated energy system.

[0032] The multi-objective optimization model of the integrated energy system is: ; (1) Wherein, is the objective function of the multi-objective optimization of the integrated energy system, is the annual operating cost of the integrated energy system, is the total annual carbon emissions, is the electricity purchase amount of the integrated energy system.

[0033] The annual operating cost of the integrated energy system The expression is: ; (2) In the formula, is the initial purchase cost of the equipment of the integrated energy system, is the replacement cost after the equipment reaches the end of its life, is the daily operation and maintenance cost of all devices, is the gas usage cost of gas turbines and gas boilers, represents the cost of purchasing electricity from the power grid.

[0034] The total annual carbon emissions of the integrated energy system The expression is: ; (3) In the formula, is the carbon emission factor generated by the combustion of natural gas per unit of energy, is the thermal input power of the gas turbine, is the thermal output power of the gas boiler, is the thermal efficiency of the gas boiler, is the carbon emission factor generated by the power grid electricity per unit of energy, is the electricity quantity purchased by the energy system from the power grid; t represents the time period serial number.

[0035] The electricity quantity purchased by the integrated energy system The expression of ; (4) In the formula, is the electricity quantity purchased by the energy system from the power grid.

[0036] Step 2: Determine the constraint conditions of the multi-objective optimization model; The constraint conditions for the power generation power and heat production power of the gas turbine are: ; (5) ; (6) In the formula, is the power generation power of the gas turbine, is the heat production power of the gas turbine, is the gas consumption power of the gas turbine, , are the power conversion efficiency and heat energy conversion efficiency of the gas turbine respectively, , are the minimum limit and maximum limit of the gas turbine power output respectively, represents the real-time output power value of the gas turbine.

[0037] The constraint conditions for the heat production power of the gas boiler are: ; (7) ; (8) In the formula, is the heat production power of the gas boiler, is the fuel consumption power of the gas boiler, is the heat conversion efficiency of the gas boiler, and are the minimum and maximum limits of the heat production power of the gas boiler.

[0038] The constraints on the charging power and discharging power of the energy storage battery are as follows: ; (9) In the formula, and are respectively , the stored electricity of the energy storage battery system during the period, is the self-discharge coefficient of the energy storage battery, and are the charging and discharging powers of the energy storage battery, and are the charging and discharging powers of the energy storage battery.

[0039] The constraints on the output cooling power of the absorption chiller are as follows: ; (10) ; (11) ; (12) In the formula, is the output cooling power of the absorption chiller, is the efficiency of the absorption chiller, is the input heat efficiency of the absorption chiller, is the cooling load of the absorption chiller, is the maximum cooling load of the absorption chiller, is the cooling load required by the system.

[0040] The output cooling power of the electric chiller is: ; (13) In the formula, is the output cooling power of the electric chiller, is the input electric power, is the refrigeration efficiency; The heating power of the electric boiler is: ; (14) ; (15) In the formula, is the actual heating power of the electric boiler, is the heating efficiency of the electric boiler, is the power consumption of the electric boiler; The heat storage power and heat release power of the heat storage tank are as follows: ; (16) In the formula, and are respectively the moment and the moment of the heat storage amount in the heat storage tank, is the heat loss coefficient of the heat storage tank, and are respectively the heat storage and heat release powers of the heat storage tank, and are respectively the heat storage and heat release efficiencies of the heat storage tank.

[0041] The constraint conditions of the integrated energy system also include the power balance of electric energy, thermal energy, and cold energy: The electric energy balance equation is: ; (17) The thermal energy balance equation is: ; (18) The cold energy balance equation is: ; (19) In the formula, is the electric power provided by the wind, is the electric power provided by the photovoltaic, , and are respectively the electric load, thermal load, and cold load delivered by the system to the user side.

[0042] Step 3: Introduce the probability flow to improve the non-dominated sorting genetic algorithm, and obtain a two-stage probability flow solution algorithm.

[0043] The two-stage probability flow solution algorithm of the embodiment includes: (1) Exploration stage: Use the probability flow to introduce randomness to inspire the solution set, and heuristically explore the solution space by adjusting the probability weights. Integrate the adaptive threshold optimization strategy to optimize the misclassification problem between non-dominated solutions and dominated solutions, which can effectively balance the convergence and diversity of the algorithm and accelerate the identification and promotion of the potential optimal solution search process.

[0044] (2) Convergence stage: Introduce a selection mechanism based on the normal distribution probability idea to select the optimal solution set flow, effectively balance the weights between objectives, and avoid selecting extreme individuals on certain objectives. Consider the overall distribution of the population more comprehensively and explore the solution space more comprehensively.

[0045] The pairing selection of population individuals aims to select excellent parents for genetic operations to produce high-quality offspring. However, in high-dimensional spaces, the proportion of non-dominated solutions in the population rises rapidly, resulting in the ineffectiveness of the main selection criterion based on the dominance relationship. Therefore, the present invention introduces a probability distribution and a dynamic adjustment mechanism. The randomness is incorporated into the optimization process in an orderly and controllable manner using a probability flow. Specifically, the solution set is sorted according to performance metrics, and the candidate solutions are divided into three different flows, namely the optimal solution set flow, the mixed solution set flow, and the worst solution set flow, as Figure 3 shown. Each solution set flow is assigned different probability weights based on its performance metrics, and the probability weights are calculated by simulating the Poisson probability cumulative distribution function; and the probability weights of the solution set are dynamically adjusted to optimize the distribution and performance of the population. First, sort the solutions from the best to the worst according to performance, determine the performance boundary through the probability flow cumulative distribution function, and ensure that the optimal solution set flow contains a certain proportion of the highest-performance solutions. Then calculate the boundary of the mixed solution set flow, and the remaining solutions are automatically classified into the worst solution set flow. Dividing the solution set flow based on the probability flow is not just through random selection, but by purposefully adjusting the probability weights, effectively avoiding premature convergence of the algorithm. The probability flow weight calculation formula is as follows: ; (20) In the formula, is the performance parameter controlling the probability distribution, is a certain objective value of the th individual, is the maximum value among all individual objective values, is the total number of individuals.

[0046] After the initial population is generated, calculate the performance metrics of each individual, use the probability flow weight calculation formula, i.e., formula (20), to assign the initial probability flow weights. In each generation, according to the performance of the current population, dynamically adjust the probability flow weights to optimize the search path and distribution of the population. The specific adjustment formula is as follows: ; (21) In the formula, is the weight adjustment amount at time , is the gradient of the current weight distribution, is the time-related control parameter, represents the guiding strength coefficient of the control gradient direction for weight update, is the contribution coefficient of adjusting the random perturbation to the search diversity.

[0047] According to the weight adjustment formula, i.e., formula (21), combined with the current weight distribution gradient and the random perturbation term, dynamically adjust the probability weights of each individual, and the updated weights are used for the individual selection and mating pool construction in the next generation.

[0048] According to the probability shunting method used above, the optimal solution set flow, the mixed solution set flow, and the worst solution set flow are divided. To maximize the distribution distance between the mixed solution set flow and the worst solution set flow and avoid excessive concentration and overlap between the mixed solution set flow and the worst solution set flow, which may slow down the convergence speed of the algorithm and reduce its adaptability. To further optimize the population structure and improve the efficiency of the selection process, based on the division of the solution set flow by probability flow, this paper uses the core idea of an adaptive threshold to supervise the probability flow.

[0049] Use binary search to iteratively explore the optimal solution in the threshold domain. In each iteration, the algorithm calculates the threshold optimization variance under the current threshold division and compares it with the previous optimal variance. If the currently calculated threshold optimization variance is larger, the optimal threshold is updated. Repeat the above steps until the maximum threshold optimization variance between the solution set flows is reached. The formula for the threshold optimization variance is as follows: ; (22) In the formula, represents the threshold standard deviation, is the number of individuals, is the global mean, is the probability of the mixed solution set flow, while is the probability of the worst solution set flow,

[0050] By embedding the threshold optimization variance evaluation mechanism into the probability flow search framework, according to the criterion of maximizing the optimization objective, dynamically adjust the random perturbation intensity through the solution set variance, and adaptively adjust the search direction and step size in real time, which can effectively traverse the solution space; combined with the differential evolution method, calculate the difference vector between the current solution and other neighboring solutions, and normalize the difference vector. Guided by the difference vector, the candidate solution is perturbed and updated in each search process. During the exploration stage, a larger step size is adopted to enhance the perturbation amplitude; during the convergence stage, the step size is gradually reduced to reduce the perturbation amplitude. Determine the optimal threshold by the threshold optimization variance method, distinguish the mixed solution set flow and the worst solution set flow into different categories, and accelerate the convergence speed.

[0051] In the process of exploring high-dimensional optimization problems, the variance optimization principle and the probability flow method are the keys. By introducing randomness in the search process and dynamically adjusting the search trajectory and its step size in turn, efficient traversal of the random space is achieved. Specify the boundary threshold by accurately calculating the statistical dispersion of the solution set, and the combination of the two ensures a balanced exploration and exploitation during the iteration process.

[0052] In the convergence stage of the two-stage probability flow solution algorithm, a selection mechanism introducing the probability idea of normal distribution is adopted to select the optimal solution set flow, balance the weights between optimization objectives, and avoid selecting extreme individuals in some optimization objectives; more comprehensively consider the overall distribution of the population and explore the solution space more comprehensively.

[0053] Directly calculating the minimum vertical distance to select the dominant or non-dominant solutions is difficult to effectively filter extreme values; based on the normal distribution probability method, according to the shape of the normal distribution function, most values (solutions) are concentrated near the mean of the normal distribution function, and the occurrence frequency of extreme values far from the mean decreases rapidly, and this characteristic can be used to reduce the influence of extreme values; according to the characteristic of the rapid decay of the tail of the normal distribution, it helps to filter out atypical solutions that lead to a decline in search performance; and through the natural decay in the probability density function, a lower selection weight is given to extreme values, without the need to artificially determine the selection pressure and without additional adjustment mechanisms.

[0054] By combining the probability idea of normal distribution, a probability density function is constructed based on the vertical distance between the solution and the reference direction to perform precise probability density evaluation on the solution set, and select the optimal solution during the optimization process; calculate the mean and standard deviation of the vertical distance from the solution to the reference direction, and use the normal distribution probability density function to guide the selection process of the optimal solution, effectively balancing the weights between optimization objectives and ensuring that the optimal solution solving process will not deviate from the optimal path due to random fluctuations; it can consider both local optimized individuals and the overall population distribution, and has better adaptability and robustness.

[0055] Solution vector to the reference vector The vertical distance is: ; (23) In the formula, represents the orthogonal projection of the solution vector in the direction of the reference vector .

[0056] ; (24) ; (25) In the formula, is the mean of the vertical distance of the solution vector , is the number of solution vectors, is the th vertical distance of the solution vector; is the standard deviation of the vertical distance of the solution vector .

[0057] ; (26) In the formula, is the perpendicular distance of the solution vector is the probability density of the normal distribution.

[0058] ; (27) In the formula, is the adjustment coefficient, is the selection pressure function.

[0059] Directly using as the selection probability may lead to some solutions being overselected, thus increasing the selection pressure; To balance this, an adjustment coefficient and a selection pressure function are introduced to select the probability; finally, a threshold is used to screen the solution set, and this threshold is dynamically calculated according to the cumulative distribution function of the global selection probability; ; (28) In the formula, represents the allowable error range, which is a decimal close to 0; is the inverse function of the cumulative distribution function ; when is greater than the calculated threshold , the solution is selected as the optimal solution.

[0060] Throughout the process, the normal distribution characteristics ensure that the algorithm does not overly favor the solutions in any specific direction, nor ignore potential non - typical solutions, so as to increase the selection pressure of the Pareto - optimal solutions.

[0061] Step 4: Use the two - stage probability flow solution algorithm to solve the multi - objective optimization model, and obtain the optimal solution, that is, the optimal scheduling plan.

[0062] Step 5: According to the optimal scheduling plan obtained in Step 4, perform the operation scheduling on the integrated energy system.

[0063] In the embodiment, in order to evaluate the effectiveness of the two - stage probability flow solution algorithm of the present invention, it is compared with the PSO algorithm and the NSGA - III algorithm by means of the evolutionary multi - objective optimization platform PlatEMO. In order to comprehensively evaluate the performance of the solution algorithm of the present invention, some functions in the commonly used DTLZ test function set and WFG test function set in the field of ultra - multi - objective optimization are selected for comparative experiments.

[0064] The algorithm parameter settings involved in the experiment are as follows: population size : ; SBX crossover probability: ; crossover distribution index: ; Polynomial mutation probability: ; Mutation index: ; Probability of introducing suboptimal solutions: , and the algorithm runs independently 30 times.

[0065] To quantitatively measure the performance of the algorithm in solving problems, the Inverted Generational Distance (IGD), which is widely used in multi-objective optimization problems, is adopted. The IGD is the solution obtained by shooting from the reference point on the true PF to the algorithm. The IGD can evaluate both the convergence and diversity of the algorithm simultaneously. The calculation formula of the IGD index is as follows: ; (29) In the formula, is the set of points uniformly distributed on the true Pareto front, is the number of individuals in the set of points distributed on the true Pareto front, is the obtained optimal Pareto-optimal solution set flow, is the minimum Euclidean distance from the individuals in to the population

[0066] Table 1 IGD values of three algorithms on DTLZ and WFG

[0067] It can be seen from Table 1 that the IGD values of the two-stage probability flow algorithm on the test problems are significantly better than those of PSO and NSGA-III. The two-stage probability flow guiding strategy algorithm can effectively avoid falling into local optima. In the global exploration stage, the probability flow model is used to dynamically adjust the search direction of the population, so as to cover the search space more comprehensively; in the local development stage, the two-stage search strategy enhances the approximation ability to the Pareto front.

[0068] To consider the feasibility of the two-stage probability flow algorithm adopted in the operation optimization of the integrated energy system, it is compared with the PSO algorithm and the NSGA-III algorithm, and the operation optimization solution results are analyzed. The operation optimization results are shown in Table 2.

[0069] Table 2 Three-objective operation optimization results of three algorithms

[0070] In the optimization of integrated energy systems, total cost, carbon emissions, and electricity purchase volume are all key evaluation indicators. As can be seen from Table 2, the two-stage probability flow algorithm has significantly reduced in all three aspects compared with the traditional PSO algorithm and NSGA-III algorithm, indicating that this algorithm has significant advantages in terms of economy, can more effectively reduce the system operation cost, better optimize the environmental benefits, more efficiently utilize local energy resources, reduce the dependence on external power, and improve the energy self-sufficiency rate and stability of the system. Figure 4 The figure shows the result curve obtained after 500 iterations when the two-stage probability flow algorithm is compared with the PSO and NSGA-III algorithms. Compared with the traditional PSO and NSGA-III algorithms, the curve slope of the two-stage strategy algorithm is steeper in the initial stage, and the optimization speed in the initial stage is slightly faster, which is conducive to exploring a wider solution set space and finding the global optimal solution; in the later stage when entering the second stage, the curve of the two-stage algorithm is flatter than the other two, which is conducive to finding the local optimal solution and improving the solution efficiency.

Claims

1. An integrated energy system optimization scheduling method based on a two-stage probability flow solution algorithm is characterized by: The following steps are involved: Step 1: construct a multi-objective optimization model for an integrated energy system, wherein the optimization objectives of the multi-objective optimization model include minimizing the operating cost of the integrated energy system, minimizing the total carbon emissions, and minimizing the amount of electricity purchased by the integrated energy system; Step 2: Determine the constraints of the multi-objective optimization model; Step 3: Introduce probability flow, improve the non-dominated sorting genetic algorithm, and obtain a two-stage probability flow solution algorithm, which includes an exploration stage and a convergence stage; Step 4: using the two-stage probability flow solving algorithm to solve the multi-objective optimization model to obtain the optimal solution, i.e., the optimal scheduling solution; Step 5: According to the optimal scheduling plan obtained in step 4, the integrated energy system is scheduled for operation.

2. The integrated energy system optimization scheduling method according to claim 1 is characterized in that: The integrated energy system includes gas turbines, gas boilers, energy storage batteries, absorption chillers, electric chillers, electric boilers and thermal storage tanks; The multi-objective optimization model of the integrated energy system is: ； （1） in, The objective function for multi-objective optimization of the integrated energy system is is the annual operating cost of the integrated energy system, is the total annual carbon emissions, The amount of electricity purchased for the integrated energy system; Annual operating costs of the integrated energy system The expression is: ；（2） In the formula, is the initial purchase cost of the integrated energy system equipment, The replacement cost of the equipment after its life expires. The daily operation and maintenance costs of all equipment. is the gas usage cost for gas turbines and gas boilers, represents the cost of purchasing electricity from the power grid; Total annual carbon emissions from the integrated energy system The expression is: ；（3） In the formula, is the carbon emission factor produced by natural gas combustion per unit energy, is the heat input power of the gas turbine, is the thermal output power of the gas boiler, is the thermal efficiency of the gas boiler, is the carbon emission factor per unit of energy generated by grid electricity, The amount of electricity purchased from the grid for the energy system; t Indicates the time period sequence number; Electricity Purchased by Integrated Energy System The expression is: ；（4） In the formula, The amount of electricity purchased from the grid for the energy system.

3. The integrated energy system optimization scheduling method according to claim 2 is characterized in that: In step 2, the constraints on the power generation and heat generation of the gas turbine are: ；（5） ； （6） In the formula, The power generated by the gas turbine is is the heat production power of the gas turbine, is the gas consumption power of the gas turbine, , are the power conversion efficiency and heat conversion efficiency of the gas turbine, , They are the minimum and maximum limits of gas turbine power output, Indicates the real-time output power value of the gas turbine; The constraints of the heating power of the gas boiler are: ；（7） ； （8） In the formula, is the heating power of the gas boiler, The fuel consumption power of the gas boiler is is the heat conversion efficiency of the gas boiler, and The minimum and maximum limits for the heat generation capacity of gas boilers; The constraints of the charging power and discharging power of the energy storage battery are: ； （9） In the formula, and They are , The storage capacity of the time-slot energy storage battery system, is the self-discharge coefficient of the energy storage battery, and To charge and discharge the energy storage battery, and Charging and discharging power for energy storage batteries; The constraints on the output cooling power of the absorption chiller are: ；（10） ；（11） ；（12） In the formula, is the output cooling power of the absorption chiller, is the efficiency of the absorption chiller, is the input heat efficiency of the absorption chiller, is the cooling load of the absorption chiller, is the maximum cooling load of the absorption chiller, is the cooling load required by the system.

4. The integrated energy system optimization scheduling method according to claim 3 is characterized in that: In step 2, the refrigeration power output of the electric refrigerator is: ； （13） In the formula, is the cooling power output by the electric refrigerator, is the input electrical power, is the refrigeration efficiency; The heating power of the electric boiler is: ；（14） ； （15） In the formula, is the actual heating power of the electric boiler, is the heating efficiency of the electric boiler, The power consumption of electric boiler; The heat storage power and heat release power of the heat storage tank are: ； （16） In the formula, and They are Moment and The heat storage tank's heat storage capacity at all times, is the heat loss coefficient of the heat storage tank, and are the heat storage and heat release powers of the heat storage tank, respectively. and are the heat storage and heat release efficiencies of the thermal storage tank respectively.

5. The integrated energy system optimization scheduling method according to claim 4 is characterized in that: In step 2, the constraints of the integrated energy system also include the power balance of electrical energy, thermal energy and cold energy: The electric energy balance equation is: ；（17） The heat balance equation is: ；（18） The cold energy balance equation is: ；（19） In the formula, Electricity provided by wind power, The electricity provided by photovoltaic , and They are respectively the electrical load, thermal load and cooling load transmitted by the system to the user side.

6. The integrated energy system optimization scheduling method according to claim 1 is characterized in that: In step 3, the two-stage probability flow solving algorithm specifically includes: 1) Exploration stage: Use probability flow to introduce randomness to conduct heuristic search for the solution set of multi-objective optimization problems, and heuristically explore the solution space by adjusting the probability weight of the solution set flow; Adopting an adaptive threshold optimization strategy to optimize the misclassification problem between non-dominated solutions and dominated solutions can effectively balance the convergence and diversity of the algorithm, and accelerate the search process of identifying and advancing potential optimal solutions. 2) Convergence stage: Introduce the selection mechanism based on the normal distribution probability idea to select the optimal solution set, effectively balance the weights between optimization objectives, avoid selecting extreme individuals on certain objectives, and explore the solution space more comprehensively.

7. The integrated energy system optimization scheduling method according to claim 6 is characterized in that: In the exploration phase of the two-stage probability flow solving algorithm, the solution set is sorted according to the performance index, and the candidate solutions are divided into three types of solution set flows, namely the optimal solution set flow, the mixed solution set flow and the worst solution set flow; each solution set flow is assigned a different probability weight based on its performance index, and the probability weight is calculated by simulating the Poisson probability cumulative distribution function; and the probability weight of the solution set is dynamically adjusted to optimize the distribution and performance of the population; First, the solutions are sorted from best to worst according to their performance, and the performance boundary is determined according to the cumulative distribution function of the probability flow to ensure that the optimal solution set contains a certain proportion of the highest performance solutions; then the mixed solution set boundary is calculated to divide the mixed solution set; the remaining solutions are divided into the worst solution set; Dividing the solution set flow based on probability flow is not only to randomly select the solution set, but also to purposefully adjust the probability weight of the solution to avoid premature convergence of the solution algorithm; The calculation formula of the probability weight of the probability flow is: ； （20） In the formula, Indicates The probability flow weight of individuals in a population, is the performance parameter that controls the probability distribution, It is The target value of each individual in the population, is the maximum value among all individual target values, is the total number of individuals; After the initial population is generated, the performance index of each individual in the population is calculated, and the initial probability weight is assigned using the probability flow weight calculation formula (Equation (20)). In each generation of the population, the probability flow weight is dynamically adjusted according to the performance of the current population to optimize the search path and distribution of the population. The formula for dynamically adjusting the probability flow weight is: ；（21） In the formula, express The weight adjustment amount at the moment, is the gradient of the current weight distribution, is a random disturbance term, which obeys Gaussian distribution. Represents the guiding strength coefficient that controls the gradient direction for weight update, is the contribution coefficient of regulating random disturbance to search diversity; According to formula (21), the probability weight of each individual in the population is dynamically adjusted in combination with the gradient of the current weight distribution and the random disturbance term. The updated probability weight is used for individual selection and mating pool construction in the next generation.

8. The integrated energy system optimization scheduling method according to claim 7 is characterized in that: In the exploration phase of the two-stage probability flow solution algorithm, after the solution set is divided into the optimal solution set flow, the mixed solution set flow and the worst solution set flow, in order to maximize the distribution distance between the mixed solution set flow and the worst solution set flow, to avoid excessive overlap between the mixed solution set flow and the worst solution set flow, which would slow down the convergence speed of the solution algorithm and reduce its adaptability; in order to further optimize the population structure and improve the efficiency of the selection process, on the basis of the probability flow division solution set flow, an adaptive threshold supervision probability flow mechanism is adopted; The optimal solution is explored using the threshold domain of the binary search iterative algorithm. In each iteration, the solution algorithm calculates the threshold optimization variance under the current threshold partition and compares it with the previous optimal variance. If the currently calculated threshold optimization variance is larger, the optimal threshold is updated. The above steps are repeated until the threshold optimization variance between the solution sets is maximized. The threshold optimization variance is calculated as: ； （22） In the formula, represents the threshold standard deviation, is the number of individuals, is the global mean, is the probability of mixed solution flow, Then it is the worst solution flow probability; represents the solution flow adjustment coefficient; By embedding the threshold optimization variance evaluation mechanism into the probability flow search framework, the random perturbation intensity is dynamically adjusted based on the solution set variance according to the criterion of maximizing the optimization goal, and the search direction and step size are adaptively adjusted in real time, which can effectively traverse the solution space; combined with the differential evolution method, the difference vector between the current solution and other neighborhood solutions is calculated, and the difference vector is normalized; Guided by the difference vector, the candidate solution is perturbed and updated in each iteration; a larger step size is used in the exploration phase , increase the disturbance amplitude; in the convergence stage, gradually reduce the step size , reduce the disturbance amplitude; determine the optimal threshold by optimizing the variance of the threshold, distinguish the mixed solution flow and the worst solution flow into different categories, and speed up the convergence.

9. The integrated energy system optimization scheduling method according to claim 8, characterized in that: In the convergence phase of the two-stage probability flow solution algorithm, a probability density function is constructed based on the vertical distance between the solution and the reference direction, and the solution set is accurately evaluated for probability density, so as to select the optimal solution during the optimization process. The mean and standard deviation of the vertical distance from the solution to the reference direction are calculated, and the normal distribution probability density function is used to guide the selection process of the optimal solution, effectively balancing the weights between the optimization objectives, and ensuring that the optimal solution solution process will not deviate from the optimal path due to random fluctuations. Solution Vector To the reference vector Vertical distance for: ； （23） In the formula, Represents the solution vector In the reference vector Orthogonal projection in the direction of ; ；（24） ； （25） In the formula, is the vertical distance The average value of is the number of solution vectors, For the The vertical distance between the solution vectors; is the perpendicular distance of the solution vector The standard deviation of ； （26） In the formula, is the vertical distance The normal distribution probability density of ; ；（27） In the formula, is the adjustment coefficient, is the selected pressure function.

10. The integrated energy system optimization scheduling method according to claim 9, characterized in that: In the convergence phase of the two-stage probability flow solving algorithm, the threshold Filter the solution set, threshold Cumulative distribution function based on global selection probability Dynamic calculation, the calculation formula is: ；（28） In the formula, Indicates the error range; is the cumulative distribution function The inverse function of Greater than the calculated threshold When Selected as the optimal solution.

Images