Optical storage collaborative robust planning method based on improved Bayesian optimization

Abstract

The invention discloses an optical storage collaborative robust planning method based on improved Bayesian optimization. According to the method, a combined site selection and capacity configuration planning model of the photovoltaic unit and the energy storage unit is constructed, a photovoltaic extreme attack scene is introduced, and specific description of operation risks is realized; a heat engine capacity adjusting mechanism based on photovoltaic equivalent load bearing capacity is adopted, heat engine redundancy is reduced, and carbon emission is reduced; multiple targets are processed by using random weights based on Dirichlet distribution, and diversified exploration of a complex decision space is realized; through Monte Carlo sampling calculation expectation improvement of a random forest agent with noise, a search space is controllably reduced, and simulation overhead is reduced; the optical storage configuration scheme is optimized and identified based on the annual simulation result and multi-target performance evaluation, the problem that reliability, economical efficiency and low-carbon performance are difficult to consider in traditional planning is solved, and efficient robust planning under complex and uncertain conditions is achieved.

Description

Technical Field

[0001] This invention relates to the field of power system planning, and specifically to a robust planning method for photovoltaic-storage collaborative planning based on improved Bayesian optimization. Background Technology

[0002] The cost of photovoltaic (PV) power generation has continued to decline, leading to large-scale and rapid deployment in recent years. Statistics show that global PV installations approximately tripled between 2018 and 2023, with utility-scale PV accounting for the majority of new installations. While PV resources are abundant and offer significant emission reduction benefits, their output is heavily influenced by weather conditions, resulting in considerable uncertainty. Furthermore, as PV systems are increasingly integrated with smart devices, their communication and control links are exposed to more complex network environments, increasing the potential for attacks and potentially impacting system stability.

[0003] To enhance system flexibility and reliability, energy storage systems can serve as an important regulation tool. Joint planning with photovoltaic (PV) systems can help improve grid performance while reducing carbon emissions. However, existing PV-storage collaborative planning methods typically assume a relatively stable operating environment and do not adequately consider extreme situations such as PV output fluctuations, load uncertainties, and network attacks, leading to insufficient robustness of the planning schemes in actual operation.

[0004] On the other hand, traditional planning and optimization methods often face problems such as high computational load and low convergence efficiency in multi-objective, strongly coupled, and simulation-driven computational scenarios, making it difficult to obtain a planning scheme that balances reliability, economy, and security within a reasonable timeframe. Therefore, there is an urgent need for a robust optical-storage collaborative planning method that can simultaneously handle uncertainty and attack risks and has efficient solution capabilities, in order to improve the operational security and planning quality of the system in complex environments. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing photovoltaic-storage collaborative planning methods in handling photovoltaic output uncertainty, load fluctuations, and potential photovoltaic attacks, such as insufficient robustness, low solution efficiency, and difficulty in simultaneously considering power supply reliability and economy. This invention proposes a photovoltaic-storage collaborative robust planning method based on improved Bayesian optimization.

[0006] The objective of this invention is achieved through the following technical solution: a robust planning method for photovoltaic-storage collaborative planning based on improved Bayesian optimization, the method comprising: By acquiring grid data, establishing basic operation models for photovoltaic, energy storage, heat engines, and grid power flow, quantifying the effective load carrying capacity of photovoltaic, reducing heat engine capacity based on the effective load carrying capacity of photovoltaic, modeling the strongest photovoltaic attack from time and space perspectives and integrating it into the basic operation model, and taking the minimization of unmet electricity demand, line loss, investment cost, and carbon emission penalty as the objective function, a photovoltaic and energy storage collaborative site selection and capacity optimization model is obtained. A two-layer solution structure is used for the photovoltaic energy storage collaborative site selection and capacity optimization model. The lower layer runs simulation, and the upper layer uses random forest as a surrogate model to perform Bayesian optimization, so as to obtain the installation location and installed capacity of photovoltaic energy storage in the power grid.

[0007] Furthermore, the power grid data includes load data of each bus of the power grid, line parameters, original heat engine configuration, typical photovoltaic fluctuation parameters for one year, and typical load fluctuation parameters; the typical photovoltaic fluctuation parameters for one year are modeled based on the daily and annual fluctuations of photovoltaic power generation; the typical load fluctuation parameters for one year are modeled based on the load fluctuation data of the power grid for the most recent year.

[0008] Furthermore, the basic operating model for photovoltaic, energy storage, heat engine, and power grid flow includes the following constraints: The capacity and installation location constraints of photovoltaic and energy storage units must meet the physical feasibility constraints of thermal engine constraints, line power flow constraints, and bus power balance constraints.

[0009] Furthermore, the quantification of the effective load-carrying capacity of photovoltaic power, and the reduction of heat engine capacity based on the effective load-carrying capacity of photovoltaic power, specifically involves defining a weight proportional to the load. The effective load carrying capacity of photovoltaic power is calculated by weighted averaging of photovoltaic output values. The modeling formula is as follows: For photovoltaic Rated power, For a moment The proportion of photovoltaic power generation available at that time according to The reduction ratio of the rated capacity of the heat engine The calculation formula is as follows, and based on... The new capacity of the heat engine is obtained, and the modeling formula is as follows: in For heat engine The original rated power, For heat engine The new rated power, It is a heat engine assembly.

[0010] Furthermore, the integration of the strongest photovoltaic attack modeling from both temporal and spatial perspectives into the basic operating model specifically includes: in the time dimension, filtering out moments within a year when both load and photovoltaic output are simultaneously high, and using parameters... and The adjustment determines the unique attack moment. This allows for coverage of the most destructive, high-risk periods; spatially, it combines nodalism and photovoltaic capacity to select key photovoltaic units to form an attack ensemble. The strongest attack model is as follows: in busbar Last moment The actual load, For the middle and upper part of the annual load The value, The proportion of photovoltaic power generation available throughout the year is among the top The value, For the attack time set, Indicates photovoltaic Is it installed on the busbar? , busbar The degree, It is the collection of all busbars in the power grid. It is a photovoltaic array.

[0011] Furthermore, the objective function is specifically: in, The objective function is the annual unmet electricity demand. busbar Last moment The amount of load shedding, Let the objective function be the line loss. For the line At any moment The meritorious trend, For the line The resistance, This is the system's power reference value; The objective function is cost. Cost per unit of photovoltaic capacity For photovoltaic Rated power, The unit cost of the energy storage power component. The unit cost of the energy storage component. For energy storage units Rated energy capacity, This is the carbon emission penalty coefficient. For heat engine The new rated power, For time sets, It is the collection of all busbars in the power grid. For the collection of routes, It is a heat engine assembly.

[0012] Furthermore, the lower-level simulation in the two-layer solution structure specifically includes: running simulations based on the candidate solutions selected in each iteration of Bayesian optimization to obtain the annual running results, and then calculating the objective function.

[0013] Furthermore, the Bayesian optimization process employs a random weighting mechanism based on the Dirichlet distribution, with average weights used in the initialization phase, specifically as follows: in, Represents the m-th tree in a random forest M. The random weight vectors generated are uniformly distributed on the multidimensional probability simplex and sum to 1. This is the normalized objective function.

[0014] Furthermore, the Bayesian optimization iteration process directly uses the random forest surrogate model Sampling from the predicted distribution To estimate the desired acquisition function, noise is introduced into the surrogate model. To simulate the uncertainties of photovoltaics and load, specifically: Random forest is composed of Composed of trees, For the first Tree pairs of candidate points The predicted value, This is the historically optimal target value. Used to measure the current candidate point Compared to Potential improvements This is the solution for the next iteration.

[0015] According to another aspect of the specification, an improved Bayesian-optimized optical-storage collaborative robust planning device is also provided, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the improved Bayesian-optimized optical-storage collaborative robust planning method.

[0016] The beneficial effects of this invention are: This invention introduces extreme photovoltaic attack modeling, photovoltaic effective load carrying capacity (ELCC) calculation, and heat engine capacity adjustment mechanism, so that the constructed planning model can maintain safe operation and reliable power supply even under drastic changes in photovoltaic output or even under attack. This invention employs an improved multi-objective Bayesian optimization algorithm, utilizing random weights based on the Dirichlet distribution to process multiple objectives and achieve diversified exploration of complex decision spaces. By improving the expectation calculation through Monte Carlo sampling with a noisy random forest agent, the high-dimensional search space is controlled within a computable range, reducing simulation overhead. Even under conditions of high simulation costs, it can still effectively search for optimal site selection and capacity determination schemes for photovoltaics and energy storage, thereby significantly improving solution efficiency and obtaining high-quality, diversified Pareto solution sets.

[0017] Based on the simulation results throughout the year and multi-objective performance evaluation, the photovoltaic and energy storage configuration schemes are optimized and identified, which solves the problem that it is difficult to balance reliability, economy and low carbon in traditional planning, and realizes efficient and robust planning under complex and uncertain conditions. Attached Figure Description

[0018] Figure 1 A flowchart of a robust planning method for optical-storage collaborative planning based on improved Bayesian optimization is provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] In this invention, a photovoltaic unit refers to an important renewable energy power generation unit installed on a busbar of the power grid, whose actual output is determined by the rated capacity and the proportion of available sunlight.

[0022] An energy storage unit refers to an electrochemical energy storage device that has the ability to charge and discharge and regulates energy through the state of charge (SOC), including power capacity, energy capacity and efficiency parameters.

[0023] Attack scenarios refer to extreme operating conditions in which photovoltaic units experience a sudden drop in output at critical moments due to human error or abnormal events. These scenarios are used to assess the robustness of planning schemes in high-risk environments.

[0024] like Figure 1 As shown in the figure, this embodiment provides a robust planning method for optical-storage collaborative planning based on improved Bayesian optimization, with the following steps: Step 1: Obtain raw power grid data. This data includes load data for each busbar, line parameters, original thermal engine configuration, typical annual photovoltaic fluctuation parameters, and typical load fluctuation parameters. This data is used to construct the basic power grid model, providing a consistent data framework for subsequent simulations and optimizations.

[0025] Step two: Based on the original power grid data, and under preset constraints, a photovoltaic energy storage collaborative site selection and capacity optimization model is constructed with the objective function of minimizing unmet power demand, line losses, investment costs, and carbon emission penalties. This model mainly includes five steps: (2.1) Construct the combined decision space of photovoltaic and energy storage units. The specific implementation details are as follows: Photovoltaic capacity, installation location, and available output are all incorporated into the optimization decision, and the modeling formula is as follows: in For photovoltaic At any moment The actual power generation, For photovoltaic Rated power, For a moment The proportion of photovoltaic power generation available at that time Indicates photovoltaic Is it installed on the busbar? , It is the collection of all busbars in the power grid.

[0026] Energy storage capacity configuration, power specifications, location selection, and operating status are all involved in joint optimization, and the modeling formula is as follows; in For energy storage At any moment The actual discharge (charge) capacity, For energy storage Rated power, For energy storage Rated discharge (charge) duration, For energy storage Rated energy capacity, For energy storage At any moment The state of charge, For energy storage charging efficiency, For the discharge efficiency of energy storage, To store energy at the state of charge at the start of the simulation, To store energy at the state of charge at the end of the simulation, Indicates energy storage Is it installed on the busbar? .

[0027] (2.2) The strongest photovoltaic attack is modeled from both temporal and spatial perspectives. The specific implementation details are as follows: In terms of time, the system filters out periods within a year when both load and photovoltaic output are relatively high, using parameters... and The adjustment determines the unique attack moment. This allows for coverage of the most destructive, high-risk periods; spatially, it combines nodalism and photovoltaic capacity to select key photovoltaic units to form an attack ensemble. The strongest attack model is as follows: in busbar Last moment The actual load, For the middle and upper part of the annual load The value, The proportion of photovoltaic power generation available throughout the year is among the top The value, For load quantile parameters, This is the quantile parameter for the proportion of photovoltaic power generation available. For time sets, This is a collection of key high-risk moments. busbar The degree, It is a photovoltaic array.

[0028] (2.3) An ELCC approximation calculation method based on load hour weight is adopted, and the specific implementation details are as follows: Define weights that are proportional to the load. The photovoltaic output value is calculated by weighted average. ; according to The reduction ratio of the rated capacity of the heat engine The calculation formula is as follows, and based on... The new capacity of the heat engine is obtained: in For heat engine The original rated power, For heat engine The new rated power, It is a heat engine assembly.

[0029] (2.4) Establish physical feasibility-satisfying thermal engine constraints, line power flow constraints, and bus power balance constraints. The specific implementation details are as follows: a) The active power output of each heat engine must be within its operating limits: in For heat engine At any moment The actual power generation, and These are the rate of increase and rate of decrease of thermal power generation, respectively. and These represent the minimum and maximum permissible output ratios of the heat engine, respectively.

[0030] b) The modeling of line power flow constraints and bus power balance constraints is as follows: in For the line At any moment The meritorious trend, For the line The equivalent admittance, and The lines are respectively The starting and ending busbars at time phase angle, For the line The maximum permissible current.

[0031] in busbar Last moment The amount of load shedding, , and They represent the busbars respectively. The collection of heat engines, photovoltaic and energy storage units on the surface and Representing the busbar A set of routes with a destination and a starting point.

[0032] (2.5) Define four objective functions: annual unmet electricity demand, line losses, investment costs, and carbon emission penalties. The specific implementation details are as follows: in For the line The resistance, This is the system's power reference value. The investment cost per unit of photovoltaic capacity, The unit cost of the energy storage power component. The unit cost of the energy storage component. This is the penalty coefficient for carbon emissions.

[0033] Step 3 involves solving the photovoltaic energy storage collaborative site selection and capacity optimization model based on an improved multi-objective Bayesian algorithm to obtain the installation location and capacity of photovoltaic energy storage in the power grid. This mainly includes three steps: (3.1) A two-layer solution structure is established. The upper layer uses Bayesian optimization with a random forest surrogate function, and the lower layer performs simulation calculations based on the candidate solutions selected in each iteration of Bayesian optimization. Running simulation Calculate the annual operating results Then calculate the objective function. The specific implementation details are as follows: The upper-layer surrogate model approximates the complex photovoltaic-storage collaborative planning model by learning from existing simulation results. This allows for rapid prediction of candidate solution performance indicators without needing to call real simulations every time, thus guiding the search direction. The lower-layer simulation module verifies the candidate solutions recommended by the surrogate function through real-world operation to obtain accurate system operation data, including photovoltaic output, energy storage charge / discharge status, thermal engine output, line power flow, and potential load shedding. The mathematical formulas for multi-objective optimization, decision variables, and simulation output are as follows: (3.2) The optimization phase adopts a random weighting mechanism based on Dirichlet distribution, while the initialization phase uses average weights. The specific implementation details are as follows: in The initial weight vector, The weighted objective function constructed with the initial weights. The random weight vectors generated are uniformly distributed on the multidimensional probability simplex and sum to 1. This is the normalized objective function.

[0034] (3.3) Directly from the Random Forest Proxy Model Sampling from the predicted distribution To estimate the desired acquisition function, noise is introduced into the surrogate model. To simulate the uncertainties of photovoltaics and load, the specific implementation details are as follows: Random forest is composed of Composed of trees, For the first Tree pairs of candidate points The predicted value, It follows a Gaussian distribution. For the first Tree prediction value The first sample drawn from the predicted distribution centered on One sample, For each tree, the number of Monte Carlo samples used to improve computation is desired. This is the historically optimal target value. Used to measure the current candidate point Compared to Potential improvements This is the set of candidate points used to search for the solution in the next iteration. This is the solution for the next iteration.

[0035] In summary, this application provides a robust planning method for photovoltaic (PV) and energy storage (ESS) collaboration based on improved Bayesian optimization. First, a joint site selection and capacity configuration planning model for PV and ESS units is constructed, introducing extreme PV attack scenarios. A thermal engine capacity adjustment mechanism based on the PV equivalent load carrying capacity is adopted to reduce thermal engine redundancy and lower carbon emissions. Multi-objective processing is handled using random weights based on Dirichlet distribution. The expectation is improved through Monte Carlo sampling calculation with a noisy random forest proxy, controllably reducing the search space and simulation overhead. Based on year-round simulation results and multi-objective performance evaluation, the PV-ESS configuration scheme is optimized and identified, solving the problem of balancing reliability, economy, and low carbon emissions in traditional planning. Ultimately, efficient and robust planning under complex and uncertain conditions is achieved.

[0036] Corresponding to the aforementioned embodiment of a robust planning method for photovoltaic-storage collaborative planning based on improved Bayesian optimization, the present invention also provides an embodiment of a robust planning device for photovoltaic-storage collaborative planning based on improved Bayesian optimization.

[0037] See Figure 2 The present invention provides a robust planning device for optical-storage collaborative planning based on improved Bayesian optimization, comprising a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement a robust planning method for optical-storage collaborative planning based on improved Bayesian optimization as described in the above embodiments.

[0038] The embodiment of the optical-storage collaborative robust planning device based on improved Bayesian optimization provided by this invention can be applied to any device with data processing capabilities, such as a computer or other similar device. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 2 The diagram shown is a hardware structure diagram of any data processing-capable device, including a robust optical-storage collaborative planning device based on improved Bayesian optimization provided by the present invention. (Except for...) Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0039] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0040] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0041] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a robust planning method for optical-storage collaborative operation based on improved Bayesian optimization as described in the above embodiments.

[0042] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0043] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned robust planning method for optical-storage collaborative planning based on improved Bayesian optimization.

[0044] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0045] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. This application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A robust planning method for photovoltaic-storage collaborative planning based on improved Bayesian optimization, characterized in that, The method includes: By acquiring grid data, establishing basic operation models for photovoltaic, energy storage, heat engines, and grid power flow, quantifying the effective load carrying capacity of photovoltaic, reducing heat engine capacity based on the effective load carrying capacity of photovoltaic, modeling the strongest photovoltaic attack from time and space perspectives and integrating it into the basic operation model, and taking the minimization of unmet electricity demand, line loss, investment cost, and carbon emission penalty as the objective function, a photovoltaic and energy storage collaborative site selection and capacity optimization model is obtained. A two-layer solution structure is used for the photovoltaic energy storage collaborative site selection and capacity optimization model. The lower layer runs simulation, and the upper layer uses random forest as a surrogate model to perform Bayesian optimization, so as to obtain the installation location and installed capacity of photovoltaic energy storage in the power grid.

2. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The power grid data includes load data of each busbar of the power grid, line parameters, original heat engine configuration, typical photovoltaic fluctuation parameters for one year, and typical load fluctuation parameters. The typical photovoltaic fluctuation parameters for one year are modeled based on the daily and annual fluctuations of photovoltaic power generation. The typical load fluctuation parameters for one year are modeled based on the load fluctuation data of the power grid for the most recent year.

3. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The basic operating model for photovoltaic, energy storage, heat engine, and power grid flow includes the following constraints: The capacity and installation location constraints of photovoltaic and energy storage units must meet the physical feasibility constraints of thermal engine constraints, line power flow constraints, and bus power balance constraints.

4. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The quantification of the effective load-carrying capacity of photovoltaic power, and the reduction of heat engine capacity based on the effective load-carrying capacity of photovoltaic power, specifically involves: Define weights that are proportional to the load. The effective load carrying capacity of photovoltaic power is calculated by weighted averaging of photovoltaic output values. The modeling formula is as follows: For photovoltaic Rated power, For a moment The proportion of photovoltaic power generation available at that time according to The reduction ratio of the rated capacity of the heat engine The calculation formula is as follows, and based on... The new capacity of the heat engine is obtained, and the modeling formula is as follows: in For heat engine The original rated power, For heat engine The new rated power, It is a heat engine assembly.

5. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The integration of the modeling of the strongest photovoltaic attack from both temporal and spatial perspectives into the basic operation model specifically includes: in the time dimension, screening for moments within a year when both load and photovoltaic output are simultaneously high, and using parameters... and The adjustment determines the unique attack moment. This allows for coverage of the most destructive, high-risk periods; spatially, it combines nodalism and photovoltaic capacity to select key photovoltaic units to form an attack ensemble. The strongest attack model is as follows: in busbar Last moment The actual load, For the middle and upper part of the annual load The value, The proportion of photovoltaic power generation available throughout the year is among the top The value, For the attack time set, Indicates photovoltaic Is it installed on the busbar? , busbar The degree, It is the collection of all busbars in the power grid. It is a photovoltaic array.

6. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The objective function is specifically: in, The objective function is the annual unmet electricity demand. busbar Last moment The amount of load shedding, Let the objective function be the line loss. For the line At any moment The meritorious trend, For the line The resistance, This is the system's power reference value; The objective function is cost. Cost per unit of photovoltaic capacity For photovoltaic Rated power, The unit cost of the energy storage power component. The unit cost of the energy storage component. For energy storage units Rated energy capacity, This is the carbon emission penalty coefficient. For heat engine The new rated power, For time sets, It is the collection of all busbars in the power grid. For the collection of routes, It is a heat engine assembly.

7. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The lower-level simulation in the two-layer solution structure specifically includes: running simulations based on candidate solutions selected in each iteration of Bayesian optimization to obtain the annual running results, and then calculating the objective function.

8. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The Bayesian optimization process employs a random weighting mechanism based on the Dirichlet distribution, with average weights used in the initialization phase, specifically as follows: in, Represents the m-th tree in a random forest M. The random weight vectors generated are uniformly distributed on the multidimensional probability simplex and sum to 1. This is the normalized objective function.

9. The robust planning method for optical-storage collaborative planning based on improved Bayesian optimization according to claim 1, characterized in that, The Bayesian optimization iteration process directly uses the random forest proxy model Sampling from the predicted distribution To estimate the desired acquisition function, noise is introduced into the surrogate model. To simulate the uncertainties of photovoltaics and load, specifically: Random forest is composed of Composed of trees, For the first Tree pairs of candidate points The predicted value, This is the historically optimal target value. Used to measure the current candidate point Compared to Potential improvements This is the solution for the next iteration.

10. A robust planning device for optical-storage collaborative planning based on improved Bayesian optimization, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that... When the processor executes the executable code, it implements a robust planning method for optical-storage collaborative planning based on improved Bayesian optimization as described in any one of claims 1-9.

Images