Cloud energy storage scheduling optimization algorithm based on differential evolution

Abstract

The invention discloses a cloud energy storage scheduling optimization algorithm based on differential evolution, which has global search capability: a wide solution space is effectively explored by utilizing the global search characteristic of the differential evolution algorithm. Through the steps of population initialization, variation, crossover, selection and the like, local optimum is avoided, and a better solution is found; adaptive parameter adjustment: a differential evolution algorithm generally relates to a small number of key control parameters, such as a variation factor (F) and a crossover probability (CR). According to the method, the parameters are self-adaptively or dynamically adjusted, so that the robustness and efficiency of the algorithm are enhanced, the algorithm can be well expressed in different situations, and the method relates to the technical field of cloud energy storage. The cloud energy storage scheduling optimization algorithm based on differential evolution has an efficient constraint processing mechanism, and the cloud energy storage scheduling problem relates to various complex constraints, such as power balance and energy storage capacity limitation. The invention provides an innovative method to effectively process the constraints, all solutions are ensured to be feasible, and actual application requirements are met.

Description

Technical Field

[0001] The present invention relates to the field of cloud energy storage technology, and in particular to a cloud energy storage scheduling optimization algorithm based on differential evolution. Background Art

[0002] With the rapid development of new power systems, the widespread application of distributed energy (such as solar energy and wind energy) has become an important part of energy transformation. The scale of distributed energy access continues to expand, which poses new challenges to the stable operation of the power system and the efficient use of energy. The randomness and intermittent characteristics of distributed energy have brought difficulties to the dispatching and management of the power system, especially when ensuring the balance of power supply and demand, it is also necessary to consider how to maximize the use of these clean and renewable energy sources. In order to meet this challenge, energy storage technology has become one of the indispensable solutions.

[0003] Energy storage systems can store energy when there is an oversupply of electricity and release energy when there is a power shortage, thereby improving energy efficiency and ensuring the stable operation of the power system. Although traditional energy storage technologies, such as battery energy storage systems, have been used to a certain extent, their high cost and limited flexibility limit their widespread application in large-scale power systems. In recent years, with the continuous development of cloud computing and big data technologies, cloud energy storage has received widespread attention as an emerging energy storage technology.

[0004] Cloud energy storage technology can not only serve as a centralized or distributed energy storage resource, but also provide users with on-demand real-time energy storage services through a sharing economy model. Through cloud energy storage technology, efficient utilization of energy storage resources can be achieved, energy storage costs can be reduced, energy utilization efficiency can be improved, and the operation mode and structure of the energy storage system can be simplified. However, to achieve the effective application of cloud energy storage technology in microgrids, it is also necessary to solve the problem of how to perform efficient scheduling optimization. Traditional optimization algorithms may not meet the needs of cloud energy storage scheduling optimization because they may easily fall into local optimal solutions or perform poorly when dealing with high-dimensional optimization problems. Therefore, an optimization algorithm with strong global search capabilities and easy-to-adjust parameters is needed to solve these problems.

[0005] Differential Evolution (DE) is an efficient global optimization algorithm with strong global search capabilities, which can avoid falling into local optimal solutions, and has relatively few parameters and is easy to adjust. The DE algorithm is particularly suitable for solving high-dimensional optimization problems, especially for solving large-scale optimization problems. Therefore, applying the differential evolution algorithm to the scheduling optimization of cloud energy storage can effectively improve the utilization rate of distributed energy, reduce the operating cost of the power system, and ensure the stable operation of the power system.

[0006] With the development of energy internet and the widespread application of distributed energy, the demand for efficient and flexible energy storage technology is increasing. As an emerging energy storage technology, cloud energy storage, combined with differential evolution algorithm for scheduling optimization, provides a feasible solution to the challenges brought by distributed energy access.

[0007] Reference document 1: A multi-objective driven microgrid group cloud energy storage optimization configuration method and device with publication number CN116436048A, the method comprising: based on the economic optimization objective function, the low-carbon optimization objective function and the reliability optimization objective function, after the investment decision constraint condition, constructing the microgrid group cloud energy storage optimization configuration planning layer model; constructing the microgrid group cloud energy storage optimization configuration scheduling layer model; using the second generation non-dominated genetic algorithm, iteratively solving the microgrid group cloud energy storage optimization configuration planning layer model and the microgrid group cloud energy storage optimization configuration scheduling layer model to generate a Pareto solution set; performing economic evaluation, low-carbon evaluation and reliability evaluation on the Pareto solution set respectively, and determining the microgrid group cloud energy storage optimization configuration scheme based on the evaluation results, and the microgrid group cloud energy storage optimization configuration scheme is used for multi-objective driven microgrid operation. The method comprehensively considers the economic, low-carbon and reliability benefits, and realizes the optimal configuration of the microgrid group cloud energy storage system capacity.

[0008] Reference document 2: A distributed energy storage resource multi-time and space scale aggregation method and system with publication number CN115330186A, the present invention includes spatial division of users and distributed energy storage resources based on geographical areas as the spatial scale, and determining users and distributed energy storage resources in each geographical area, and supplying power to users in the same geographical area through energy storage resources; time division based on power consumption time period as the time scale, based on the user's power consumption data, determining high power users and low power users in each power consumption time period; when it is necessary to dispatch power consumption for a certain power consumption time period, the surplus power provided by the energy storage resources to low power users is transferred to high power users to achieve power complementarity between high power users and low power users. The present invention can solve the problem of large amounts of electricity in existing cloud energy storage that is not used but has to be retained, resulting in resource waste, improve the utilization efficiency of distributed energy storage resources, and play a role in saving energy and reducing costs.

[0009] Reference document 3: A method for sharing cloud energy storage in a microgrid based on master-slave game, with publication number CN117852709A, which constructs a microgrid operator revenue model, a user aggregator revenue model, and a cloud energy storage service provider revenue model respectively, and establishes a master-slave game model for the above models, and then performs genetic iterative calculations on the master-slave game model to complete the sharing. The method for sharing cloud energy storage in a microgrid based on master-slave game models the relationship between microgrid operators, user aggregators, and cloud energy storage service providers as a master-slave game model, taking into account the interactions and conflicts between the parties; there is a price and load game between the microgrid operator and the user aggregator to achieve the optimal operation of the entire system, and uses a solver to locally optimize the load distribution of the user aggregator and the rental price of the cloud energy storage service provider, which helps to improve the convergence speed and overall performance of the system, while taking into account the characteristics of decision variables at different levels.

[0010] Reference document 4: A cloud energy storage flexible control method with publication number CN117689226A that considers the efficient absorption of distributed photovoltaic power in distribution networks, specifically: 1) Establishing an upper-level optimization model; taking the minimum daily operating cost of the centralized energy storage power station and the regional microgrid of the distribution network as the optimization goal, different centralized energy storage power station electricity sales and purchase prices are decided for different types of microgrids; 2) Establishing a lower-level optimization model; based on the electricity sales and purchase prices of the centralized energy storage power station determined by the upper-level optimization model, taking the minimum operating cost of the regional microgrid as the optimization goal, decide the electric power that the regional microgrid needs to purchase from the energy storage power station and the power grid, and reasonably arrange the power and time of energy storage charging and discharging, so as to ensure that the distributed photovoltaic power in the regional microgrid is efficiently absorbed while reducing the electricity cost of the microgrid; the method is based on the cloud energy storage management strategy, establishes a two-layer optimization model of multi-regional microgrids, and realizes efficient absorption of multi-regional distributed photovoltaic power in active distribution networks under the cloud energy storage mode.

[0011] Reference document 5: The cloud energy storage system access distribution network coordinated optimization scheduling method with publication number CN117713242A establishes a multi-agent interactive structure of the distribution network-cloud energy storage system, and establishes the upper model and lower model of the two-layer optimization scheduling model of the distribution network-cloud energy storage system by collaboratively modeling the distribution network, cloud energy storage system and cloud energy storage users. The upper model takes the lowest operating cost of the distribution network as the goal, and solves to obtain the optimal output plan of each power source in the distribution network and the charging and discharging plan of the cloud energy storage system. The lower model takes the maximum profit of the cloud energy storage system and formulates the charging and discharging plan of the cloud energy storage user; the distribution network-cloud energy storage two-layer optimization scheduling model is solved by the iterative method to obtain the optimized scheduling strategy. The present invention realizes mutual benefit and win-win results among the distribution network, cloud energy storage operators and cloud energy storage users, and can effectively reduce the operating cost of the distribution network, reduce the peak-to-valley difference of the load curve, and improve the economy and stability of the system operation.

[0012] Reference 1 specifically includes the following technical defects:

[0013] Computational efficiency: Using the second generation non-dominated genetic algorithm (NSGA-II) for multi-objective optimization may result in long computational time, especially when dealing with large-scale microgrid groups and complex constraints. NSGA-II requires a large population size and a large number of iterations to converge to a satisfactory solution, which may lead to a large consumption of computing resources, while the differential evolution algorithm (DE) can usually obtain a good solution in fewer iterations.

[0014] Model complexity: The optimization model proposed in the patent contains multiple objective functions and complex constraints, which increases the complexity of the model and may cause problems in understanding and implementation. For practical applications, overly complex models may limit their application and promotion in the industry, while differential evolution algorithms are easier to understand and accept due to their simplicity.

[0015] Parameter sensitivity: Genetic algorithms (including NSGA-II) usually contain some parameters that need to be set manually, such as population size, crossover probability, mutation probability, etc. The selection of these parameters is crucial to the performance of the algorithm, and improper selection may lead to poor algorithm performance, such as premature convergence or failure to find a good solution. In contrast, differential evolution algorithms usually involve fewer key parameters (such as mutation factor F and crossover probability CR), and these parameters can be better adjusted through adaptive methods, thereby reducing the difficulty of parameter selection.

[0016] Reference 2 specifically includes the following technical defects:

[0017] Lack of global optimization capability: The method mentioned in the patent focuses on local optimization through the division of geographical areas and time scales, that is, by complementing the power of low-power users with high-power users in a specific time period. This method may not be able to fully explore the global optimal solution, especially when dealing with complex multi-objective optimization problems, and may not be able to find the global optimal or near-optimal solution. The differential evolution algorithm has a strong global search capability and can explore the solution space more effectively to avoid falling into the local optimum.

[0018] Limited flexibility: This method relies on predefined spatial and temporal division rules and may not be able to flexibly adapt to changing electricity supply and demand. For example, if the electricity demand in a certain area changes suddenly, this method may not be able to respond quickly, while the differential evolution algorithm can better adapt to the dynamically changing environment through iteration and adaptive parameter adjustment.

[0019] Lack of scalability: This method may encounter challenges when dealing with larger-scale distributed energy storage resource networks. As the number of geographical areas and energy storage resources increases, it becomes increasingly difficult to manually divide time and space and perform scheduling. In contrast, differential evolution algorithms are able to handle higher-dimensional optimization problems and automatically discover better scheduling strategies through population search.

[0020] Reference 3 specifically includes the following technical defects:

[0021] Computational efficiency: The master-slave game model usually involves a multi-layer decision-making process, and this structure may lead to a long computation time. In this patent, iterative calculations are required through genetic algorithms to solve the master-slave game model, which may lead to a large consumption of computing resources, especially when dealing with large-scale microgrids and complex constraints.

[0022] Model complexity: The model proposed in the patent involves interactions between multiple participants, and it is necessary to build their own profit models and build a master-slave game model on this basis. This multi-level modeling approach increases the overall complexity of the model and may lead to problems with understanding and implementation. In addition, the complexity of the model may also limit its widespread promotion in the industry.

[0023] Difficulty in parameter adjustment: The genetic algorithm in this method requires setting a series of parameters, such as population size, crossover probability, mutation probability, etc. The selection of these parameters is crucial to the performance of the algorithm, and improper selection may lead to poor algorithm performance, such as premature convergence or failure to find a good solution. Compared with differential evolution algorithms, which usually involve fewer key parameters (such as mutation factor F and crossover probability CR) and can be better adjusted through adaptive methods, parameter adjustment in the master-slave game model may be more difficult.

[0024] Reference 4 specifically includes the following technical defects:

[0025] Computational efficiency: This method uses a two-layer optimization model, in which the upper optimization model is used to decide the electricity sales and purchase prices of centralized energy storage power stations, while the lower optimization model is used to decide the power purchase power and energy storage charging and discharging power of regional microgrids. This hierarchical structure may lead to long computational times, especially when dealing with large-scale microgrids and complex constraints. Compared with the differential evolution algorithm, this method may require more computing resources to obtain satisfactory results.

[0026] Model complexity: This method requires the construction of two optimization models, and there is a mutual dependence between the two models. This multi-level modeling approach increases the overall complexity of the model and may lead to problems that are difficult to understand and implement. In addition, the complexity of the model may also limit its widespread promotion in the industry.

[0027] Difficulty in parameter adjustment: The optimization model in this method may involve multiple parameters, such as the electricity sales and purchase prices of centralized energy storage power stations. The selection of these parameters is crucial to the performance of the model, and improper selection may lead to poor model performance, such as premature convergence or failure to find a good solution. Compared with differential evolution algorithms, which usually involve fewer key parameters and can be better adjusted through adaptive methods, parameter adjustment in this method may be more difficult.

[0028] Reference 5 specifically includes the following technical defects:

[0029] Computational efficiency: This method adopts a two-layer optimization model, in which the upper model aims to minimize the operating cost of the distribution network, while the lower model aims to maximize the profit of the cloud energy storage system. This hierarchical structure may lead to long computational time, especially when dealing with large-scale distribution networks and complex constraints. Compared with the differential evolution algorithm, this method may require more computing resources to obtain satisfactory results.

[0030] Model complexity: This method requires the construction of two optimization models, and there is a mutual dependence between the two models. This multi-level modeling approach increases the overall complexity of the model and may lead to problems that are difficult to understand and implement. In addition, the complexity of the model may also limit its widespread promotion in the industry.

[0031] Difficulty in parameter adjustment: The optimization model in this method may involve multiple parameters, such as the optimal output plan of each power source in the distribution network and the charging and discharging plan of the cloud energy storage system. The selection of these parameters is crucial to the performance of the model, and improper selection may lead to poor model performance, such as premature convergence or failure to find a good solution. Compared with differential evolution algorithms, which usually involve fewer key parameters and can be better adjusted by adaptive methods, parameter adjustment in this method may be more difficult. Summary of the invention

[0032] 1. Technical issues to be resolved

[0033] In view of the shortcomings of the prior art, the present invention provides a cloud energy storage scheduling optimization algorithm based on differential evolution, which covers global search, adaptive parameter adjustment, efficient constraint processing and multi-objective optimization capabilities, together forming a powerful and practical cloud energy storage scheduling optimization algorithm.

[0034] (II) Technical solution

[0035] To achieve the above objectives, the present invention is implemented through the following technical solutions: a cloud energy storage scheduling optimization algorithm based on differential evolution, specifically comprising the following steps:

[0036] S1. Optimization goal: Optimize cost expenditure through effective scheduling of distributed power sources, intelligent charging and discharging management of energy storage systems, and analysis of energy exchange with external power grids;

[0037] S2. Objective function: Define an objective function to comprehensively consider various costs, with the goal of minimizing the overall economic cost. The objective function can be expressed as:

[0038]

[0039] Where: C D (t): depreciation cost at time t;

[0040] C O (t): operation and maintenance cost at time t;

[0041] C G (t): is the electricity purchase cost at time t;

[0042] S3, cost calculation model processing;

[0043] S4. Constraint analysis: including two constraints: power balance constraint and energy storage constraint;

[0044] S5. Differential evolution algorithm processing: In order to cope with the possible limitations of traditional optimization algorithms, the differential evolution algorithm (DE) is introduced into cloud energy storage scheduling optimization because it has powerful global search capabilities and easy-to-adjust parameters. The following is a detailed description of the algorithm. The differential evolution algorithm includes objective functions, cost models, constraints, and optimization combined with these elements, specifically including the following steps:

[0045] T1. Population initialization: Randomly generate an initial population P0 containing N individuals, each individual x i Represents a possible scheduling scheme, which is a vector containing the output power P of each distributed energy unit at each time step t. i (t) and the charging and discharging power P of the energy storage system S (t), x i The expression is as follows:

[0046]

[0047] Among them, P i,k (t) represents the output power of the kth distributed energy unit of the ith individual at time step t, and P S (t) represents the charging and discharging power of the energy storage system at time step t;

[0048] T2, mutation: usually affects individual xi Make some kind of change to produce a variant individual v i The mutation operation can be simple, such as randomly changing one or some gene values, or complex, such as using a certain probability distribution to determine the new gene value. Suppose a simple mutation operation is used, that is, i Each gene g in the random increase or decrease of a small value Δ g , this value is drawn from a specific probability distribution, normal distribution or uniform distribution, and the mutation operation can be defined as:

[0049] v i =x i +Δ;

[0050] where Δ is a i A vector of the same dimension, each element of which is Δ g Independently extract from a probability distribution, which is adjusted according to actual needs, for example, a normal distribution with a mean of 0 can be selected;

[0051] T3. Crossover: The crossover operation is used to combine the information of two or more individuals to generate a new individual. In the differential evolution algorithm, the crossover operation is often used to mix the information of the target vector and the mutation vector to generate a new test vector u i ;

[0052] T4, Selection: The selection operation is used to determine which individual (target vector or test vector) will enter the next generation population. In the differential evolution algorithm, it is as follows:

[0053] e1, for each target vector x i and the corresponding test vector u i , calculate their fitness values ​​f(x i ) and f(u i );

[0054] e2. If the test vector u i The fitness value is better than the target vector x i The fitness value, that is, f(u i )<f(x i ), then the test vector u i Replace the target vector x i Become part of the next generation population;

[0055] e3. Otherwise, keep the target vector x i In the next generation population. This process can be expressed as:

[0056]

[0057] Here, g represents the number of iterations. In each generation (each iteration), the algorithm will perform mutation, crossover and selection operations on the population, thereby gradually improving the fitness of the population and eventually converging to the global optimal solution or the local optimal solution.

[0058] T5. Constraint processing: In the cloud energy storage scheduling optimization problem, there may be various constraints, such as power balance constraints, energy storage system capacity constraints, etc. These constraints are added to the fitness calculation through the penalty function, or the individuals that do not meet the constraints are directly modified through the repair mechanism. The penalty function is used if the individual x i If the constraint condition is not met, a penalty term p(x i ):

[0059] f(x i ) = f original (x i )-p(x i );

[0060] Among them, f original (x i ) is the fitness value without considering the constraints;

[0061] T6. Iteration: The genetic algorithm will go through multiple iteration cycles until a certain termination condition is reached. Each iteration cycle includes steps such as fitness evaluation, selection, crossover, and mutation. The iteration process can be expressed in pseudo code as follows:

[0062] p1, initialize the population P0;

[0063] p2. Repeat until the termination condition is met (such as the maximum number of iterations, fitness threshold, etc.):

[0064] p3. Output the optimal individual: In the cloud energy storage scheduling optimization problem, define an objective function f(x) to comprehensively consider various costs, with the goal of minimizing the overall economic cost.

[0065] S6. Through the differential evolution algorithm process of step S5, individuals in the population are continuously iterated and optimized to find the optimal scheduling solution that can achieve the lowest economic cost. The optimal individual finally output is a detailed scheduling strategy, which includes the charging and discharging status, charging power, discharging power and remaining capacity of the energy storage system at each scheduling moment, so as to ensure that the overall economic cost is minimized under the premise of meeting all constraints.

[0066] Preferably, the cost calculation model processing in step S3 specifically includes the following steps:

[0067] a1. Depreciation cost analysis:

[0068] Depreciation costs are usually calculated using the equal annual value method, expressed as:

[0069]

[0070] Where: P i (t): is the output power of the i-th micro power source at time t;

[0071] r i : is the fixed annual interest rate of the i-th micro power source;

[0072] L: investment repayment period;

[0073] a2. Operation and maintenance cost analysis:

[0074] The operation and maintenance costs can be expressed as:

[0075]

[0076] Where: α i : Operation and maintenance cost coefficient of the i-th micro power source unit output;

[0077] N: is the number of micro power sources;

[0078] P i (t): is the output power of the i-th micro power source at time t;

[0079] a3. Analysis of electricity purchase cost: The electricity purchase cost is:

[0080] C G (t) = C P (t)*P G (t);

[0081] Where: C P (t): the electricity purchase price at time t;

[0082] P G (t): is the purchased electricity power at time t.

[0083] Preferably, the power balance constraint in step S4 ensures that at each time point t, the total output power in the microgrid matches the load:

[0084]

[0085] Where: L(t): is the load at time t;

[0086] P S (t): is the charging and discharging power of the energy storage system (discharging is positive, charging is negative).

[0087] Preferably, the energy storage constraint in step S4 ensures that the charging and discharging process of the energy storage system does not exceed its capacity limit:

[0088] E min ≤E(t)≤E max ;

[0089] Where: E(t) is the capacity of the energy storage system at time (t);

[0090] E min and E max are the minimum and maximum capacities of the energy storage system respectively.

[0091] Preferably, in step T1, if it is considered that there are 24 time steps in a day, each individual may be a number of distributed energy units with a length of 24×(n+1), plus an energy storage system.

[0092] Preferably, a specific mutation operation step in step T2 is:

[0093] b1. For each time step t and each distributed energy unit k, a variation value Δ is drawn from a distribution k,t ;

[0094] b2. For each time step t, extract a variation value Δ of the energy storage system from a distribution S,t ;

[0095] b3. Add the mutation value to the original value to generate the mutation individual v i , the specific formula is as follows:

[0096]

[0097] b4. Assume that there are 3 distributed energy units and 1 energy storage system. Each time step t extracts a variation value from a normal distribution with a mean of 0 and a standard deviation of 10 kW. Then, the variation individual v i is written as:

[0098]

[0099] Among them, the variation value Δ k,t and Δ S,t By independently drawing from a chosen probability distribution, this mutation operation helps explore the solution space and increase population diversity.

[0100] Preferably, the crossover operation in step T3 is performed for each target vector x i , using the mutation vector v i To generate the test vector u i , the crossover operation can be defined as:

[0101]

[0102] Where: rand j is a random number in the interval [0,1], CR is the crossover probability (usually set to a value between 0 and 1), j rand is an index randomly selected in the range [1,D], ensuring that at least one gene is selected from the mutation vector v i Inherited from , D is the dimension that usually represents the decision variable, that is, the length of each individual (vector). In this particular problem, each individual represents a scheduling plan, which includes the output power of each distributed energy unit at each time step and the charging and discharging power of the energy storage system.

[0103] Preferably, step p2 is repeated until the termination condition is met as follows:

[0104] c1. For each individual x i ∈P0;

[0105] c2. Select individuals for reproduction of the next generation;

[0106] c3. Perform crossover and mutation operations on the selected individuals to generate new individuals;

[0107] c4. Update population P.

[0108] Cloud energy storage technology can not only serve as a centralized or distributed energy storage resource, but also provide users with on-demand real-time energy storage services through a sharing economy model. Through cloud energy storage technology, efficient utilization of energy storage resources can be achieved, energy storage costs can be reduced, energy utilization efficiency can be improved, and the operation mode and structure of the energy storage system can be simplified. However, to achieve the effective application of cloud energy storage technology in microgrids, it is also necessary to solve the problem of how to perform efficient scheduling optimization. Traditional optimization algorithms may not meet the needs of cloud energy storage scheduling optimization because they may easily fall into local optimal solutions or perform poorly when dealing with high-dimensional optimization problems. Therefore, an optimization algorithm with strong global search capabilities and easy-to-adjust parameters is needed to solve these problems.

[0109] Differential Evolution (DE) is an efficient global optimization algorithm with strong global search capabilities, which can avoid falling into local optimal solutions, and has relatively few parameters and is easy to adjust. The DE algorithm is particularly suitable for solving high-dimensional optimization problems, especially for solving large-scale optimization problems. Therefore, applying the differential evolution algorithm to the scheduling optimization of cloud energy storage can effectively improve the utilization rate of distributed energy, reduce the operating cost of the power system, and ensure the stable operation of the power system.

[0110] (III) Beneficial effects

[0111] The present invention provides a cloud energy storage scheduling optimization algorithm based on differential evolution. Compared with the prior art, it has the following beneficial effects:

[0112] (1) The cloud energy storage scheduling optimization algorithm based on differential evolution has global search capabilities: the global search characteristics of the differential evolution algorithm are used to effectively explore a vast solution space. Through steps such as population initialization, mutation, crossover and selection, it avoids falling into local optimality and seeks a better solution; adaptive parameter adjustment: differential evolution algorithms usually involve a small number of key control parameters, such as mutation factor (F) and crossover probability (CR). The present invention enhances the robustness and efficiency of the algorithm by adaptively or dynamically adjusting these parameters, so that the algorithm can perform well in different scenarios.

[0113] (2) The cloud energy storage scheduling optimization algorithm based on differential evolution has an efficient constraint processing mechanism: The cloud energy storage scheduling problem involves a variety of complex constraints, such as power balance and energy storage capacity limitations. This invention proposes an innovative method to effectively handle these constraints, ensuring that all solutions are feasible and meet the needs of practical applications.

[0114] (3) The cloud energy storage scheduling optimization algorithm based on differential evolution has multi-objective optimization capabilities: In addition to the main goal of minimizing economic costs, the algorithm also considers other important factors. By achieving multi-objective optimization (such as system reliability, environmental impact, etc.), the algorithm provides more comprehensive and practical optimization results, helping decision makers make more informed choices. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 This is a block diagram of the cloud energy storage scheduling optimization algorithm based on differential evolution of the present invention. DETAILED DESCRIPTION

[0116] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0117] See also Figure 1 The embodiment of the present invention provides two technical solutions: a cloud energy storage scheduling optimization algorithm based on differential evolution, which specifically includes the following embodiments:

[0118] Example 1: A cloud energy storage scheduling optimization algorithm based on differential evolution, specifically comprising the following steps:

[0119] S1. Optimization goal: Optimize cost expenditure through effective scheduling of distributed power sources, intelligent charging and discharging management of energy storage systems, and analysis of energy exchange with external power grids;

[0120] S2. Objective function: Define an objective function to comprehensively consider various costs, with the goal of minimizing the overall economic cost. The objective function can be expressed as:

[0121]

[0122] Where: C D (t): depreciation cost at time t;

[0123] C O (t): operation and maintenance cost at time t;

[0124] C G (t): is the electricity purchase cost at time t;

[0125] S3, cost calculation model processing;

[0126] S4. Constraint analysis: including two constraints: power balance constraint and energy storage constraint;

[0127] S5. Differential evolution algorithm processing: In order to cope with the possible limitations of traditional optimization algorithms, the differential evolution algorithm (DE) is introduced into cloud energy storage scheduling optimization because it has powerful global search capabilities and easy-to-adjust parameters. The following is a detailed description of the algorithm. The differential evolution algorithm includes objective functions, cost models, constraints, and optimization combined with these elements, specifically including the following steps:

[0128] T1. Population initialization: Randomly generate an initial population P0 containing N individuals, each individual x i Represents a possible scheduling scheme, which is a vector containing the output power P of each distributed energy unit at each time step t. i (t) and the charging and discharging power P of the energy storage system S (t), x i The expression is as follows:

[0129]

[0130] Among them, P i,k (t) represents the output power of the kth distributed energy unit of the ith individual at time step t, and P S (t) represents the charging and discharging power of the energy storage system at time step t;

[0131] T2, mutation: usually affects individual xi Make some kind of change to produce a variant individual v i The mutation operation can be simple, such as randomly changing one or some gene values, or complex, such as using a certain probability distribution to determine the new gene value. Suppose a simple mutation operation is used, that is, i Each gene g in the random increase or decrease of a small value Δ g , this value is drawn from a specific probability distribution, normal distribution or uniform distribution, and the mutation operation can be defined as:

[0132] v i =x i +Δ;

[0133] where Δ is a i A vector of the same dimension, each element of which is Δ g Independently extract from a probability distribution, which is adjusted according to actual needs, for example, a normal distribution with a mean of 0 can be selected;

[0134] T3. Crossover: The crossover operation is used to combine the information of two or more individuals to generate a new individual. In the differential evolution algorithm, the crossover operation is often used to mix the information of the target vector and the mutation vector to generate a new test vector u i ;

[0135] T4, Selection: The selection operation is used to determine which individual (target vector or test vector) will enter the next generation population. In the differential evolution algorithm, it is as follows:

[0136] e1, for each target vector x i and the corresponding test vector u i , calculate their fitness values ​​f(x i ) and f(u i );

[0137] e2. If the test vector u i The fitness value is better than the target vector x i The fitness value, that is, f(u i )<f(x i ), then the test vector u i Replace the target vector x i Become part of the next generation population;

[0138] e3. Otherwise, keep the target vector x i In the next generation population. This process can be expressed as:

[0139]

[0140] Here, g represents the number of iterations. In each generation (each iteration), the algorithm will perform mutation, crossover and selection operations on the population, thereby gradually improving the fitness of the population and eventually converging to the global optimal solution or the local optimal solution.

[0141] T5. Constraint processing: In the cloud energy storage scheduling optimization problem, there may be various constraints, such as power balance constraints, energy storage system capacity constraints, etc. These constraints are added to the fitness calculation through the penalty function, or the individuals that do not meet the constraints are directly modified through the repair mechanism. The penalty function is used if the individual x i If the constraint condition is not met, a penalty term p(x i ):

[0142] f(x i ) = f original (x i )-p(x i );

[0143] Among them, f original (x i ) is the fitness value without considering the constraints;

[0144] T6. Iteration: The genetic algorithm will go through multiple iteration cycles until a certain termination condition is reached. Each iteration cycle includes steps such as fitness evaluation, selection, crossover, and mutation. The iteration process can be expressed in pseudo code as follows:

[0145] p1, initialize the population P0;

[0146] p2. Repeat until the termination condition is met (such as the maximum number of iterations, fitness threshold, etc.):

[0147] p3. Output the optimal individual: In the cloud energy storage scheduling optimization problem, define an objective function f(x) to comprehensively consider various costs, with the goal of minimizing the overall economic cost.

[0148] S6. Through the differential evolution algorithm process of step S5, individuals in the population are continuously iterated and optimized to find the optimal scheduling solution that can achieve the lowest economic cost. The optimal individual finally output is a detailed scheduling strategy, which includes the charging and discharging status, charging power, discharging power and remaining capacity of the energy storage system at each scheduling moment, so as to ensure that the overall economic cost is minimized under the premise of meeting all constraints.

[0149] Specifically, the meaning of the optimal individual is as follows:

[0150] 1) Scheduling time: Each scheduling time corresponds to a specific time period in a day.

[0151] 2) Energy storage status: the charging and discharging status (charging or discharging) of the energy storage device in each time period.

[0152] 3) Charging power: the charging power of the energy storage device in each time period.

[0153] 4) Discharge power: the discharge power of the energy storage device in each time period.

[0154] 5) Remaining capacity: the remaining capacity of the energy storage device at the end of each time period.

[0155] In the algorithm process, each individual is a possible scheduling solution, which consists of a series of decision variables that define the state and power level of the energy storage system at each scheduling moment. By evaluating the fitness value of each individual, that is, calculating the objective function value (total economic cost) corresponding to each scheduling solution, the individual with the lowest fitness value is selected from the population, that is, the scheduling solution with the lowest total economic cost.

[0156] Outputting the optimal individual The last step of the algorithm is to output the optimal individual, which is a scheduling solution that can achieve the lowest total economic cost. This means that the solution minimizes depreciation costs, operation and maintenance costs, and electricity purchase costs while satisfying all constraints, thereby providing the optimal scheduling strategy for the cloud energy storage system. Termination conditions The termination conditions of the algorithm can be reaching a predetermined maximum number of iterations, the fitness value reaching a certain threshold, or there is no significant improvement for several consecutive generations. Once the termination conditions are met, we have obtained the optimal individual, that is, the best scheduling solution that achieves the goal of minimizing the overall economic cost.

[0157] In the embodiment of the present invention, the cost calculation model processing in step S3 specifically includes the following steps:

[0158] a1. Depreciation cost analysis:

[0159] Depreciation costs are usually calculated using the equal annual value method, expressed as:

[0160]

[0161] Where: P i (t): is the output power of the i-th micro power source at time t;

[0162] r i : is the fixed annual interest rate of the i-th micro power source;

[0163] L: investment repayment period;

[0164] a2. Operation and maintenance cost analysis:

[0165] The operation and maintenance costs can be expressed as:

[0166]

[0167] Where: α i : Operation and maintenance cost coefficient of the i-th micro power source unit output;

[0168] N: is the number of micro power sources;

[0169] P i (t): is the output power of the i-th micro power source at time t;

[0170] a3. Analysis of electricity purchase cost: The electricity purchase cost is:

[0171] C G (t) = C P (t)*P G (t);

[0172] Where: C P (t): the electricity purchase price at time t;

[0173] P G (t): is the purchased electricity power at time t.

[0174] In the embodiment of the present invention, the power balance constraint in step S4 ensures that at each time point t, the total output power in the microgrid matches the load:

[0175]

[0176] Where: L(t): is the load at time t;

[0177] P S (t): is the charging and discharging power of the energy storage system (discharging is positive, charging is negative).

[0178] In the embodiment of the present invention, the energy storage constraint in step S4 ensures that the charging and discharging process of the energy storage system does not exceed its capacity limit:

[0179] E min ≤E(t)≤E max ;

[0180] Where: E(t) is the capacity of the energy storage system at time (t);

[0181] E min and E max are the minimum and maximum capacities of the energy storage system respectively.

[0182] In the embodiment of the present invention, if it is considered that there are 24 time steps in a day in step T1, each individual may be a number of distributed energy units with a length of 24×(n+1) plus an energy storage system.

[0183] Embodiment 2: The difference between the embodiment of the present invention and embodiment 1 is that: a specific variant operation step in step T2:

[0184] b1. For each time step t and each distributed energy unit k, a variation value Δ is drawn from a distribution k,t ;

[0185] b2. For each time step t, extract a variation value Δ of the energy storage system from a distribution S,t ;

[0186] b3. Add the mutation value to the original value to generate the mutation individual v i , the specific formula is as follows:

[0187]

[0188] b4. Assume that there are 3 distributed energy units and 1 energy storage system. Each time step t extracts a variation value from a normal distribution with a mean of 0 and a standard deviation of 10 kW. Then, the variation individual v i is written as:

[0189]

[0190] Among them, the variation value Δ k,t and Δ S,t By independently drawing from a chosen probability distribution, this mutation operation helps explore the solution space and increase population diversity.

[0191] In the embodiment of the present invention, the crossover operation in step T3 is performed for each target vector x i , using the mutation vector v i To generate the test vector u i , the crossover operation can be defined as:

[0192]

[0193] Where: rand j is a random number in the interval [0,1], CR is the crossover probability (usually set to a value between 0 and 1), j rand is an index randomly selected in the range [1,D], ensuring that at least one gene is selected from the mutation vector v i Inherited from , D is the dimension that usually represents the decision variable, that is, the length of each individual (vector). In this particular problem, each individual represents a scheduling plan, which includes the output power of each distributed energy unit at each time step and the charging and discharging power of the energy storage system.

[0194] Specifically, in this example, assuming that there are 24 time steps in a day, there are n distributed energy units and 1 energy storage system, the length D of each individual (vector) will be composed of the following parts:

[0195] a. Output power of n distributed energy units at each time step, with a total of 24×n decision variables.

[0196] b. The charging and discharging power of the energy storage system at each time step, a total of 24 decision variables.

[0197] Therefore, the total dimension D is 24n+24. For example, if there are 3 distributed energy units, the length D of each individual is. In the crossover operation, j rand is a randomly selected index in the range [1, D], which means it randomly picks an integer between 1 and D. This ensures that at least one gene (i.e., the value of the decision variable) is inherited from the mutation vector to the test vector 24×3+24=96.

[0198] Crossover operation example: Suppose we have a target vector x i and a mutation vector v i , each vector contains 3 distributed energy units and 1 energy storage system, and each time step is 24. We randomly select a j rand Value, such as j rand =5, and set the crossover probability CR = 0.8. For each j (i.e., each time step), we randomly select a rand from the interval [0,1] j Value, if rand j ≤CR or j=j rand , then use the mutation vector v i Otherwise, the target vector x is used. i In this way, we get the test vector u i .

[0199] In the embodiment of the present invention, step p2 is repeated until the termination condition is met as follows:

[0200] c1. For each individual x i ∈P0;

[0201] c2. Select individuals for reproduction of the next generation;

[0202] c3. Perform crossover and mutation operations on the selected individuals to generate new individuals;

[0203] c4. Update population P.

[0204] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0206] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0207] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A cloud energy storage scheduling optimization algorithm based on differential evolution, characterized by: The specific steps include: S1. Optimization goal: Optimize cost expenditure through effective scheduling of distributed power sources, intelligent charging and discharging management of energy storage systems, and analysis of energy exchange with external power grids; S2. Objective function: Define an objective function to comprehensively consider various costs, with the goal of minimizing the overall economic cost. The objective function can be expressed as: Where: C D (t): depreciation cost at time t; C O (t): operation and maintenance cost at time t; C G (t): is the electricity purchase cost at time t; S3, cost calculation model processing; S4. Constraint analysis: including two constraints: power balance constraint and energy storage constraint; S5. Differential evolution algorithm processing: The differential evolution algorithm includes objective function, cost model, constraint conditions and optimization based on these elements, specifically including the following steps: T1. Population initialization: Randomly generate an initial population P0 containing N individuals, each individual x i Represents a possible scheduling scheme, which is a vector containing the output power P of each distributed energy unit at each time step t. i (t) and the charging and discharging power P of the energy storage system S (t), x i The expression is as follows: Among them, P i,k (t) represents the output power of the kth distributed energy unit of the ith individual at time step t, and P S (t) represents the charging and discharging power of the energy storage system at time step t; T2, mutation: usually affects individual x i Make some kind of change to produce a mutant individual v i , assuming a simple mutation operation is used, that is, for individual x i Each gene g in the random increase or decrease of a small value Δ g , this value is drawn from a specific probability distribution, normal distribution or uniform distribution, and the mutation operation can be defined as: v i =x i +Δ; where Δ is a i A vector of the same dimension, each element of which is Δ g Independently draw from a probability distribution that is adjusted based on actual needs; T3. Crossover: The crossover operation is used to combine the information of two or more individuals to generate a new individual. In the differential evolution algorithm, the crossover operation is often used to mix the information of the target vector and the mutation vector to generate a new test vector u i ; T4, Selection: The selection operation is used to determine which individual will enter the next generation population. In the differential evolution algorithm, it is as follows: e1, for each target vector x i and the corresponding test vector u i , calculate their fitness values ​​f(x i ) and f(u i ); e2. If the test vector u i The fitness value is better than the target vector x i The fitness value, that is, f(u i )<f(x i ), then the test vector u i Replace the target vector x i Become part of the next generation population; e3. Otherwise, keep the target vector x i In the next generation population. This process can be expressed as: Here, g represents the number of iterations. In each generation, the algorithm performs mutation, crossover, and selection operations on the population, thereby gradually improving the fitness of the population and eventually converging to the global optimal solution or the local optimal solution. T5. Constraint processing: In the cloud energy storage scheduling optimization problem, there may be various constraints, such as power balance constraints, energy storage system capacity constraints, etc. These constraints are added to the fitness calculation through the penalty function, or the individuals that do not meet the constraints are directly modified through the repair mechanism. The penalty function is used if the individual x i If the constraint condition is not met, a penalty term p(x i ): f(x i )=f original (x i )-p(x i ); Among them, f original (x i ) is the fitness value without considering the constraints; T6. Iteration: The genetic algorithm will go through multiple iteration cycles until a certain termination condition is reached. Each iteration cycle includes steps such as fitness evaluation, selection, crossover, and mutation. The iteration process can be expressed in pseudo code as follows: p1, initialize the population P0; p2. Repeat until the termination condition is met: p3. Output the optimal individual: In the cloud energy storage scheduling optimization problem, define an objective function f(x) to comprehensively consider various costs, with the goal of minimizing the overall economic cost. S6. Through the differential evolution algorithm process of step S5, individuals in the population are continuously iterated and optimized to find the optimal scheduling solution that can achieve the lowest economic cost. The optimal individual finally output is a detailed scheduling strategy, which includes the charging and discharging status, charging power, discharging power and remaining capacity of the energy storage system at each scheduling moment, so as to ensure that the overall economic cost is minimized under the premise of meeting all constraints.

2. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized in that: The cost calculation model processing in step S3 specifically includes the following steps: a1. Depreciation cost analysis: Depreciation costs are usually calculated using the equal annual value method, expressed as: Where: P i (t): is the output power of the i-th micro power source at time t; r i : is the fixed annual interest rate of the i-th micro power source; L: investment repayment period; a2. Operation and maintenance cost analysis: The operation and maintenance costs can be expressed as: Where: α i : Operation and maintenance cost coefficient of the i-th micro power source unit output; N: is the number of micro power sources; P i (t): is the output power of the i-th micro power source at time t; a3. Analysis of electricity purchase cost: The electricity purchase cost is: C G (t)=C P (t)*P G (t); Where: C P (t): the electricity purchase price at time t; P G (t): is the purchased electricity power at time t.

3. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized in that: The power balance constraint in step S4 ensures that at each time point t, the total output power in the microgrid matches the load: Where: L(t): is the load at time t; P S (t): is the charging and discharging power of the energy storage system.

4. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized in that: The energy storage constraint in step S4 ensures that the charging and discharging process of the energy storage system does not exceed its capacity limit: E min ≤E(t)≤E max ; Where: E(t) is the capacity of the energy storage system at time (t); E min and E max are the minimum and maximum capacities of the energy storage system respectively.

5. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized in that: In step T1, if it is considered that there are 24 time steps in a day, each individual may be a distributed energy unit with a length of 24×(n+1) plus an energy storage system.

6. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized in that: A specific mutation operation step in step T2: b1. For each time step t and each distributed energy unit k, a variation value Δ is drawn from a distribution k,t ; b2. For each time step t, extract a variation value Δ of the energy storage system from a distribution S,t ; b3. Add the mutation value to the original value to generate the mutation individual v i , the specific formula is as follows: b4. Assume that there are 3 distributed energy units and 1 energy storage system. Each time step t extracts a variation value from a normal distribution with a mean of 0 and a standard deviation of 10 kW. Then, the variation individual v i is written as: Among them, the variation value Δ k,t and Δ S,t Draw independently from a chosen probability distribution.

7. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized in that: The crossover operation in step T3 is performed for each target vector x i , using the mutation vector v i To generate the test vector u i , the crossover operation can be defined as: Where: rand j is a random number in the interval [0,1], CR is the crossover probability, j rand is an index randomly selected in the range [1,D], ensuring that at least one gene is selected from the mutation vector v i Inherited from , D is the dimension that usually represents the decision variable, that is, the length of each individual. In this particular problem, each individual represents a scheduling plan, which includes the output power of each distributed energy unit at each time step and the charging and discharging power of the energy storage system.

8. The cloud energy storage scheduling optimization algorithm based on differential evolution according to claim 1 is characterized by: The step p2 is repeated until the termination condition is met as follows: c1. For each individual x i ∈P0; c2. Select individuals for reproduction of the next generation; c3. Perform crossover and mutation operations on the selected individuals to generate new individuals; c4. Update population P.

Images