A method, apparatus and product for optimizing the scheduling of microgrid systems

Abstract

This invention discloses a method, apparatus, and product for optimizing the scheduling of microgrid systems, relating to the field of smart grids, to improve the efficiency and quality of solving scheduling schemes for microgrid systems. The invention includes: constructing a microgrid optimization model describing the scheduling schemes of each distributed unit; configuring the operational constraints of the microgrid optimization model; and using a particle swarm optimization algorithm to iteratively solve for the optimal scheduling scheme solution set of the microgrid optimization model, treating each scheduling scheme as a particle, with the goal of minimizing the optimization objective. During the iteration process, the particle swarm is optimized hierarchically, and non-dominated solutions are recorded and updated using external archives. A linear decreasing strategy is employed to update the inertia factors of the two layers of particles. This invention improves convergence efficiency and Pareto front quality.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method, apparatus and computer program product for optimizing the scheduling of microgrid systems. Background Technology

[0002] In recent years, with the continuous optimization of the global energy structure and the increasing penetration rate of renewable energy, the traditional power system, which is mainly based on centralized energy supply, is facing profound changes. To enhance the flexibility, resilience, and sustainability of the power system, microgrid systems, as a new type of distributed energy system, have emerged and are gradually becoming an important supporting technology for achieving efficient energy utilization, carbon emission control, and smart grid construction. As an important component of the next-generation power system, microgrid systems are widely used in production and daily life due to their flexible structure, controllable operation, and local power supply. They typically integrate multi-energy complementary distributed units such as wind power generation, photovoltaic power generation, diesel engines, and energy storage systems. The operation and scheduling of microgrid systems requires coordinating the output allocation of various microgrid units to achieve a comprehensive optimization of multiple objectives, including economy, environmental protection, and reliability, under grid-connected or islanded operation modes. This process is affected by various factors such as the volatility of renewable energy, dynamic changes in load, and equipment operating constraints, constituting a typical multi-objective, multi-constraint nonlinear optimization problem.

[0003] To address such complex problems, numerous researchers have introduced intelligent optimization algorithms. In the early stages, heuristic methods such as ant colony optimization, genetic algorithms, and simulated annealing were widely used. However, these methods suffer from low search efficiency and difficulty in parameter tuning in multi-objective scenarios. In recent years, particle swarm optimization (PSO) algorithms have gradually become an important tool for multi-objective optimization scheduling of microgrids due to their simple structure, fast convergence speed, and ease of coding. The multi-objective Particle Swarm Optimization (MOPSO) algorithm enhances the diversity of the solution set and its boundary expansion capabilities by introducing non-dominated sorting mechanisms and external archive management mechanisms.

[0004] However, when using the traditional MOPSO algorithm to solve microgrid system scheduling schemes, the optimization process is often hampered by weak local search capabilities and premature convergence of the solution set. This is especially true in high-dimensional objective spaces where the solution set is sparsely distributed, making it difficult to obtain a comprehensive Pareto front, i.e., the optimal solution set. Furthermore, the convergence speed is slow, making it difficult to quickly obtain a reliable scheduling scheme. Summary of the Invention

[0005] The purpose of this invention is to provide a method, apparatus, and product for optimizing the scheduling of microgrid systems, thereby improving the efficiency and quality of solving multi-objective scheduling schemes for microgrid systems, addressing all or part of the problems mentioned above.

[0006] The technical solution adopted in this invention is as follows: A method for optimizing the scheduling of a microgrid system, the microgrid system comprising multiple distributed units; the method includes: S1. Construct a microgrid optimization model that describes the scheduling schemes of each distributed unit. The optimization objectives include operating costs and environmental protection costs. S2. Configure the operating constraints of the microgrid optimization model to calculate the constraint violation degree, which characterizes the degree to which each distributed unit violates the operating constraints, based on the constrained terms. S3. Treating each scheduling scheme as a particle, the particle swarm optimization algorithm is used to iteratively solve for the optimal scheduling scheme solution set of the microgrid optimization model with the goal of minimizing the optimization objective; each iteration process includes: S31. Record the non-dominated solution for each iteration using an external archive, and update the external archive based on the constraint violation degree and particle density of the particles; S32. Divide the particle swarm into a first layer and a second layer to perform local optimum exploration and global optimum exploration respectively; S33. The inertia factors of the first layer and the second layer are updated respectively using a linear decreasing strategy.

[0007] This application also provides a microgrid system optimization scheduling device, which includes a processor and a storage medium. The storage medium stores computer instructions, and the processor runs the computer instructions to execute the above-described microgrid system optimization scheduling method.

[0008] This application also provides a computer program product, including a computer program that, when run by a processor, executes the above-described microgrid system optimization scheduling method.

[0009] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. Based on the traditional MOPSO algorithm, this application introduces a hierarchical dynamic learning structure tailored to the characteristics of microgrid systems. The particle population is divided into two layers, each responsible for local development and global search, respectively. By dynamically hierarchizing the particles and setting differentiated learning factors, the first layer of particles focuses on the refined development of existing optimal solutions, while the second layer is responsible for expanding the search boundary. Furthermore, this application employs a linear decreasing strategy to adjust the inertia factor, guiding the particles to gradually transition from initial global search to later local development during the optimization process. This improves the global coverage and boundary diversity of the solution set, effectively overcoming the shortcomings of traditional algorithms that easily get trapped in local optima, and enhancing the overall optimization capability and convergence efficiency of the algorithm.

[0010] 2. This application designs an external archive update mechanism based on dominance relations and constraint violation degrees. It combines a grid density roulette wheel strategy to screen, clean, and select the global optimum position for non-dominated solutions. This not only ensures the feasibility of the solution set but also enhances the balance and representativeness of the solution set distribution. While maintaining the quality of the Pareto front solution set, this mechanism avoids the problem of solution set accumulation, improving the engineering applicability and stability of multi-objective optimization scheduling solutions. Attached Figure Description

[0011] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 This is a construction diagram of a microgrid system in one embodiment.

[0012] Figure 2 This is a flowchart of one embodiment of the microgrid system optimization scheduling method.

[0013] Figure 3 A line graph of ambient temperature index.

[0014] Figure 4 A line graph of ambient light intensity index.

[0015] Figure 5 Line graph of environmental wind speed index.

[0016] Figure 6 Load power demand indicators.

[0017] Figure 7 This is a distribution diagram of feasible and non-dominated solutions in the traditional MOPSO algorithm.

[0018] Figure 8 This is a distribution diagram of the Pareto front solution set of the traditional MOPSO algorithm.

[0019] Figure 9 This is a distribution diagram of feasible and non-dominated solutions in this application.

[0020] Figure 10 This is a distribution diagram of the Pareto front solution set in this application.

[0021] Figure 11 It is a traditional MOPSO output scheduling scheme under grid-connected mode.

[0022] Figure 12 It is a traditional MOPSO output scheduling scheme in the island mode.

[0023] Figure 13 This is the power output scheduling scheme for this application under grid-connected mode.

[0024] Figure 14 This is the power output scheduling scheme for this application under the isolated mode. Detailed Implementation

[0025] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0026] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0027] A typical microgrid system is constructed as follows: Figure 1 As shown. Addressing the challenges of early intelligent optimization algorithms being difficult to apply to solving multi-objective optimization problems in microgrid system scheduling, and the traditional MOPSO algorithm's difficulty in obtaining the optimal scheduling scheme set and its slow convergence speed, this application proposes a microgrid system optimization scheduling method, apparatus, and product, aiming to efficiently solve for an effective and reliable scheduling scheme set for microgrid systems.

[0028] Microgrids primarily operate in two modes: grid-connected and islanded. In the system architecture, the microgrid is connected to the main grid via a Point of Common Coupling (PCC). Controlling this PCC allows for dynamic switching between the two operating modes. Under normal circumstances, the PCC is on, and the system operates in grid-connected mode, enabling bidirectional energy exchange between the microgrid and the main grid. When the external grid fails or requires independent operation, the PCC is off, and the microgrid switches to islanded mode, relying on internal distributed energy resources to maintain local load power supply.

[0029] like Figure 2 As shown in the embodiments of this application, the microgrid system optimization scheduling method includes the following process: S1. Construct a microgrid optimization model describing the scheduling schemes of each distributed unit. The optimization objectives include operating cost C and environmental protection cost E.

[0030] As an optional implementation, operating costs include the operation and maintenance expenses of the microgrid system. Fuel costs and microgrid electricity purchase and sale fees Microgrid power purchase and sale refers to the purchase and sale of electricity from a microgrid system to the main power grid.

[0031] Specifically, the operating cost C of a microgrid system is calculated as follows: Equation (1): ; In the formula, T represents the time period of the scheduling scheme, t represents each time period in T, and PCC represents the operating state of the microgrid. A state of on indicates that the grid-connected mode is activated, and a state of off indicates that the connection is disconnected and the island mode is entered.

[0032] In formula (1), the operation and maintenance cost is calculated by formula (2): Formula (2): ; wherein, are sequentially the operating cost of the wind power generation unit, the operating cost of the photovoltaic power generation unit, the operating cost of the diesel power generation unit, the operating cost of the micro gas turbine unit and the operating cost of the storage battery unit.

[0033] Fuel cost only involves some distributed units that require fuel, and it is calculated by formula (3): Formula (3): ; wherein, are the fuel cost of the diesel power generation unit and the fuel cost of the micro gas turbine unit respectively.

[0034] Environmental protection cost E includes the treatment cost of pollutants generated by each distributed unit. The environmental protection cost E of the microgrid system depends on the environmental protection cost of each distributed unit, and some distributed units (such as clean energy power generation units) may not involve this part of the cost.

[0035] Specifically, the calculation method of environmental protection cost E is: Formula (4): ; In the formula, N represents the number of units, n represents a unit unit, represents the unit environmental protection cost of the nth unit, sequentially represent the pollutant production amount of the diesel power generation unit, the pollutant production amount of the micro gas turbine unit and the pollutant production amount of the power purchase and sale link of the microgrid. The pollutant production amount of each part is calculated based on the corresponding electric energy power: Formula (5): ; In the formula, are sequentially the actual output power of the diesel power generation unit, the actual output power of the micro gas turbine unit and the power purchased and sold by the microgrid; represent the pollutant gas emission coefficients of the diesel generator, the micro gas turbine and the large power grid respectively.

[0036] According to the operating cost C and the environmental protection cost E of the microgrid system, the optimization objective function F of the microgrid optimization model is constructed: Formula (6): ; The optimization objective function F represents minimizing both the operating cost C and the environmental protection cost E.

[0037] S2. Configure the operating constraints of the microgrid optimization model to calculate the constraint violation degree, which characterizes the degree to which each distributed unit violates the operating constraints, based on the constrained terms.

[0038] As an optional implementation, the constrained terms in the operating constraints include power balance constraints, distributed unit output constraints, energy storage state constraints, and grid switching constraints.

[0039] S21. Set a power balance constraint so that the output power of the microgrid system must equal the power required by the load. The power balance constraint is configured as follows: Equation (7): ; in, The values ​​represent, in order, the total power demand at the microgrid load end, the output power of the wind power generation unit, the output power of the photovoltaic power generation unit, and the charging and discharging power of the battery during the operating cycle.

[0040] S22. Distributed unit output constraints limit the actual output power of each distributed unit to between its upper and lower power thresholds. Additionally, fuel power generation units (diesel power generation units and micro gas turbine units), as controllable distributed units, require limits on ramp-up power to prevent large output power fluctuations. Specifically, the distributed unit output constraints are configured as follows: Equation (8): ; Equation (9): ; Equation (10): ; Equation (11): .

[0041] In equations (8)-(11), These represent the upper and lower limits of the output power of the wind power generation unit, respectively. These represent the upper and lower limits of the output power of the photovoltaic power generation unit, respectively. These represent the upper and lower limits of the output power of the diesel generator unit, respectively. These represent the upper and lower limits of the output power of the photovoltaic power generation unit, respectively. This indicates the ramping power of the diesel generator unit. These represent the output power of the diesel generator unit in the later time period t and the earlier time period t-1, respectively, and the difference in output power between the two adjacent time periods. The climbing power must not be exceeded. ; Indicates the ramp-up power of the micro gas turbine unit. The output power of the micro gas turbine unit in the next two adjacent time periods t and t-1 are respectively, and the difference in output power between the two adjacent time periods is also represented. The climbing power must not be exceeded. .

[0042] S23. Energy storage state constraints include power constraints and charge state constraints. The specific configuration is as follows: Equation (12): ; In the formula, Indicates the output power of the battery unit. These represent the upper and lower limits of the battery unit's output power, respectively. Indicates the state of charge of the battery cell. These represent the upper limit and lower limit of the state of charge of the battery cell, respectively.

[0043] S24, Grid Exchange Constraints describe the limitations on the power exchange between the microgrid system and the main grid, and are configured as follows: Equation (13): ; In the formula, These represent the upper and lower limits of the power that a microgrid system can purchase and sell to the main grid, respectively.

[0044] This completes the configuration of each constraint in the operational constraints.

[0045] In the process of resolving the scheduling scheme, it is necessary to calculate the degree to which each distributed unit in the scheduling scheme violates the operational constraints based on the configured operational constraints. In this application, a constraint violation degree is defined to quantify the degree of violation. In one optional implementation, the method for calculating the constraint violation degree includes: S25. Calculate the number of violations of operating constraints by each distributed unit throughout the entire operating cycle.

[0046] Specifically, the amount of violation of operational constraints by each distributed unit throughout the entire operating cycle includes the following aspects: Total output power The degree to which total power demand is violated Diesel generator ramping power violation degree Micro gas turbine unit ramp power violation degree And the degree of violation of the state of charge of the battery cell .

[0047] Equation (14): ; In the formula, This represents the difference in total output power between two adjacent time periods t and t-1.

[0048] Equation (15): .

[0049] Equation (16): ; Equation (17): .

[0050] S26. Summarize the violation amounts of each distributed unit to obtain the constraint violation degree. V .

[0051] Equation (18): .

[0052] The constraint violation degree calculated by the above method can measure the degree of feasibility of the solution represented by the particle, which is helpful for particle selection and feasibility determination.

[0053] S3. Treating each scheduling scheme as a particle, the particle swarm optimization algorithm is used to iteratively solve the optimal scheduling scheme solution set of the microgrid optimization model with the goal of minimizing the optimization objective.

[0054] The particle swarm optimization algorithm finds the optimal scheduling solution set through multiple rounds of iterative optimization. In one optional embodiment of this application, the following improved features are included in each iteration: S31. Use an external archive to record the non-dominated solution for each iteration, and update the external archive based on the constraint violation degree of the particles and the particle density.

[0055] Specifically, as an optional implementation, step S31 includes the following process: S311. After each iteration, determine the dominance relationship of all particles in the current particle swarm and select non-dominated solutions as candidate particles.

[0056] S312. Write particles that satisfy the operating constraints into the external archive; for particles that violate the operating constraints, if the number of particles in the external archive does not reach the capacity threshold and the constraint violation degree of the particle does not exceed the set threshold, temporarily store the particle in the external archive. That is, for feasible non-dominated solutions that satisfy the operating constraints, they are directly added to the external archive, while for non-dominated solutions that violate some operating constraints, if their constraint violation degree is low and the number of non-dominated solutions in the external archive is insufficient, they are allowed to be temporarily retained as potential high-quality solutions to enhance the diversity of the solution set.

[0057] S313. If the number of particles in the external archive exceeds the capacity threshold, particles that violate the operating constraints are first removed. If the capacity threshold is still exceeded, the particle reachable space (i.e., the optimization target space) is divided into grids and the particle density of each grid is calculated. Based on the particle density, the grids are sampled using a roulette wheel to remove particles, where the grid selection probability is positively correlated with the particle density.

[0058] The roulette wheel sampling for removing particle meshes is based on the mesh density weighting function, and the probability of selection is: Equation (19): ; In the formula, The grid density weighting function represents the grid density of the m-th grid during particle removal; This represents the probability that the m-th grid is selected; The value represents the number of non-dominated solutions in the m-th grid, i.e., the particle density of the m-th grid; K is the number of grid cells; in the grid density weighting function, This is the first coefficient, also known as the exponential amplification factor, and it takes a positive value, indicating that the probability of grid selection is positively correlated with particle density. When When the value is large, the weight of dense regions will be significantly amplified, while the weight of sparse regions will be reduced; as... As the probability of selecting each grid cell decreases, the probability of each grid cell being selected tends to be equal. When =0, each grid is randomly selected.

[0059] S314. When selecting the globally optimal position for updating particle velocity from external archives, the reachable space of particles is divided into grids and the particle density of each grid is calculated. Based on the particle density, the grids are sampled by roulette wheel to select particles, where the grid selection probability is inversely related to the particle density.

[0060] Using external archives can effectively speed up the search for the optimal solution and improve search efficiency.

[0061] The roulette wheel sampling of the grid for selecting the globally optimal particle is also based on the grid density weighting function, with the probability of selection being: Equation (20): ; In the formula, The grid density weighting function represents the grid density of the m-th grid when the global optimal position is selected. This represents the probability that the m-th grid is selected; This is the second coefficient, the exponential amplification factor, which takes a negative value, indicating that the probability of grid selection is negatively correlated with particle density. When When the weight is small, the weight of dense regions will be significantly reduced, and the weight of sparse regions will be increased; as the weight decreases, the weight of dense regions will be increased. As the value increases, the probability of each grid being selected tends to be equal. When =0, each grid is randomly selected.

[0062] Furthermore, in an optional implementation, the method for performing roulette wheel sampling on the grid in any of the above-described roulette wheel sampling methods is as follows: S315. Generate a random probability value; sequentially traverse each grid and calculate the cumulative probability of the traversed grid being selected, and select the grid whose cumulative probability reaches the random probability value and minimizes the cumulative probability.

[0063] Step S315 can be represented as: Equation (21): , ; In the formula, G represents the selected grid. Represents a random number. , This represents the probability that the j-th grid cell is selected during the iteration. This represents the cumulative probability that the m-th grid is selected.

[0064] The above method ensures that each iteration yields the globally optimal particle for the next iteration, thereby improving the convergence speed. Furthermore, randomly removing particles from dense regions using the roulette wheel algorithm enhances the distribution balance and diversity of the solution set, increasing the selectivity of excellent particles.

[0065] S32. Divide the particle swarm into a first layer and a second layer to conduct local optimization and global optimization respectively.

[0066] In this application, the optimal position is explored from both local and global dimensions in each iteration, which can effectively solve the problems of weak local search ability and premature convergence of solution set in the traditional MOPSO algorithm, and obtain a comprehensive Pareto front.

[0067] In one alternative implementation, before each iteration, the dominance relationship of all particles in the current particle swarm is determined. Particles belonging to non-dominated solutions and with a constraint violation degree V of zero are assigned to the first layer for local optimum exploration, while the remaining particles are assigned to the second layer for global optimum exploration. Particles in different layers achieve local optimum exploration and global optimum exploration respectively by setting different learning factors.

[0068] This structure effectively balances the global exploration and local exploitation capabilities of the population by layering particles, thereby improving the quality and distribution uniformity of the Pareto solution set.

[0069] Specifically, based on the initial particle swarm, after the particle swarm is divided into layers in each iteration, the particle update method during the optimal exploration process of the first and second layers is as follows: Speed ​​update: Equation (22): ; Position update: Equation (23): .

[0070] in, and These represent the updated velocity and position of the particle after each iteration; and These represent the particle's velocity and position before each iteration; These are the learning factors for the particle's optimal position relative to itself and the optimal position of the particle swarm, respectively. In the first layer... Comparison Large, second floor Comparison Small; These are random numbers within the range [0,1]. w Indicates the inertia factor; This represents the particle's historical best position. This represents the historical best position of the particle swarm.

[0071] By assigning differentiated learning factors to particles at different levels, it is helpful to improve the overall convergence effect and the boundary diversity of the solution set.

[0072] For example, the learning factor of the first layer of particles c Set to: Equation (24): ; The learning factor of the second layer of particles c Set to: Equation (25): .

[0073] Of course, without violating the aforementioned constraints, the learning factor of the two-layer particles c Other weight values ​​can also be set.

[0074] S33. The inertia factors of the first and second layers are updated respectively using a linear decreasing strategy.

[0075] As can be seen from equation (22), during the iteration process of the two-layer particles, when updating the particle velocity, the inertial margin of the particle velocity in the previous round is calculated through the inertial factor. In this embodiment, the inertial factor of each iteration is reduced round by round through a linear decreasing strategy, guiding the particles from global exploration of high-quality solutions to local development of the current high-quality solution region, accelerating the algorithm to find the optimal solution, thereby improving the convergence speed.

[0076] As an optional implementation, the method of updating the inertia factors of the first layer and the second layer respectively using a linear decreasing strategy is as follows: Equation (26): ; in, Indicates the first i Maximum number of iterations per layer; Indicates the first i Current iteration round of the layer ; These are the maximum and minimum values ​​of the set inertia factor, respectively.

[0077] By introducing a linearly decreasing inertia factor strategy, the inertia factor gradually decreases with the number of iterations, guiding the particles from global exploration of high-quality solutions to local development of the current high-quality solution region, thereby automatically improving the accuracy of particle search.

[0078] Through the design of S31, S32, and S33 above, based on the initialized particle swarm, the particle swarm velocity is updated as shown in Equation (22). The position of the new generation of particles is calculated according to Equation (23). The constraint violation degree V and the optimization objective value F of the particles are re-evaluated based on the updated position, thereby performing the next round of particle swarm layering and iteration. This process is repeated until the global optimal position obtained at convergence is the optimal scheduling scheme solution set. The convergence condition can be set to reach a predetermined maximum number of iterations, or the optimization objective value of the optimal position explored by the particle swarm meets the set minimum allowable error.

[0079] Finally, embodiments of this application may further include: S4. Output the optimal scheduling scheme solution set. The distributed units of the microgrid system are scheduled based on the optimal scheduling scheme solution set.

[0080] Compared with known methods for solving microgrid system scheduling schemes based on the MOPSO algorithm, this application incorporates the following optimization designs when using the particle swarm optimization algorithm for scheduling scheme optimization: First, in traditional particle swarm optimization (PSO) algorithms, all particles follow the same evolutionary pattern, which can easily lead to premature convergence or getting trapped in local optima, and lacks targeted search strategies. The embodiments of this application dynamically divide particles into two levels based on their non-dominated relationships and constraint violations. The first level consists of particles with currently feasible non-dominated solutions, primarily responsible for in-depth exploration of the optimal solution region of the objective function; the second level includes infeasible or dominated particles, focusing on global search and boundary expansion tasks. This structure effectively improves the search diversity of the population and the ability to extend the edge of the solution set.

[0081] Second, based on the hierarchical structure, to further refine the control of particle behavior bias, this application sets different learning factor combinations for the first and second layers. The learning factor for the first-layer particles emphasizes guidance from individual historical optimal solutions and global optimal solutions, thereby enhancing local exploration capabilities; the learning factor setting for the second-layer particles enhances their sensitivity to group information, encouraging them to perform wide-area jump searches. Simultaneously, to balance the breadth of the search and the convergence speed, a linearly decreasing inertia factor strategy is designed, allowing particles to have greater inertia in the early stages of the algorithm to perform large-scale exploration, while gradually reducing inertia in the later stages to promote convergence.

[0082] Third, in each iteration, a non-dominated external archive is established and maintained, and particles are screened using a dual criterion of dominance relationship and constraint violation degree: feasible non-dominated solutions that satisfy the constraints are directly added to the external archive; for non-dominated solutions that violate some constraints, if their constraint violation degree is low and the number of solutions in the current external archive is insufficient, they are allowed to be temporarily retained as potential high-quality solutions to enhance the diversity of the solution set. When the external archive exceeds the limit, individuals with high violation degree are removed first; if redundancy still exists, particles are mapped to a multi-objective space grid, and a probability roulette wheel betting function is constructed through grid density statistics to preferentially remove individuals from dense particle regions, maintaining the overall capacity and distribution balance of the archive. In addition, this mechanism is also used to select the globally optimal particle for velocity update, preferentially selecting from sparse regions of the target space to enhance boundary expansion and diversity maintenance capabilities.

[0083] Based on the microgrid system optimization scheduling method designed above, this application embodiment also conducts a practical study on its performance.

[0084] The microgrid system comprises five distributed units: wind power generation unit, photovoltaic power generation unit, diesel power generation unit, micro gas turbine unit, and battery unit.

[0085] The output power of the wind power generation unit is calculated using the wind speed-power function relationship, as shown in equation (27), and its operating cost is shown in equation (28).

[0086] Equation (27): ; Equation (28): .

[0087] In the formula, This indicates the output power of the wind power generation unit, in kW. , , , These represent the actual wind speed, cut-in wind speed, rated wind speed, and cut-out wind speed of the wind power generation unit, respectively, in m / s. This indicates the rated output power of the wind power generation unit, in kW. , , These represent wind speed parameters; This represents the operating cost of the wind power generation unit during the current time period, in yuan. This indicates the operation and maintenance parameters of the wind power generation unit.

[0088] The output power expression of the photovoltaic power generation unit is established by the functional relationship between light intensity, temperature and power, as shown in equation (29), and the operating cost is expressed as equation (30).

[0089] Equation (29): ; Equation (30): .

[0090] In the formula, This indicates the output power of the photovoltaic power generation unit, in kW. , These represent the actual battery temperature and the battery temperature under standard test conditions during the current time period, respectively, in °C. This indicates the rated power of the photovoltaic power generation unit under standard test conditions, in kW. , These represent the actual light intensity and the light intensity under standard test conditions for the current time period, respectively, in kW / ; Indicates the power temperature coefficient; This represents the operating cost of the photovoltaic power generation unit during the current time period, in yuan. This indicates the operation and maintenance parameters of the photovoltaic power generation unit.

[0091] The relationship between the fuel consumption and the output power of the diesel generator unit can be quantitatively analyzed by equation (31), the constraint condition is expressed as equation (10), and its operating cost is shown in equation (32).

[0092] Equation (31): ; Equation (32): .

[0093] In the formula, This indicates the fuel consumption of the diesel generator unit during the current time period, in liters (L). , These represent the intercept coefficient and slope coefficient of the fuel curve for the diesel generator unit, respectively. This indicates the actual output power of the diesel generator unit during the current time period, in kW. This indicates the rated output power of the diesel generator unit, in kW. This represents the operating cost of the diesel generator unit during the current time period, in yuan. This indicates the operating and maintenance parameters of the diesel generator unit. The difference in output power between adjacent time periods of the diesel generator unit cannot exceed this ramp power, unit: kW.

[0094] Based on the relationship between the fuel consumption and output power of a diesel generator unit, the fuel cost of the diesel generator can be calculated by multiplying the fuel consumption during operation by the unit price of diesel fuel. The function expression is as follows: Equation (33): .

[0095] In the formula, This represents the fuel cost of the diesel generator during the current time period, in yuan. This indicates the unit price of diesel fuel on that day, in yuan / L.

[0096] The relationship between the power generation efficiency and output power of a micro gas turbine unit can be expressed by the following functional relationship (34), and its operating cost is given by equation (35) and fuel cost by equation (36).

[0097] Equation (34): ; Equation (35): ; Equation (36): .

[0098] In the formula, This indicates the actual output power of the micro gas turbine unit during the current time period, in kW. This indicates the power generation efficiency of the micro gas turbine unit during the current time period; , , , This represents the efficiency coefficient of a micro gas turbine unit; This represents the operating cost of the micro gas turbine unit during the current time period, in yuan. This indicates the operating and maintenance parameters of the micro gas turbine unit, in yuan / kW. This indicates the ramp power of the micro gas turbine unit. The difference in output power between adjacent time periods of the micro gas turbine unit cannot exceed this ramp power. Unit: kW. This represents the fuel cost of the micro gas turbine unit during the current time period, in yuan. This indicates the low thermal power of natural gas, in kW / ; This indicates the price of natural gas on that day, in yuan / .

[0099] The battery unit uses a lead-acid battery pack. Its state of charge and operating costs are shown in equations (37) and (38): Equation (37): ; Equation (38): .

[0100] In the formula, , These represent the battery capacity (i.e., state of charge) of the battery cell at time t and the battery capacity of the battery cell at time t-1, respectively, in C. , This indicates the discharge efficiency and charging efficiency of a battery cell, in units of C / kW. This represents the charging and discharging power of the battery cell during time period t, in kW. When, it indicates that the battery cell is in a discharging state; when When this time, it indicates that the battery unit is in a charging state. This indicates the operating cost of the battery unit during the current time period, in yuan. This indicates the operating and maintenance parameters of the battery unit; These represent the upper limit and lower limit of the state of charge of the battery cell, respectively, in C.

[0101] Secondly, to analyze the optimization performance of the proposed method, this application embodiment conducted 10 independent experiments on the traditional MOPSO algorithm and the proposed method based on the same microgrid multi-objective scheduling model. The experiments used environmental data from typical climates, such as... Figures 3 to 6 As shown in Table 1, during the experiment, the solution with the lowest total cost was selected as the optimal solution for each group of experiments. The total cost (i.e., the optimization objective) was the sum of the operating cost and the environmental protection cost. Finally, the average total cost of the optimal solutions in 10 experiments for both algorithms was calculated as their respective average comprehensive cost. The experimental results are shown in Table 1.

[0102] Table 1 Comparative Experiment Table

[0103] In terms of optimization performance, the average comprehensive cost of this application is RMB 33,166.4, significantly lower than the RMB 42,259.3 of the traditional MOPSO algorithm, indicating that this application performs better in the dual-objective scheduling of optimizing operation and environmental protection. Regarding solution set quality, the average number of non-dominated solutions obtained by this application is 74.5, an increase of approximately 35% compared to the 55.1 of the traditional MOPSO algorithm, demonstrating a significant advantage in solution diversity and Pareto front distribution accuracy. Although the average running time of this application (56.9 seconds) is slightly higher than the 43.8 seconds of the traditional MOPSO algorithm, considering its significant improvement in the number of solution sets and optimization results, this cost is reasonable and acceptable.

[0104] To further analyze the performance of this application in multi-objective optimization, this paper selects a representative experimental result from one of 10 independent experiments conducted by each of the two algorithms. The experimental results are as follows: Figures 7 to 10 As shown. Figure 7 This displays the distribution of feasible and non-dominated solutions in the traditional MOPSO algorithm. Green circles represent current feasible solutions, and blue asterisks represent non-dominated solutions saved in external archives. Figure 8 The corresponding Pareto front solution set is shown, reflecting the boundary characteristics and distribution of the final solution. As can be seen from the figure, the traditional MOPSO algorithm has 57 non-dominated solutions, and the solution set has a relatively small coverage area in the target space. In particular, its boundary expansion is weak in the dimensions of operating cost and environmental cost, exhibiting a certain degree of concentration and lack of diversity.

[0105] Figure 9 and Figure 10 This is the simulation result of this application, with the yellow pentagram representing the current elite layer particles. Figure 9 The results show that the algorithm obtained more non-dominated solutions, significantly increased the number of feasible solutions, and the solution set expanded significantly towards the target boundary region. Figure 10 The Pareto front solution set in this algorithm is uniformly distributed across both objective dimensions and has sufficiently extended boundaries, reflecting stronger global exploration capabilities and solution set diversity. Experimental results verify that this algorithm outperforms the traditional MOPSO algorithm in terms of inherent convergence performance, solution set diversity, and boundary coverage, thus validating the optimization potential and practical effectiveness of this application in the multi-objective scheduling problem of microgrids.

[0106] Finally, in this embodiment, a microgrid system in a climate with abundant wind and solar resources is selected as the research object. The performance of traditional MOPSO and this application in multi-objective optimization scheduling is compared under two modes: grid-connected operation and islanded operation. The operating cost and environmental protection cost data of each experiment are shown in Table 2.

[0107] Table 2 Cost Comparison Table

[0108] Figure 13This paper demonstrates the optimal power output scheme obtained in this application under grid-connected operation mode. During nighttime when wind resources are abundant and loads are low, the algorithm prioritizes wind turbine output while appropriately reducing the output of other units to lower costs. During peak daytime load periods, wind turbines and photovoltaic arrays cannot independently support all load demands. The improved algorithm coordinates the synchronous output of diesel generators and micro gas turbines, and supplements the system's power supply through battery discharge during certain periods. Furthermore, since diesel generators produce harmful gases that increase environmental protection costs, the micro gas turbines have higher output power than diesel generators, effectively controlling the system's environmental impact. In addition, the system maintains small-scale power purchases or sales for most periods, indicating that under grid-connected conditions, the algorithm tends to rely on natural resources for energy supply, only using the main grid for assistance when resources are insufficient.

[0109] The optimal scheduling result of the traditional MOPSO algorithm under the same climatic conditions and the same operating mode is as follows: Figure 11 As shown, the output results of this application differ significantly from those of the traditional MOPSO algorithm. The output trends of the wind turbines and photovoltaic arrays in this application follow the trends of environmental resource changes, demonstrating the full utilization of renewable energy; while the output of wind turbines and photovoltaic arrays in the traditional MOPSO algorithm is relatively low, and the utilization of renewable energy during operation is not as sufficient as in this application. Due to the insufficient utilization of natural resources, the output of diesel generators, micro gas turbines, and the main power grid in the traditional MOPSO algorithm is higher than that in this application, resulting in correspondingly higher operating costs and environmental protection costs.

[0110] Figure 14 This paper demonstrates the optimal power output scheme obtained in this application under islanded operation mode. In this scheme, the system is completely disconnected from the main grid, relying solely on distributed power sources and energy storage systems for autonomous power supply. From the power output structure perspective, wind turbine units have high output power at night when resources are abundant, providing primary support during low-load periods. During the day, the photovoltaic array reaches its peak output between 11:00 and 15:00, alleviating daytime peak load pressure. However, since wind and solar resources cannot cover all electricity demand, diesel generators and micro gas turbines undertake important regulation tasks, with the micro gas turbines having a higher output than the diesel generators, thus reasonably controlling the increase in environmental protection costs. Batteries play a balancing role in system operation, discharging moderately during the day and providing auxiliary power supply during periods of insufficient wind and solar power or increased load, enhancing system stability.

[0111] The scheduling results of the traditional MOPSO algorithm are as follows: Figure 12As shown in this application, the output of wind turbines and photovoltaic arrays closely follows resource fluctuations, improving the system's renewable energy utilization efficiency. In contrast, the output of diesel generators and micro gas turbines is significantly lower in the scheduling results of the traditional MOPSO algorithm. To meet system load demands, the traditional MOPSO algorithm requires higher output from diesel generators and micro gas turbines, making it more reliant on traditional energy sources overall, resulting in higher operating and environmental costs compared to the improved algorithm.

[0112] The above embodiments sufficiently demonstrate the significant advantages of this application in terms of quality and efficiency in solving microgrid system scheduling schemes.

[0113] Based on the ideas of this application, this application also proposes a microgrid system optimization scheduling device, which includes a processor and a storage medium. The storage medium stores computer instructions, and the processor runs the computer instructions to execute the above-described microgrid system optimization scheduling method.

[0114] Furthermore, this application also proposes a computer program product, including a computer program that, when run by a processor, executes the aforementioned microgrid system optimization scheduling method.

[0115] This invention is not limited to the specific embodiments described above. The invention extends to any new feature or combination disclosed in this specification, as well as any new method or process step or combination disclosed herein.

Claims

1. A method for optimizing the scheduling of a microgrid system, wherein the microgrid system comprises multiple distributed units; characterized in that, The methods include: S1. Construct a microgrid optimization model that describes the scheduling schemes of each distributed unit. The optimization objectives include operating costs and environmental protection costs. S2. Configure the operating constraints of the microgrid optimization model to calculate the constraint violation degree, which characterizes the degree to which each distributed unit violates the operating constraints, based on the constrained terms. S3. Treating each scheduling scheme as a particle, the particle swarm optimization algorithm is used to iteratively solve for the optimal scheduling scheme solution set of the microgrid optimization model with the goal of minimizing the optimization objective; each iteration process includes: S31. Record the non-dominated solutions for each iteration using an external archive, and update the external archive based on the constraint violation degree and particle density of the particles, including: S311. After each iteration, the dominance relationship of all particles in the current particle swarm is judged, and non-dominated solutions are selected as candidate particles. S312. Write particles that meet the operating constraints into the external file; for particles that violate the operating constraints, if the number of particles in the external file does not reach the capacity threshold and the constraint violation degree of the particles does not exceed the set threshold, temporarily store the particles in the external file. S313. If the number of particles in the external file exceeds the capacity threshold, particles that violate the operating constraints are first removed. If the number still exceeds the capacity threshold, the reachable space of the particles is divided into grids and the particle density of each grid is calculated. Based on the particle density, the grids are sampled by roulette wheel to remove particles, wherein the grid selection probability is positively correlated with the particle density. S314. When selecting the globally optimal position for updating particle velocity from the external archive, the reachable space of the particles is divided into grids and the particle density of each grid is calculated. Based on the particle density, the grids are sampled by roulette wheel to select particles, wherein the grid selection probability is inversely correlated with the particle density. The method for roulette wheel sampling of the grid is as follows: generate a random probability value; traverse each grid sequentially and calculate the cumulative probability of the traversed grid being selected, and select the grid whose cumulative probability reaches the random probability value and minimizes the cumulative probability; S32. Divide the particle swarm into a first layer and a second layer to explore local optima and global optima respectively, including: before each iteration, determine the dominance relationship of all particles in the current particle swarm, assign particles that belong to non-dominated solutions and have zero constraint violation degree to the first layer for local optima exploration, and assign the remaining particles to the second layer for global optima exploration. S33. The inertia factors of the first layer and the second layer are updated respectively using a linear decreasing strategy.

2. The microgrid system optimization scheduling method as described in claim 1, characterized in that, The operating costs include the operation and maintenance costs of the microgrid system, fuel costs, and the cost of purchasing and selling electricity from the microgrid; the environmental protection costs include the cost of treating pollutants generated by each distributed unit.

3. The microgrid system optimization scheduling method as described in claim 1, characterized in that, The constrained terms in the operational constraints include power balance constraints, distributed unit output constraints, energy storage state constraints, and grid switching constraints; the constraint violation degree, which characterizes the degree to which each distributed unit violates the operational constraints based on the constrained terms, includes: The number of violations of the operational constraints by each distributed unit during the entire operating cycle is counted separately. The constraint violation degree is obtained by summing the violation amounts of each distributed unit.

4. The microgrid system optimization scheduling method as described in claim 1, characterized in that, During the optimal exploration process in the first and second layers, the particle update method is as follows: Speed ​​updates: ; Location update: ; in, and These represent the updated velocity and position of the particle after each iteration; and These represent the particle's velocity and position before each iteration; These are the learning factors for the particle's optimal position relative to itself and the optimal position of the particle swarm, respectively. In the first layer... Comparison Large, second floor Comparison Small; These are random numbers within the range [0,1]. w Indicates the inertia factor; This represents the particle's historical best position. This represents the historical best position of the particle swarm.

5. The microgrid system optimization scheduling method as described in claim 4, characterized in that, The method for updating the inertia factors of the first and second layers respectively using a linear decreasing strategy is as follows: ； in, Indicates the first i Maximum number of iterations per layer; Indicates the first i Current iteration round of the layer ; These are the maximum and minimum values ​​of the set inertia factor, respectively.

6. A microgrid system optimized scheduling device, characterized in that, It includes a processor and a storage medium, the storage medium storing computer instructions, and the processor executing the computer instructions to perform the microgrid system optimization scheduling method as described in any one of claims 1-5.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is run by the processor, it executes the microgrid system optimization scheduling method as described in any one of claims 1-5.

Images