Virtual power plant mobile flexible resource dynamic configuration and capacity planning method and system

Abstract

The application discloses a kind of virtual power plant mobile flexible resource dynamic configuration and capacity planning method, system, belong to virtual power plant optimization scheduling technical field.The method includes: the typical scene and its probability are generated to photovoltaic annual output data by clustering;Electric vehicle charging load is simulated by Monte Carlo;Double-layer optimization model is constructed, the access node and capacity of mobile photovoltaic and energy storage are optimized in upper layer, output and charge-discharge power are optimized in lower layer, and mobile resource access position is adjusted by constraint;Improved Harris eagle optimization algorithm is used to solve, the improved algorithm includes elite reservation boundary processing and mixed variable collaborative evolution;The optimal access position, capacity and annual total cost are output.The application actively tracks seasonal fluctuation of photovoltaic output and spatial migration of electric vehicle load by dynamic relocation of mobile photovoltaic storage system, effectively reduces virtual power plant annual total cost and improves distribution network voltage quality.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant optimization scheduling technology, specifically to a method and system for mobile and flexible dynamic resource allocation and capacity planning in virtual power plants. Background Technology

[0002] Both distributed photovoltaic (PV) power generation and the electric vehicle (EV) industry are experiencing explosive growth. By the end of 2025, my country's new energy vehicle ownership had reached 43.97 million vehicles, and the cumulative installed PV capacity exceeded 1200 GW. However, distributed PV power output is highly random and intermittent, with its output mainly concentrated during daytime hours; EV charging load, on the other hand, exhibits significant spatiotemporal dynamic distribution characteristics. Its disordered charging behavior easily overlaps with the base load during evening peak hours, creating a "peak-on-peak" effect, leading to a series of problems such as voltage exceeding limits at distribution network nodes, transformer overload, and increased network losses.

[0003] Virtual power plants, by aggregating distributed resources such as photovoltaic (PV), energy storage systems, electric vehicles, and flexible loads, allow them to participate in grid dispatch and the electricity market like a unified power plant, representing an effective technical means to solve the aforementioned problems. However, within the flexible resources of a virtual power plant, traditional fixed PV-storage systems permanently install PV and energy storage equipment at specific nodes, and their deployment locations cannot be changed once determined. This static configuration method is difficult to adapt to the fluctuations in PV output due to seasonal and weather changes, and it cannot effectively respond to the dynamic spatial migration of electric vehicle charging loads. For example, in seasons with weak PV output or in areas with concentrated electric vehicle charging loads, fixed PV-storage systems, limited by their installation locations, often require long-distance power transmission or rely on supplemental power from the upper-level grid, leading to increased network losses and severe local voltage drops; while during periods of abundant PV output, fixed systems may experience curtailment due to insufficient local absorption capacity.

[0004] To address the aforementioned issues, the concept of mobile photovoltaic-storage systems has emerged in recent years. These systems install photovoltaic and energy storage units on mobile carriers, serving as dispatchable resources for virtual power plants, allowing for flexible adjustments to the grid connection location based on grid demand. Compared to fixed photovoltaic-storage systems, mobile systems offer several significant advantages: First, spatial flexibility, enabling the movement of photovoltaic power generation units to nodes with abundant sunlight or high load demand for localized consumption and voltage support. Second, temporal flexibility, allowing for adjustments to the connection location on a daily or even hourly basis based on seasonal changes, weather, and daily load curve variations, achieving a shift from "source follows load" to "source interacts with load." Third, improved economic efficiency, by optimizing the mobile path and connection nodes, reducing the system's dependence on long-distance transmission lines, minimizing grid losses and electricity purchase costs, and avoiding grid expansion and upgrades required due to localized overvoltage or undervoltage issues. However, current research on planning methods for mobile and flexible resources in virtual power plants is still incomplete, mainly suffering from the following technical defects: First, there is a lack of a two-layer planning model for virtual power plants that simultaneously considers the uncertainty of photovoltaic output and the spatiotemporal distribution of electric vehicle loads, with upper-layer site selection and capacity determination and lower-layer operation scheduling often being optimized separately; second, the operational constraints of mobile resources are not adequately characterized, such as the lack of systematic modeling of characteristics like movement costs and unique access locations for each time period; and third, existing solution algorithms are prone to getting trapped in local optima or experiencing slow convergence when dealing with planning problems involving high-dimensional, non-convex, and mixed variables (discrete site selection + continuous capacity). Summary of the Invention

[0005] The technical problem this invention aims to solve is: addressing the issues of existing virtual power plant resource allocation, such as the difficulty of fixed photovoltaic-storage systems adapting to spatiotemporal variations in source and load, high investment and operating costs, and significant voltage exceedance risks; and the imperfections and low solution efficiency of existing mobile resource planning methods. This invention provides a dynamic configuration and capacity planning method for virtual power plants where both photovoltaic and energy storage units can be deployed mobilely. To solve these problems, this invention adopts the following technical solution:

[0006] First, this invention proposes a method for mobile and flexible dynamic resource allocation and capacity planning in virtual power plants, comprising the following steps:

[0007] Acquire annual power output data of distributed photovoltaic power within the coverage area of ​​the virtual power plant, use K-means clustering algorithm to cluster the annual power output data, generate at least two typical photovoltaic power output scenarios, and determine the probability of occurrence of each typical photovoltaic power output scenario;

[0008] The probability distributions of return time, travel time, daily mileage, and initial state of charge of electric vehicles are obtained, and Monte Carlo simulation is used to generate disordered charging load curves for large-scale electric vehicles.

[0009] A two-layer optimization model for flexible resource dynamic allocation in a virtual power plant is constructed, consisting of an upper-layer model and a lower-layer model. The upper-layer model optimizes the access nodes and installed capacity of mobile photovoltaic units and mobile energy storage units with the minimum annualized investment cost, while the lower-layer model optimizes the actual output of photovoltaic power, charging and discharging power and node electricity price with the minimum annual operating cost, and constrains the access location of mobile resources to be adjustable within the scheduling cycle.

[0010] The improved Harris Eagle optimization algorithm is used to solve the bi-level optimization model. The improved Harris Eagle optimization algorithm includes an elite-preserving boundary processing mechanism and a mixed variable co-evolution mechanism.

[0011] Output the optimal access node location and installation capacity of the mobile photovoltaic unit and the mobile energy storage unit, as well as the corresponding annualized total cost.

[0012] Preferably, Monte Carlo simulation is used to generate disordered charging load curves for large-scale electric vehicles. The return time of electric vehicles follows a normal distribution, the travel time follows a normal distribution, the daily mileage follows a log-normal distribution, and the initial state of charge follows a normal distribution.

[0013] Preferably, the objective function of the upper-level model is to minimize the sum of photovoltaic and energy storage investment costs of all candidate access nodes under each typical scenario, wherein the investment cost is calculated by multiplying the unit capacity investment cost by the installed capacity; the constraint is that the installed capacity of photovoltaic and energy storage on each candidate node shall not exceed a preset upper limit.

[0014] Preferably, the objective function of the lower-level model is to minimize the annual operating cost, which includes operation and maintenance costs, electricity purchase costs from the upper-level grid, curtailment costs, and distribution network loss costs. The constraints of the lower-level optimization model include distribution network power flow balance constraints, upper and lower limits of voltage amplitude at each node, upper limits of current at each branch, lower limits of photovoltaic utilization, upper and lower limits of energy storage charging and discharging power, mutual exclusion constraints of energy storage charging and discharging states, and dynamic range constraints of energy storage state of charge.

[0015] Preferably, the lower limit constraint of photovoltaic utilization requires that the actual output of photovoltaic power is not less than a specified percentage of its theoretical maximum output, that is, not less than 90% of the output of the photovoltaic unit; the dynamic range constraint of energy storage state of charge requires that the energy storage state of charge always remain between the preset minimum and maximum values ​​during operation.

[0016] Preferably, the improved Harris Eagle optimization algorithm for solving the bi-level optimization model includes: initializing the population and algorithm parameters; calculating the fitness values ​​of individuals in the population to determine the current optimal individual; determining whether to enter the exploration or development phase based on prey escape energy and updating the population position; performing elite preservation boundary processing on individuals that have crossed the boundary after each iteration, resetting them to the position of the current globally optimal individual; using threshold binarization to achieve co-evolution of mixed variables for the location variables; repeating the iteration until the convergence condition is met and outputting the optimal solution.

[0017] Meanwhile, this invention proposes a mobile, flexible, dynamic resource allocation and capacity planning system for virtual power plants, comprising:

[0018] The scenario generation module is configured to perform the following process: acquire annual power output data of distributed photovoltaics within the coverage area of ​​the virtual power plant, cluster the annual power output data using the K-means clustering algorithm, generate at least two typical photovoltaic power output scenarios, and determine the probability of occurrence of each typical photovoltaic power output scenario;

[0019] The electric vehicle load prediction module is configured to perform the following process: obtain the probability distribution of return time, travel time, daily mileage and initial state of charge of electric vehicles, and generate disordered charging load curves of electric vehicles on a large scale using Monte Carlo simulation.

[0020] The model building module is configured to perform the following process: build a two-layer optimization model for flexible resource dynamic configuration of a virtual power plant, including an upper-layer model and a lower-layer model; the upper-layer model optimizes the access nodes and installed capacity of mobile photovoltaic units and mobile energy storage units with the minimum annualized investment cost, and the lower-layer model optimizes the actual output of photovoltaics, charging and discharging power and node electricity price with the minimum annual operating cost, while constraining the access location of mobile resources to be adjustable within the scheduling cycle;

[0021] The model solving module is configured to perform the following process: solve the bi-level optimization model using an improved Harris Eagle optimization algorithm, which includes an elite-preserving boundary treatment mechanism and a mixed-variable co-evolution mechanism;

[0022] The solution output module is configured to perform the following process: output the optimal access node location and installation capacity of the mobile photovoltaic unit and the mobile energy storage unit, as well as the corresponding annualized total cost.

[0023] Meanwhile, the present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the present invention.

[0024] Meanwhile, the present invention proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed, it implements the method described in the present invention.

[0025] Meanwhile, the present invention proposes a computer program product, including a computer program or computer-readable instructions, which, when executed by a processor, implement the method described in the present invention.

[0026] The present invention, by adopting the above technical solution, has the following beneficial effects:

[0027] (1) The mobile photovoltaic-storage system can actively migrate to the optimal access node according to the node electricity price and load distribution, effectively avoiding the network loss and electricity purchase cost caused by long-distance power transmission in the fixed system. Simulation results show that compared with the fixed scheme, the present invention can reduce the annualized total cost of the virtual power plant by more than 15%, significantly improve the economic efficiency of system investment and operation, and help expand the profit margin of operators.

[0028] (2) Mobile photovoltaic energy storage units can provide reactive power support and discharge support at nearby nodes with weak voltage, effectively overcoming the lack of flexibility caused by the fixed access location of fixed photovoltaic energy storage. Simulation results show that the node voltage over-limit rate drops significantly from 19.18% in the fixed scheme to 0.55%, effectively ensuring the voltage level of the distribution network and reducing the risk of equipment damage caused by voltage over-limit.

[0029] (3) The improved Harris Eagle optimization algorithm introduces an elite-preserving boundary processing mechanism and a mixed variable co-evolution mechanism. Compared with the particle swarm optimization algorithm and the standard Harris Eagle algorithm, it has better convergence characteristics, especially in the later stage of iteration, the global optimization ability is stronger, and it can obtain high-quality solutions in a reasonable time, which meets the computational efficiency requirements of engineering planning.

[0030] (4) The present invention adopts a dynamic configuration strategy based on days, which can flexibly adjust the access location of mobile resources according to the daily changes in photovoltaic output and load level, providing a practical technical solution for the daily operation of virtual power plants. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of a typical photovoltaic power output scenario.

[0032] Figure 2 This is a probability distribution diagram of the return trip time for electric vehicles.

[0033] Figure 3 This is a probability distribution diagram of electric vehicle travel times.

[0034] Figure 4A daily load curve showing the electric vehicle charging load superimposed with the base load.

[0035] Figure 5 This is a schematic diagram of the daily site selection node for a fixed photovoltaic energy storage system.

[0036] Figure 6 This is a graph showing the voltage change at nodes after a fixed photovoltaic-storage system is connected.

[0037] Figure 7 This is a map showing the annual site selection distribution of mobile photovoltaic energy storage systems.

[0038] Figure 8 This is a map showing the annual access capacity distribution of mobile photovoltaic energy storage systems.

[0039] Figure 9 This is a diagram showing the voltage amplitude distribution at nodes after the mobile photovoltaic-storage system is connected.

[0040] Figure 10 This is a comparison of the convergence curves of different optimization algorithms.

[0041] Figure 11 This is a schematic diagram of the virtual power plant system architecture of the present invention. Detailed Implementation

[0042] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0043] Example 1: A specific embodiment of the present invention is described in detail through the following steps:

[0044] S1. Data Acquisition and Photovoltaic Output Scenario Generation:

[0045] Historical output data of distributed photovoltaic (PV) power stations within the coverage area of ​​the virtual power plant were collected. Due to the strong randomness of PV output influenced by irradiance, the annual data volume, at hourly resolution, reached 8760 points, making direct use for planning calculations too computationally intensive. This embodiment employs the K-means clustering algorithm to reduce the number of scenarios in the annual PV output data. The number of clusters was set to 4, and four initial cluster centers were randomly selected. The Euclidean distance from each data point to the cluster center was calculated, and the closest points were grouped into the same cluster. Then, the mean of each cluster was recalculated as the new cluster center, iterating until the sum of squared errors converged. Finally, four typical PV output scenarios were obtained, and the proportion of days each scenario occurred throughout the year was calculated as the probability of occurrence for that scenario. Clustering revealed probabilities of 0.2356, 0.2685, 0.3507, and 0.1452 for the four scenarios, respectively. Their unit capacity PV output data are shown below. Figure 1 As shown.

[0046] By reducing the number of scenarios described above, this embodiment provides a typical operating scenario for the dynamic configuration of mobile photovoltaic storage on a daily basis. The mobile photovoltaic unit can flexibly select the access node according to the daily scenario type (sunny day, cloudy day). In scenarios with high photovoltaic output, it will be moved to the load center for local consumption first, while in scenarios with low output, it will be moved to the node with strong voltage support requirements, thus achieving the optimal matching between resource spatial distribution and source-load characteristics.

[0047] S2. Prediction of uncontrolled electric vehicle charging load on a large scale within a virtual power plant:

[0048] Based on data from the U.S. Department of Transportation's National Household Mobility Survey (NHTS), a probabilistic model of electric vehicle travel and charging behavior within a virtual power plant is established. Specifically, this includes:

[0049] (1) Probability model of return time:

[0050] To simplify this model, a day is divided into 24 time periods, assuming that the moment a vehicle begins charging is the moment its last trip of the day ends. Furthermore, this paper assumes sufficient charging resources and that vehicles do not need to wait in line to charge, focusing instead on the analysis of user charging behavior itself. The probability distribution function for the EV return trip is shown in equation (1).

[0051] (1)

[0052] In the formula, Indicates the return time; Let be the mean of this distribution. Let be the standard deviation of this distribution. The probability distribution of the EV return time is as follows: Figure 2 As shown.

[0053] (2) Travel time probability model:

[0054] The probability model for EV travel time is as follows: As shown:

[0055] (2)

[0056] In the formula, Indicates travel time; Let be the mean of this distribution. Let $\frac{ ... Figure 3 As shown.

[0057] Depend on Figure 2 , 3 It can be seen that the period from 16:00 to 21:00 corresponds to the highest probability of users starting charging when they return home from get off work. The period from 7:00 to 11:00 corresponds to EV users leaving for work, which aligns with people's daily travel patterns.

[0058] The return time probability density function and travel time data used in this embodiment are derived from travel records in the National Household Travel Survey (NHTS) of the United States.

[0059] (3) State of charge and daily mileage:

[0060] The State of Charge (SOC) of an EV is a crucial factor influencing user charging behavior, typically expressed as the remaining battery charge. Statistical data shows that when an EV begins charging, the battery's initial state of charge (SOC) is... This can be approximated as a normal distribution, and its probability distribution characteristics can be expressed as:

[0061] (3)

[0062] In the formula, Let be the standard deviation of this distribution. Let be the standard deviation of this distribution.

[0063] Daily mileage refers to the total distance an EV travels in a single day. This mileage is influenced by various factors, including driving behavior and road conditions. The probability distribution function for EV daily mileage is as follows:

[0064] (4)

[0065] In the formula, Let be the standard deviation of this distribution. Let be the standard deviation of this distribution.

[0066] EV manufacturers, grid operators, and urban planners can use statistical data such as EV charging patterns and travel models to provide crucial information for guiding orderly charging, ensuring the safe operation of the power grid, and constructing charging infrastructure.

[0067] Using the Monte Carlo simulation method, simulation parameters were first set, with 500 electric vehicles in the virtual power plant, each with a battery capacity of 35kWh, charging power of 7.2kW, and charging efficiency of 0.9. Then, random sampling was performed on each vehicle, drawing corresponding values ​​from the normal distributions of return time (mean 17.6, standard deviation 3.4), travel time (mean 8.2, standard deviation 3.2), daily mileage (mean 3.2, standard deviation 0.88), and initial state of charge (mean 0.6, standard deviation 0.2). Based on the daily mileage and initial state of charge, the required charging amount and charging time were calculated, thus obtaining the charging power curve of the vehicle within the charging time interval. Finally, the charging power curves of all 500 vehicles were superimposed over time to obtain the total unordered charging load curve of the large-scale electric vehicles. Simulation results show that disordered charging leads to a load peak between 18:00 and 21:00, which, when superimposed on the base load, results in a "peak-on-peak" phenomenon, posing a threat to the safe operation of the virtual power plant. Figure 4 As shown.

[0068] S3. Construct a virtual power plant two-layer optimization model unit:

[0069] The two-layer optimization model constructed in this embodiment consists of an upper layer (capacity planning layer) and a lower layer (operation scheduling layer), and the two layers are coupled through decision variables and cost functions.

[0070] Upper-level optimization model:

[0071] The objective function of the upper layer is to minimize the investment cost of MPVSS, that is, to minimize the initial investment in the construction and equipment purchase of the photovoltaic storage system, as shown in equations (5) and (6).

[0072] (5)

[0073] (6)

[0074] In the formula, The investment cost of the photovoltaic and energy storage system, Numbering is assigned to typical operating scenarios. ; Representative scenarios The corresponding probability of occurrence; The number of nodes; , They represent the first In the first scenario The first period Initial investment cost per unit capacity of photovoltaic and energy storage in each node; , These respectively indicate that they are installed in the first... In each scenario Time period The rated installed capacity of photovoltaics and energy storage in each node; τ is the present value to equivalent annual value coefficient; This represents the discount rate.

[0075] To meet actual needs, the photovoltaic power generation capacity and energy storage capacity must be constrained. The constraints on the photovoltaic power generation capacity and energy storage capacity are shown in equation (7).

[0076] (7)

[0077] In the formula: , Representing the candidate nodes respectively Photovoltaic and energy storage capacity installed on the roof; , Representing the candidate nodes respectively The maximum allowed photovoltaic capacity and energy storage capacity.

[0078] Lower-level optimization model:

[0079] The lower-level objective function aims to minimize the system operating costs, including operation and maintenance costs, electricity purchase costs, curtailment costs, and grid loss costs, as shown in equations (8) to (14):

[0080] (8)

[0081] Annual operating and maintenance costs of MPVSS This includes maintenance costs for photovoltaic and energy storage equipment.

[0082] (9)

[0083] In the formula, and Representing the scene respectively Middle node In time period The active power output of PV and ES, when When ES discharges, This indicates that the ES is charging; and These represent the unit operating costs of PV and ES, respectively.

[0084] Electricity purchase cost is the annual cost of purchasing electricity from the power grid. , refers to the cost of purchasing electricity from the grid when photovoltaic power generation is insufficient. As shown in equations (10) to (12).

[0085] (10)

[0086] (11)

[0087] (12)

[0088] In the formula, For nodes In time The unit cost of purchasing electricity; For nodes In time The equivalent load value; For nodes The baseline active power requirement; For nodes In time The base load value; For nodes In time period The charging load of electric vehicles; This is the time-of-use electricity price factor; This is the equivalent load factor; , , These represent the electricity purchase prices during off-peak, normal, and peak hours in the time-of-use electricity pricing system. , , These represent off-peak electricity periods, average electricity periods, and peak electricity periods, respectively. The unit of time is the scheduling unit.

[0089] Cost of abandonment This refers to the economic losses caused by the failure to effectively utilize excess photovoltaic power generation.

[0090] (13)

[0091] In the formula, This represents the cost coefficient for wasted light.

[0092] Network loss cost It is the annualized cost of network losses in the distribution network.

[0093] (14)

[0094] In the formula, This refers to the annual power loss of the power distribution network. Cost per unit of loss.

[0095] To ensure the safe and stable operation of the power system, system current constraints, node voltage constraints and branch current constraints are introduced, as shown in equations (15) to (17).

[0096] (15)

[0097] (16)

[0098] (17)

[0099] In the formula, and Representative scenarios The corresponding distribution network node The active and reactive power injected at the point; and Representative scenarios The corresponding distribution network node and nodes The voltage amplitude; and This is the node admittance matrix of the distribution network; For the scene The corresponding node With nodes The phase angle difference; Number the branch roads; This represents the total number of branches in the distribution network; branch road The corresponding current; branch road The maximum allowable current.

[0100] To ensure the safe operation and effective utilization of mobile photovoltaic and mobile energy storage devices within their physical capabilities, the following key constraints need to be added:

[0101] (1) Constraints on photovoltaic utilization rate:

[0102] To ensure the efficient utilization of photovoltaic resources, it is required that, in any operating scenario, time period, and access node, the actual active power output of photovoltaics shall not be less than 90% of its theoretical maximum output, as shown in equation (18):

[0103]

[0104] In the formula, For the scene Time period internal nodes The maximum power that photovoltaic theory can generate; The minimum utilization factor for photovoltaic power output is 0.9.

[0105] (2) Energy storage charging and discharging power constraints

[0106] The charging and discharging power of an energy storage system is limited by the rated power of its converter, and charging and discharging cannot be performed simultaneously, as shown in equation (19):

[0107] (19)

[0108] In the formula, and Installed on the nodes respectively The maximum charging power and maximum discharging power of an energy storage system are usually determined by the energy storage capacity planned at the upper level. , Let them be binary decision variables, representing the scenarios respectively. Time period ,node Is the energy storage device in a charging or discharging state?

[0109] (3) Dynamic constraints on the state of charge of energy storage

[0110] (20)

[0111] (twenty one)

[0112] In the formula, In the scene Time period ,node The energy storage state of charge; and These are the charging and discharging efficiencies of energy storage, respectively. and These represent the minimum and maximum states of charge that energy storage can tolerate, respectively.

[0113] S4, Improved Harris Eagle Optimization Algorithm Solving Unit:

[0114] To effectively solve the two-layer optimization model constructed in this embodiment, an intelligent optimization algorithm with strong global search capabilities and high convergence accuracy is required. Since the Harris Hawks Optimization (HHO) algorithm can effectively balance global exploration and local exploitation capabilities, and given the high-dimensionality, non-convexity, and multi-constraint characteristics of the model in this embodiment, using the HHO algorithm can effectively avoid premature convergence and improve the accuracy and stability of the solution. Therefore, this embodiment adopts and improves the HHO algorithm.

[0115] The Harris Eagle Optimization Algorithm (HHO), proposed by Heidari et al. in 2019, is inspired by the different encirclement strategies employed by Harris eagles based on the prey's state during the hunt. In HHO, the optimal solution to the problem is defined as the prey, and each individual in the population represents a Harris eagle. The core of the algorithm is to update the individual's position by simulating the encirclement process of the prey by the eagle flock, thereby gradually approaching the global optimum. The entire hunting process can be divided into two main phases: the exploration phase and the development phase, the switching of which is determined by the prey's escape energy E.

[0116] (1) Exploration phase (|E|≥1):

[0117] When prey is energetic (|E|≥1), the Harris Eagle will extensively explore its search space to locate potential prey areas. Its position update formula is as follows:

[0118] (twenty two)

[0119] (twenty three)

[0120] In the formula, and The first generation and first The position vector of the individual; This represents the currently known optimal solution, i.e., the prey's location; For the first An individual randomly selected from the population; This represents the average position of all individuals in the current population. , , , , All are random numbers within the interval [0, 1]. and These are the upper and lower bounds of the search space, respectively.

[0121] This phase employs two strategies to ensure the population fully explores the global scope and avoids premature convergence.

[0122] (2) Development stage ):

[0123] When the prey is exhausted ( When the prey is in a vulnerable state, the Harris Eagle will enter the attack phase, launching a tight encirclement. The outcome will depend on whether the prey still has the ability to escape (determined by a random number). (Judgment), the development phase is further subdivided into four besiege strategies:

[0124] Soft Besiege: When and At that time, the prey still has a strong ability to escape, and the eagles adopt a strategy of surrounding but not attacking immediately, as shown in equations (24) and (25).

[0125] (twenty four)

[0126] (25)

[0127] In the formula, ; The random jump intensity of the prey; A random number between 0 and 1.

[0128] Hard Besiege: When and At that time, the prey is almost powerless to escape, and the flock of eagles launches a direct attack on its current location, as shown in equation (26).

[0129] (26)

[0130] Soft / hard encirclement with Levy in flight: When At any time, regardless of the prey's energy level, the eagle flock will execute a sudden dive attack, a process simulated by Levy flight. Levy flight is a random walk with long-distance jump characteristics, which can effectively help the algorithm escape local optima traps, as shown in equations (27) to (29).

[0131] (27)

[0132] (28)

[0133] (29)

[0134] In the formula, It refers to the location after being surrounded by soft or hard confinement; This is the final update location; It is a random step size vector; For Levy flight functions; From the perspective of the problem dimension. Final location. exist and Choose the one with better adaptability.

[0135] Throughout the process, the prey's escape energy and jump intensity It changes dynamically with the number of iterations, as shown in the equation. As shown.

[0136] (30)

[0137] In the formula, Initial energy; A random number within the interval [-1, 1]; This represents the current iteration number; This represents the maximum number of iterations. The absolute value gradually decreases from greater than 1 to less than 1, naturally realizing the transition from exploration to development.

[0138] Although the standard Harris Hawks Optimization (HHO) algorithm boasts superior performance, it still suffers from two shortcomings when handling complex optimization problems with high dimensions, multiple constraints, and mixed variables, as described in this embodiment: first, it may prematurely converge in the later stages of iteration due to insufficient population diversity; second, a simple boundary truncation strategy wastes search information when individuals cross boundaries. To overcome these problems, this embodiment proposes the following improvements to construct the IHHO (Improved Harris Hawks Optimization) algorithm:

[0139] (1) Boundary handling mechanism based on elite preservation:

[0140] In the standard HHO, if the updated position Exceeding The range is typically forced to be set as a boundary value. This approach ignores the current optimal solution. The high-quality information contained therein. This embodiment draws on relevant research ideas and adopts an elite preservation strategy: when an individual goes out of bounds, it is directly reset to the current optimal solution. The mathematical expression is:

[0141] (31)

[0142] In the formula, For individual indexes; This is used as a dimension index. This strategy can effectively utilize the current best information to guide individuals that have exceeded the limit, accelerate convergence, and improve the quality of the solution.

[0143] (2) Mixed variable co-evolutionary mechanism:

[0144] The decision variables in this model include discrete site selection variables. and capacity variable , This also includes continuous variables such as changes in energy storage SOC. To enable HHO to effectively handle such mixed variables, this paper uses a threshold method for co-evolution. Specifically, taking site selection variables as an example, Within the algorithm, it participates in evolution as a continuous variable, with a value range of [0,1]. Before each fitness calculation, a threshold is used... Binarize it: If If the selected site is chosen, the capacity variable evolves as a continuous variable. Continuous variables in the model directly participate in the optimization. This method embeds discrete decision-making into a continuous optimization framework and implements discretization constraints on the capacity variable, ensuring the smoothness of the algorithm and the feasibility of the solution.

[0145] Algorithm solution process:

[0146] Based on the above theory and improvements, the specific process of using the IHHO algorithm to solve the two-layer optimization model described in this embodiment is as follows:

[0147] (1) Set the parameters required by the algorithm, including population size, maximum number of iterations, etc., and initialize the population;

[0148] (2) Calculate the fitness value of the Harris Eagle population according to formula (22), sort the calculation results, and determine the optimal individual location of the prey;

[0149] (3) Use equation (30) to calculate the escape energy of the prey and equation (25) to calculate the jump distance of the prey during its escape.

[0150] (4) When the prey enters the local mining stage, determine whether the prey's escape energy E is greater than 1. If it is, return to formula (22) for calculation; otherwise, continue to determine whether E is greater than 0.5.

[0151] (5) Update the population position according to the four besieging strategies.

[0152] (6) Determine whether the convergence condition is met, i.e., the maximum number of iterations. Is it less than the current iteration number? If the condition is met, the iteration ends and the optimal position and fitness value of the Harris Eagle population are output; if the convergence condition is not met, the calculation is restarted by returning to equation (22).

[0153] S5, Output Virtual Power Plant Configuration Scheme Unit:

[0154] Through the above solution process, the optimal configuration scheme of the virtual power plant is output, including: the access node location and installed capacity of each photovoltaic unit, the access node location and installed capacity of each energy storage unit, the annualized total cost of the virtual power plant (including investment cost and operating cost), and the node voltage distribution at each time period of a typical day.

[0155] Specifically, this embodiment uses the IEEE 33-node distribution network system as the access network for the virtual power plant. The simulation platform is Matlab 2016a. There are 15 candidate nodes for photovoltaic (PV) and energy storage, namely nodes 2, 3, 6, 10, 12, 15, 17, 18, 20, 26, 27, 29, 30, 31, and 32. The unit installed capacity of the PV unit is 0.05 MW, and the unit installed capacity of the energy storage unit is 0.01 MW. The maximum number of PV units that can be connected is 20, and the maximum number of energy storage units that can be connected is 10. The time-of-use pricing is set as follows: Off-peak period (0:00 to 6:00): 0.4 yuan per kilowatt-hour; Normal period (6:00 to 12:00 and 18:00 to 24:00): 0.8 yuan per kilowatt-hour; Peak period (12:00 to 18:00): 1.2 yuan per kilowatt-hour. The penalty cost for curtailment of solar power is 1.5 yuan per kilowatt-hour. The unit investment cost of photovoltaic power is 8,000 yuan per kilowatt, and the unit investment cost of energy storage is 12,700 yuan per kilowatt. The operation and maintenance costs of photovoltaic power and energy storage are 0.1 yuan per kilowatt and 0.05 yuan per kilowatt, respectively. The charge and discharge efficiency of energy storage is 0.9, with a lower limit of 0.1 and an upper limit of 0.9 for the state of charge (SOC). At the end of the dispatch cycle, the SOC must be restored to between 0.7 and 0.9. The lower limit requirement for photovoltaic utilization is that the actual output should not be less than 90% of the theoretical maximum output.

[0156] Table 1 shows the site selection, capacity configuration, and annualized total cost of FPVSS (Fixed Photovoltaic Storage System). Figure 5 This is the addressing node for FPVSS. Figure 6 This shows the node voltage distribution of the distribution network at 18:00 after the FPVSS was connected. Due to the large number of unordered electric vehicle charging loads connected to this system, significant voltage exceedances would occur without considering the FPVSS connection. After the FPVSS was connected, the system voltage distribution improved significantly due to the influence of photovoltaic output and energy storage. However, due to the still high uncertainty in photovoltaic output, voltage exceedances still occurred for approximately 70 days throughout the year, with an exceedance rate of 19.18%.

[0157] Table 1

[0158] PV site selection 17 ES Location 18 PV capacity (kW) 650 ES capacity (kW) 50 Total cost (RMB) 52688314.57

[0159] Table 2 shows the site selection, capacity configuration, and total daily cost of mobile optical storage, taking a single day as an example:

[0160] Table 2

[0161] PV site selection 18 ES Location 12 PV capacity (kW) 850 ES capacity (kW) 90 Daily cost (RMB) 122270.78

[0162] Figure 7This shows the annual distribution of photovoltaic and energy storage unit access nodes in MPVSS (Mobile Photovoltaic Storage System). Figure 7 It is known that considering the large-scale integration of electric vehicles, the system is under heavy load, and voltage overshooting is likely to occur around node 18. To ensure the voltage quality of the distribution network, connecting the photovoltaic output units in the MPVSS to the vicinity of node 18 can effectively reduce the load level of this node and alleviate the voltage drop problem caused by long-distance power transmission. The connection node of the energy storage unit can be flexibly adjusted according to the daily load level and photovoltaic output, thereby achieving economic optimization. Figure 8 The figure shows the annual grid connection capacity of photovoltaic and energy storage units in MPVSS. As can be seen from the figure, the grid connection capacity of photovoltaic mainly fluctuates between 500-900kW, while the grid connection capacity of energy storage is mainly between 20-80kW.

[0163] Table 3 compares the annualized costs of fixed and mobile photovoltaic-storage systems:

[0164] Table 3

[0165] Cost (RMB) 52688314.57 44609650.65 8078663.92

[0166] Because the mobile photovoltaic-storage system in this embodiment can be flexibly scheduled on a daily basis, the location and capacity of photovoltaic and energy storage units can be adjusted according to different photovoltaic output, load levels, and nodal tariffs, thereby minimizing its own operating costs. Simulations over an annual period show that the annual operating cost of MPVSS is 44,609,650.65, a 15.33% reduction compared to FPVSS. From a long-term perspective, MPVSS can effectively reduce operating costs and has high economic value.

[0167] Considering the scenario of a large number of electric vehicle charging loads being connected, the MPVSS connects to selected nodes according to the site selection scheme and analyzes the node voltage changes. The results are as follows: Figure 9 As shown.

[0168] like Figure 9 As shown, compared to FPVSS, which can only be fixedly connected to a single node, MPVSS exhibits significantly improved voltage distribution at 18:00 due to the uncertainty of photovoltaic output and the large-scale voltage exceedance caused by EV charging load. Simulation analysis shows that, over an annual period, MPVSS experiences voltage exceedances on only two days, with a voltage exceedance rate of 0.55%, a significant improvement compared to FPVSS. Experimental results indicate that when the distribution network is under heavy operational pressure, MPVSS can provide more flexible power support, ensuring the safe and reliable operation of the system.

[0169] To illustrate the superiority of the IHHO used, this embodiment employs PSO (Particle Swarm Optimization, a swarm intelligence algorithm that simulates the foraging behavior of bird flocks, guiding particle position updates through individual and global optima), HHO, and IHHO to solve the MPVSS location and capacity problem described in this chapter. The convergence curves of the three algorithms are shown below. Figure 10 As shown in the figure, compared with the traditional Harris Eagle algorithm and particle swarm optimization algorithm, the improved Harris Eagle algorithm has better convergence characteristics, especially in the later stages of iteration. IHHO can escape local optima, thus having a superior global optimization capability, indicating that it can find a better scheduling scheme.

[0170] Finally, the optimal configuration scheme for the virtual power plant was output. The results show that mobile photovoltaic units are mainly deployed near node 18, with daily capacity fluctuating between 500 kW and 900 kW; mobile energy storage units have more flexible access nodes, allowing selection of different nodes based on the day's photovoltaic output and load level, with daily capacity ranging from 20 kW to 80 kW. The annualized total cost of the virtual power plant is 44,609,650.65 yuan, a 15.33% reduction compared to the fixed photovoltaic-energy storage scheme (52,688,314.57 yuan). Voltage analysis results show that the fixed photovoltaic-energy storage scheme experiences voltage overruns on 70 days of the year (overrun rate 19.18%), while the scheme of this invention only experiences voltage overruns on 2 days of the year (overrun rate 0.55%). During the peak load period at 18:00, the voltage of all nodes in the scheme of this invention is above 0.95 pu, while the voltage of the fixed scheme near node 18 has dropped below 0.90 pu. This conclusion demonstrates that the mobile and flexible resource dynamic allocation method proposed in this invention has high engineering practical value.

[0171] Example 2: This example proposes a mobile, flexible, dynamic resource allocation and capacity planning system for virtual power plants, including:

[0172] The scenario generation module is configured to perform the following process: acquire annual power output data of distributed photovoltaics within the coverage area of ​​the virtual power plant, cluster the annual power output data using the K-means clustering algorithm, generate at least two typical photovoltaic power output scenarios, and determine the probability of occurrence of each typical photovoltaic power output scenario;

[0173] The electric vehicle load prediction module is configured to perform the following process: obtain the probability distribution of return time, travel time, daily mileage and initial state of charge of electric vehicles, and generate disordered charging load curves of electric vehicles on a large scale using Monte Carlo simulation.

[0174] The model building module is configured to perform the following process: build a two-layer optimization model for flexible resource dynamic configuration of a virtual power plant, including an upper-layer model and a lower-layer model; the upper-layer model optimizes the access nodes and installed capacity of mobile photovoltaic units and mobile energy storage units with the minimum annualized investment cost, and the lower-layer model optimizes the actual output of photovoltaics, charging and discharging power and node electricity price with the minimum annual operating cost, while constraining the access location of mobile resources to be adjustable within the scheduling cycle;

[0175] The model solving module is configured to perform the following process: solve the bi-level optimization model using an improved Harris Eagle optimization algorithm, which includes an elite-preserving boundary treatment mechanism and a mixed-variable co-evolution mechanism;

[0176] The solution output module is configured to perform the following process: output the optimal access node location and installation capacity of the mobile photovoltaic unit and the mobile energy storage unit, as well as the corresponding annualized total cost.

[0177] Figure 11 This is a schematic diagram of the system architecture of the MPVSS virtual power plant. The system mainly consists of three parts: mobile photovoltaic (PV), mobile energy storage, and EVs. The entire system meets the charging needs of EV users through the coordinated operation of the power grid, PV power generation equipment, and energy storage equipment. Solar energy is generated through PV modules, and the power generation equipment is installed on a mobile carrier, giving the system high flexibility and making it suitable for application scenarios in different locations. The electricity generated by the mobile PV can be directly used to charge EVs or base loads, or it can be transmitted to the power grid or mobile energy storage equipment for energy transportation or storage. The mobile energy storage equipment, through coordinated operation with the power grid and PV power generation modules, can provide stable power output when load demand is high or PV power generation is insufficient, thereby improving the reliability of the entire system. This paper aims to achieve global optimization of the system through upper and lower layer linkage. The decision variables of the upper layer are the location and capacity of PV and energy storage, and the optimization objective is to minimize the system investment cost; the decision variables of the lower layer are the actual output of PV and the charging and discharging of energy storage, and the objective function is to minimize the system operating cost. First, typical photovoltaic (PV) data is obtained through K-means clustering based on annual PV output data. Then, the maximum PV output is determined using the PV capacity in the upper-level model's decision variables. Considering the overall system load demand, not all of this PV output will be utilized; after accounting for curtailment, the actual PV output is obtained. The nodal price is calculated using the actual PV output, energy storage charging and discharging, and system load, thus determining the system operating cost. Finally, the lower-level model returns the system operating cost to the upper-level decision variables to optimize resource allocation. Through this feedback mechanism, the upper and lower levels collaborate to optimize, ultimately achieving the optimal total system cost.

[0178] Example 3: This example proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in this invention.

[0179] Example 4: This example proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the steps of the method described in this invention.

[0180] Example 5: This example proposes a computer program product, including a computer program or computer-readable instructions, which, when executed by a processor, implement the method described in this invention.

[0181] It should be noted that the processing flow of embodiments 2-5 corresponds to the specific steps of the method provided in embodiment 1 of the present invention, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the method provided in embodiment 1 of the present invention.

[0182] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0183] The specific implementation schemes described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific implementation schemes of the present invention and are not intended to limit the scope of the present invention. Any equivalent changes and modifications made by those skilled in the art without departing from the concept and principles of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for mobile, flexible, dynamic resource allocation and capacity planning in a virtual power plant, characterized in that, Includes the following steps: Acquire annual power output data of distributed photovoltaic power within the coverage area of ​​the virtual power plant, use K-means clustering algorithm to cluster the annual power output data, generate at least two typical photovoltaic power output scenarios, and determine the probability of occurrence of each typical photovoltaic power output scenario; The probability distributions of return time, travel time, daily mileage, and initial state of charge of electric vehicles are obtained, and Monte Carlo simulation is used to generate disordered charging load curves for large-scale electric vehicles. A two-layer optimization model for flexible resource dynamic allocation in a virtual power plant is constructed, consisting of an upper-layer model and a lower-layer model. The upper-layer model optimizes the access nodes and installed capacity of mobile photovoltaic units and mobile energy storage units with the minimum annualized investment cost, while the lower-layer model optimizes the actual output of photovoltaic power, charging and discharging power and node electricity price with the minimum annual operating cost, and constrains the access location of mobile resources to be adjustable within the scheduling cycle. The improved Harris Eagle optimization algorithm is used to solve the bi-level optimization model. The improved Harris Eagle optimization algorithm includes an elite-preserving boundary processing mechanism and a mixed variable co-evolution mechanism. Output the optimal access node location and installation capacity of the mobile photovoltaic unit and the mobile energy storage unit, as well as the corresponding annualized total cost.

2. The method according to claim 1, characterized in that, Monte Carlo simulation was used to generate disordered charging load curves for large-scale electric vehicles. The return time of electric vehicles follows a normal distribution, the travel time follows a normal distribution, the daily mileage follows a log-normal distribution, and the initial state of charge follows a normal distribution.

3. The method according to claim 1, characterized in that, The objective function of the upper-level model is to minimize the sum of photovoltaic and energy storage investment costs of all candidate access nodes under each typical scenario, where the investment cost is calculated by multiplying the unit capacity investment cost by the installed capacity; the constraint is that the installed capacity of photovoltaic and energy storage on each candidate node shall not exceed the preset upper limit.

4. The method according to claim 1, characterized in that, The objective function of the lower-level model is to minimize the annual operating cost, which includes operation and maintenance costs, electricity purchase costs from the upper-level grid, curtailment costs, and distribution network loss costs. The constraints of the lower-level optimization model include distribution network power flow balance constraints, upper and lower limits of voltage amplitude at each node, upper limits of current at each branch, lower limits of photovoltaic utilization, upper and lower limits of energy storage charging and discharging power, mutual exclusion constraints of energy storage charging and discharging states, and dynamic range constraints of energy storage state of charge.

5. The method according to claim 4, characterized in that, The photovoltaic utilization rate lower limit constraint requires that the actual output of photovoltaic power is not less than a specified percentage of its theoretical maximum output, that is, not less than 90% of the output of the photovoltaic unit; the energy storage state of charge dynamic range constraint requires that the energy storage state of charge always remain between the preset minimum and maximum values ​​during operation.

6. The method according to claim 1, characterized in that, The improved Harris Eagle optimization algorithm for solving the bi-level optimization model includes: initializing the population and algorithm parameters; calculating the fitness values ​​of individuals in the population to determine the current optimal individual; determining whether to enter the exploration or development phase based on prey escape energy and updating the population position; performing elite preservation boundary processing on individuals that have crossed the boundary after each iteration, resetting them to the position of the current globally optimal individual; using threshold binarization for the location variables to achieve co-evolution of mixed variables; repeating the iteration until the convergence condition is met and outputting the optimal solution.

7. A mobile, flexible, dynamic resource allocation and capacity planning system for virtual power plants, characterized in that: include: The scenario generation module is configured to perform the following process: acquire annual power output data of distributed photovoltaics within the coverage area of ​​the virtual power plant, cluster the annual power output data using the K-means clustering algorithm, generate at least two typical photovoltaic power output scenarios, and determine the probability of occurrence of each typical photovoltaic power output scenario; The electric vehicle load prediction module is configured to perform the following process: obtain the probability distribution of return time, travel time, daily mileage and initial state of charge of electric vehicles, and generate disordered charging load curves of electric vehicles on a large scale using Monte Carlo simulation. The model building module is configured to perform the following process: build a two-layer optimization model for flexible resource dynamic configuration of a virtual power plant, including an upper-layer model and a lower-layer model; the upper-layer model optimizes the access nodes and installed capacity of mobile photovoltaic units and mobile energy storage units with the minimum annualized investment cost, and the lower-layer model optimizes the actual output of photovoltaics, charging and discharging power and node electricity price with the minimum annual operating cost, while constraining the access location of mobile resources to be adjustable within the scheduling cycle; The model solving module is configured to perform the following process: solve the bi-level optimization model using an improved Harris Eagle optimization algorithm, which includes an elite-preserving boundary treatment mechanism and a mixed-variable co-evolution mechanism; The solution output module is configured to perform the following process: output the optimal access node location and installation capacity of the mobile photovoltaic unit and the mobile energy storage unit, as well as the corresponding annualized total cost.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed, it implements the method as described in any one of claims 1-6.

10. A computer program product comprising a computer program or computer-readable instructions, characterized in that, When the computer program or computer-readable instructions are executed by a processor, they implement the method as described in any one of claims 1 to 6.

Images