Light storage building day-ahead optimization method, device and equipment considering multiple uncertainties

Abstract

The invention provides a light storage building day-ahead optimization method, device and equipment considering multiple uncertainties, and relates to the technical field of energy scheduling. The method comprises the steps of establishing a carbon emission intensity uncertainty set, a photovoltaic output uncertainty set and an electrical load uncertainty set for a light storage building cluster, and constructing a carbon-source-load uncertainty set; correlation and time sequence feature analysis is carried out on the carbon-source-load uncertainty set, and sampling is carried out according to an analysis result to obtain candidate intra-day scenes; screening the candidate intra-day scenes according to the scene occurrence probability, the operation cost sensitivity and the carbon emission influence degree to obtain a key scene; and performing two-stage robust optimization according to the carbon emission intensity grade corresponding to the day-ahead predicted value of the carbon emission intensity of the optical storage building cluster and the key scene to obtain a day-ahead optimization scheduling result of the optical storage building cluster. According to the invention, economical, safe and low-carbon operation of the light storage building in a complex and changeable environment can be realized.

Description

Technical Field

[0001] This invention relates to the field of energy dispatching technology, and in particular to a day-ahead optimization method, apparatus and equipment for photovoltaic-storage buildings that takes into account multiple uncertainties. Background Technology

[0002] As a significant source of energy consumption and carbon emissions, the building sector is facing an urgent need for low-carbon transformation. Building-integrated photovoltaic (BIPV) systems, by deeply integrating distributed photovoltaic power generation, energy storage devices, and adjustable loads, can not only absorb clean energy locally but also flexibly participate in grid regulation, representing an effective technological path to improve building energy efficiency and reduce carbon emissions.

[0003] However, during the current optimization and scheduling phase, the operation of photovoltaic-storage building systems faces the combined impact of multiple uncertainties: photovoltaic output fluctuates significantly due to weather conditions, building load exhibits strong temporal variability and unpredictability, and grid carbon emission intensity dynamically fluctuates with changes in the power generation structure and system operating status. Most existing studies only consider the uncertainties of photovoltaics and load in their optimization models, treating grid carbon emission intensity as a fixed value, failing to reflect the coupling relationship and synchronous changes among these three factors in actual operation. This approach often leads to a significant increase in the deviation between prediction and reality when facing complex scenarios with multiple uncertainties, resulting in problems such as high electricity purchase prices, curtailment of photovoltaic power, and excessive carbon emissions, increasing system operating costs and weakening low-carbon benefits. Furthermore, traditional robust optimization methods generally employ independent modeling and full-scenario worst-case strategies when dealing with uncertainties. While these can guarantee a certain safety margin, they easily introduce excessive conservatism, causing the operating strategy to fail to fully realize its economic potential in most real-world scenarios. At the same time, these methods typically lack differentiated response mechanisms for carbon emission levels, are unable to proactively optimize power purchase and sales and energy storage strategies based on predicted carbon intensity in the day-ahead phase, and lack the ability to coordinate operation and resource complementarity scheduling among cross-regional photovoltaic-storage building clusters.

[0004] Therefore, there is an urgent need for a technical solution that can simultaneously consider the triple uncertainties of photovoltaic output, building load and carbon emission intensity during the day-ahead optimization stage, so as to achieve economic, safe and low-carbon operation of photovoltaic-storage building clusters in complex and ever-changing environments. Summary of the Invention

[0005] This invention provides a day-ahead optimization method, apparatus, and equipment for photovoltaic-storage buildings that considers multiple uncertainties, in order to solve the problem that day-ahead optimization scheduling of photovoltaic-storage buildings is difficult to achieve economic, safe, and low-carbon operation in complex and variable environments.

[0006] In a first aspect, embodiments of the present invention provide a day-ahead optimization method for photovoltaic-storage buildings that considers multiple uncertainties, including: For photovoltaic-storage building clusters, uncertain sets of carbon emission intensity, photovoltaic output, and electricity load are established, and these sets are incorporated into the same joint distribution to construct a carbon-source-load uncertainty set. Correlation and temporal characteristic analysis were performed on the carbon-source-load uncertainty set, and candidate intraday scenes were obtained by sampling based on the analysis results; The candidate intraday scenarios were screened based on the probability of occurrence, sensitivity to operating costs, and the degree of impact on carbon emissions to obtain key scenarios; Based on the carbon emission intensity level corresponding to the day-ahead forecast of the carbon emission intensity of the photovoltaic-storage building cluster and the key scenarios, a two-stage robust optimization is performed to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster.

[0007] Secondly, embodiments of the present invention provide a day-ahead optimization device for photovoltaic-storage buildings that considers multiple uncertainties, comprising: The model building module is used to establish uncertain sets of carbon emission intensity, photovoltaic output, and electricity load for photovoltaic-storage building clusters, and to incorporate the uncertain sets of photovoltaic output, electricity load, and carbon emission intensity into the same joint distribution to construct a carbon-source-load uncertain set. The scene generation module is used to perform correlation and time series feature analysis on the carbon-source-load uncertainty set, and to sample and obtain candidate intraday scenes based on the analysis results; The scenario screening module is used to screen the candidate intraday scenarios based on the probability of scenario occurrence, sensitivity to operating costs, and degree of carbon emission impact to obtain key scenarios; The day-ahead optimization module is used to perform two-stage robust optimization based on the carbon emission intensity level corresponding to the day-ahead predicted value of the carbon emission intensity of the photovoltaic-storage building cluster and the key scenario, so as to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster.

[0008] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0009] In this embodiment of the invention, for photovoltaic-storage building clusters, uncertain sets of carbon emission intensity, photovoltaic output, and electricity load are first established, and these sets are incorporated into the same joint distribution to construct a carbon-source-load uncertain set. Then, correlation and time-series characteristic analysis are performed on the carbon-source-load uncertain set, and candidate intraday scenarios are obtained by sampling based on the analysis results. Next, the candidate intraday scenarios are screened based on the scenario occurrence probability, operating cost sensitivity, and carbon emission impact to obtain key scenarios. Finally, a two-stage robust optimization is performed based on the carbon emission intensity level corresponding to the day-ahead predicted value of the photovoltaic-storage building cluster's carbon emission intensity and the key scenarios to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster. This enables economical, safe, and low-carbon operation of photovoltaic-storage buildings in complex and variable environments. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the implementation of the pre-housing-photovoltaic building optimization method considering multiple uncertainties provided in this embodiment of the invention. Figure 2 This is a flowchart illustrating the implementation of a pre-hospitalization optimization method for photovoltaic-storage buildings that considers multiple uncertainties, provided in another embodiment of the present invention. Figure 3 This is a topology diagram of the IEEE 33-node system provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the day-ahead forecast of photovoltaic output of a photovoltaic-storage building cluster provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the day-ahead predicted total load value of IEEE 33 nodes provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the day-ahead optimized scheduling results of photovoltaic-storage building considering multiple uncertainties provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the time-shiftable load power adjustment of the photovoltaic-storage building cluster at node 25 provided in an embodiment of the present invention; Figure 8 This is a schematic diagram illustrating the time-shifted load peak shaving and valley filling effect of the photovoltaic-storage building cluster provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of the photovoltaic-storage building-ahead optimization device that considers multiple uncertainties provided in the embodiments of the present invention; Figure 10 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0011] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0012] In recent years, with the rapid development of distributed photovoltaic (PV) and energy storage technologies, the application of PV-storage buildings in building energy conservation and low-carbon operation has been continuously expanding. Related research often uses current-day optimization models to uniformly schedule PV, energy storage, and building loads to improve clean energy consumption and reduce operating costs. Regarding uncertainty handling, the mainstream approach focuses on PV and load forecasting errors, employing polyhedral uncertainty sets, probabilistic constraints, or scenario simulations for optimization solutions, thereby enhancing the tolerance of plans to deviations to some extent. However, existing technologies generally ignore the time-varying and uncertain nature of grid carbon emission intensity, approximating it as a fixed value or historical average, failing to reflect the dynamic fluctuations under changes in power generation structure and system operating status. Even when some work introduces carbon costs, they are mostly superimposed on the economic cost model as additional parameters, lacking joint modeling with PV output and building load, making it difficult to reveal the physical coupling and statistical dependence among the three. In terms of robust optimization, a common strategy is to independently model each uncertainty and perform worst-case processing for all scenarios. While this provides a safety margin, it is prone to over-conservatism, leading to economic losses in most real-world scenarios. Simultaneously, the lack of a differentiated response mechanism based on carbon intensity prediction makes it difficult to proactively optimize energy storage charging and discharging, movable load transfer, and power purchase and sale plans at the day-ahead stage to avoid the use of high-carbon electricity. At the system level, research often focuses on single-region operations, with insufficient joint modeling and coordinated scheduling of cross-regional uncertainties, failing to fully leverage the flexibility and low-carbon potential brought by meteorological differences and resource complementarity. In summary, existing technologies suffer from insufficient joint correlation modeling, inadequate utilization of carbon emission intensity uncertainty, an imbalance between robustness and economics, and a lack of cross-regional coordination when dealing with the "triple uncertainty" of photovoltaic output, building load, and carbon emission intensity. These issues directly restrict the ability of photovoltaic-storage buildings to achieve economical, safe, and low-carbon operation in complex and variable environments. There is an urgent need to develop a comprehensive technical path at the day-ahead optimization stage that can simultaneously characterize the correlation and joint fluctuations of these three factors, introduce carbon emission intensity-driven adaptive scheduling, key scenario screening and dynamic adjustment mechanisms, and multi-regional coordinated optimization. These shortcomings directly limit the ability of photovoltaic-storage building clusters to achieve both economic efficiency and low carbon emissions in complex and variable environments. Therefore, this invention proposes an embodiment.

[0013] See Figure 1 The flowchart illustrates the implementation of the pre-housing-photovoltaic building optimization method considering multiple uncertainties provided in this embodiment of the invention, which is described in detail below: Step 101: For photovoltaic-storage building clusters, establish uncertain sets of carbon emission intensity, photovoltaic power output, and electricity load, and incorporate these sets into the same joint distribution to construct a carbon-source-load uncertain set.

[0014] In this embodiment, the various devices involved in the photovoltaic-storage building cluster are first modeled: 1. Photovoltaic power generation model The output of photovoltaic (PV) power generation is significantly affected by various environmental factors, including sunlight intensity, operating mode, and environmental conditions. Given the physical parameters of the PV array components, these factors collectively determine the output characteristics of the PV power generation equipment. These output characteristics can be expressed as: ; In the formula: This represents the actual light intensity radiated onto the surface of the photovoltaic array at time t. Under standard test conditions (usually) The ratio of the light intensity to the standard light intensity reflects the strength of the current lighting conditions relative to the standard lighting conditions. Indicates the rated power of photovoltaic power generation equipment. This represents the output power of the photovoltaic power generation equipment at time t; This is the derating factor for photovoltaic power generation equipment, used to indicate the degree of reduction in actual output relative to rated output; This indicates the operating temperature of photovoltaic power generation equipment under standard conditions. This represents the operating temperature of the photovoltaic power generation equipment at time t, in units of... .

[0015] 2. Distribution Network Power Interaction Model When the power generation equipment within the distribution network cannot meet load demand, electricity needs to be purchased from the distribution network to maintain supply and demand balance. Conversely, if the electricity generated by photovoltaic-storage buildings exceeds local demand, the surplus electricity can be sold back to the distribution network. This not only helps balance the supply and demand of the entire power system but also brings certain economic benefits. During power exchange, the following constraints must be met: ; In the formula: Let t represent the power purchased by the photovoltaic-storage building from the distribution network at time t. The power sold to the distribution network by the photovoltaic-storage building at time t: Let be the power of the normal load at time t; Let be the power of the controllable load at time t. Where: ; ; In the formula: This is a binary variable. A value of "1" indicates that electricity is purchased from the distribution network at time t, and a value of "0" indicates that electricity is sold to the distribution network at time t. This indicates the maximum allowable power consumption for both purchasing and selling.

[0016] The power exchange cost of a distribution network can be expressed as: ; In the formula: This represents the cost of exchanging data with the distribution network at time t; Let t be the day-ahead trading price of the distribution network.

[0017] 3. Photovoltaic-storage building load model On the load side, there are numerous time-shiftable loads that can interact amicably with the power grid. These loads possess a certain degree of flexibility and can be adjusted or transferred within a certain time frame to meet the system's operational requirements. Their model can be represented as follows: ; ; In the formula: This indicates that the time-shiftable load has adjustable power at time t; Indicates in Total electricity demand of time-shiftable loads within a given time period; This represents the lower limit of the adjustable power of the time-shiftable load at time t. This represents the upper limit of the adjustable power of the time-shiftable load at time t.

[0018] The control cost of a time-shiftable load at time t can be expressed as: ; In the formula: This represents the control cost of the time-shiftable load at time t; This represents the compensation coefficient for time-shiftable loads; This represents the original electricity consumption plan for time-shiftable loads. The formula contains absolute values, is a non-linear function, and introduces two non-negative auxiliary variables. , Consider this as a linear programming problem: ; And the following constraints must be met: ; ; 4. Energy Storage Model Energy storage devices can absorb excess electrical energy when there is overcapacity and release electrical energy when the power supply cannot meet the load demand, thus playing a role in "peak shaving and valley filling." Ignoring the self-discharge effect of the energy storage device, the stored energy during the charging and discharging process can be expressed as: ; In the formula: This represents the amount of electricity stored by the energy storage device at time t. This indicates the amount of electricity stored by the energy storage device at the initial moment, expressed in kWh. This represents the charging power of the energy storage device at time t. These represent the discharge power of the energy storage device at time t; This represents the charging efficiency coefficient of energy storage devices. This represents the discharge efficiency coefficient of the energy storage device; For time intervals.

[0019] To avoid overcharging and over-discharging during the operation of energy storage devices, the following constraints must be met: (1) Charge and discharge state constraints To prevent energy storage devices from charging and discharging simultaneously, charging / discharging state logic constraints must be met. At the same time, to extend the lifespan of energy storage devices and improve their economic efficiency, the number of charging / discharging state transitions also needs to be limited.

[0020] ; In the formula: This indicates the charging state of the energy storage device at time t. These two variables represent the discharge state of the energy storage device at time t. Both are 0-1 variables. When the energy storage battery is charging... =1, otherwise The value is 1.

[0021] ; ; In the formula: This represents the upper limit of the number of charge / discharge state transitions. Since the charge / discharge state constraints contain absolute values, they are linearized by introducing an auxiliary variable: ; ; The original constraints can then be transformed into: ; ; (2) Charging / discharging power constraints ; ; In the formula: The upper limit of charging power for energy storage devices. This represents the upper limit of the discharge power of the energy storage device.

[0022] (3) Charge state constraints To prevent energy storage devices from being overcharged or over-discharged, state of charge constraints must be met: ; In the formula: This refers to the upper limit of the amount of electricity that an energy storage device can store. This represents the lower limit of the amount of electricity that the energy storage device can store. To ensure the stability and efficiency of the energy storage device during the cyclic control process, it is generally desirable that the remaining capacity of the energy storage device is equal at the beginning and end of the control process. ; In the formula: This represents the control period, which is 24 hours in day-ahead control. The charging and discharging cost of the energy storage device at time t. It can be represented as: ; In the formula: This indicates the cost coefficient of energy storage equipment.

[0023] Next, we will conduct carbon emission analysis and modeling: Due to the intermittent and fluctuating nature of renewable energy generation, balancing the supply and demand of the power system becomes more difficult, further exacerbating the uncertainty of carbon emission intensity. Therefore, considering the uncertainty of carbon emission intensity is particularly important, and it is necessary to conduct relevant optimization operation studies to ensure system stability. To facilitate the quantitative analysis of the distribution characteristics and mechanisms of energy networks in the power system, the carbon emission flow method is used to analyze the carbon emissions of the power system.

[0024] (1) Carbon emission intensity model Carbon emission intensity is an indicator that measures the amount of carbon emissions transferred per unit of electricity exchanged. It reflects the carbon emission efficiency of a power system during power generation. The calculation of carbon emission intensity is typically based on factors such as the power generation structure of the power system, the carbon emission factors of various power sources, and the amount of electricity generated. The specific calculation method is as follows: ; In the formula: This represents the carbon emission intensity of a photovoltaic-storage building's grid-connected node, in units of... ; This represents the carbon flow rate from the branch to this node, in units of... ; This represents the active power flowing to this node (mainly considering that carbon emissions in the power system are primarily related to active power flow). This represents the grid loss rate of electricity supplied to the grid from the generation point to the grid connection point in fossil fuel energy. This indicates the grid loss rate of clean energy supplied to the grid from the generation point to the grid connection point; This indicates the mass of fossil fuels consumed to supply one unit of electricity to the power grid. It represents the amount of carbon emissions produced per unit mass of fossil fuel consumption.

[0025] (2) Uncertainty analysis of carbon emissions Due to the randomness and volatility of renewable energy sources such as photovoltaics, the composition of electricity sources in photovoltaic-storage buildings exhibits a certain degree of uncertainty. (Photovoltaic predicted power) The volatility and randomness of carbon emissions make carbon emission intensity an uncertain parameter, meaning it inherits the uncertainty characteristics of new energy output. Among various methods for representing uncertain sets, the polyhedral uncertain set, with its unique linear structure and ease of uncertainty control, has become the preferred method. Its linearity makes solving the model more efficient and better able to handle real-world uncertainties, specifically expressed as: ; In the formula: Represents an uncertain set of carbon emission intensity; This represents the predicted carbon emission intensity of the power system at time t; This represents the deviation between the predicted and actual values ​​of electricity carbon emission intensity at time t; An uncertain adjustment parameter representing the intensity of carbon emissions from electricity.

[0026] (3) Carbon emission cost model An initial carbon emission allowance allocation method based on power generation capacity is adopted. This method determines the initial carbon emission allowance for each power generation unit through a specific mathematical expression. The expression for the initial carbon emission allowance is as follows: ; In the formula: This represents the initial carbon emission allowance for the distribution network at time t; This represents the emission allocation per unit of electricity, in units of... ; This represents the carbon emission allocation per unit heat output of a micro gas turbine. This represents the electrical power output of the micro gas turbine at time t.

[0027] The carbon emissions of a power distribution network can be expressed as: ; In the formula: This represents the carbon emissions of the power distribution network at time t; =0.6101 represents the carbon emission coefficient of a micro gas turbine.

[0028] In summary, the carbon trading cost of the distribution network at time t is: ; In the formula: Let be the carbon trading cost of the distribution network at time t; The carbon trading price is taken here. .

[0029] Building upon this foundation, this paper proposes a two-stage robust optimization method for photovoltaic (PV) power output uncertainty, building load uncertainty, and grid carbon emission intensity uncertainty simultaneously present in day-ahead optimization of PV-storage buildings. This method is based on multi-uncertainty correlation analysis, a carbon intensity tiered adaptive strategy, key scenario screening and self-calibration, and multi-regional collaboration. In the modeling stage, this method comprehensively characterizes the physical correlations and statistical dependencies of the three types of uncertainties. In the solution stage, it introduces scenario tiering and dynamic adjustment mechanisms to ensure the optimization results are robust while improving economic efficiency and low-carbon emissions. The source-load uncertainty analysis includes: The model's uncertainty is mainly affected by the uncertainty of photovoltaic output and the uncertainty of electricity consumption fluctuations. To describe and analyze this problem, mathematical modeling is required. The uncertain sets of photovoltaic output and electricity load can be expressed as follows: ; ; ; In the formula: Represents the polyhedral uncertain set of photovoltaics and loads; This represents the set of uncertain photovoltaic output. Represents an uncertain set of electrical loads; This represents the predicted photovoltaic output at time t. This represents the predicted electrical load at time t; This represents the deviation between the actual and predicted photovoltaic output at time t. This represents the deviation between the actual and predicted electrical load values ​​at time t. This represents the maximum deviation between the actual and predicted photovoltaic output at time t. This represents the maximum deviation between the actual and predicted electrical load values ​​at time t. This represents the uncertain adjustment parameter of photovoltaics. This represents the uncertain adjustment parameter of the electrical load.

[0030] Then, the polyhedral uncertainty set of photovoltaics and loads, together with the uncertainty set of carbon emission intensity, constitutes the carbon-source-load uncertainty set. The carbon-source-load uncertainty set is then incorporated into a subproblem in a two-stage robust optimization model to solve for the minimum operating cost under the worst-case scenario, ensuring the stability of power supply while controlling carbon emissions.

[0031] Step 102: Perform correlation and time series characteristic analysis on the carbon-source-load uncertainty set, and obtain candidate intraday scenarios by sampling based on the analysis results.

[0032] Combination Figure 2In this embodiment, unlike traditional methods that model photovoltaic, load and carbon emission intensity separately and assume they are independent of each other, this embodiment extracts the synchronous fluctuation patterns of the three on the same time scale through joint analysis of historical forecast data and actual operating data. It also constructs a set of scenarios that can reflect the real physical process by combining meteorological conditions, power system operating characteristics and changes in carbon emission source structure.

[0033] In the scenario generation phase, prediction residuals of photovoltaic power output, building load, and carbon emission intensity are collected to establish their respective probability distributions. The Copula method is then used to uniformly characterize the correlation and temporal characteristics among the three. Based on this, a large number of possible intraday scenarios are generated using a sampling method. The residuals are then superimposed onto the prediction curve to obtain specific scenario curves. Abnormal combinations that do not conform to physical laws are eliminated, ultimately forming a candidate intraday scenario set.

[0034] Specifically, the process begins by using historical "prediction-actual" paired data to jointly analyze three types of uncertainties—photovoltaic power output, building load, and grid carbon emission intensity—on the same time scale. Tools like Copula are used to characterize their correlations and temporal characteristics. This step yields a joint distribution model, rather than a single scenario. This joint distribution simultaneously reflects the interrelationships between the three variables (e.g., low sunlight periods are often accompanied by high load and high carbon intensity) and their temporal fluctuation patterns. Next, sampling is performed within this joint distribution framework, and the resulting residuals are superimposed onto the prediction curve, generating a large number of possible intraday scenarios. The correlations and temporal characteristics determine the "probability distribution basis for candidate scenario generation." Where to sample, which combinations are high-probability, and which are almost impossible are all constrained by this analysis. The candidate intraday scenario set is the result of sampling and screening based on this distribution.

[0035] Step 103: Select candidate intraday scenarios based on scenario occurrence probability, operating cost sensitivity, and carbon emission impact to obtain key scenarios.

[0036] In one embodiment, step 103 includes: For each scenario in the candidate day, calculate the probability of occurrence, operating cost sensitivity, and carbon emission impact of that scenario.

[0037] Scenarios with a probability of occurrence lower than a set probability threshold are removed from the candidate scenarios within the candidate day to obtain the first candidate scenario.

[0038] The operating cost sensitivity and carbon emission impact of the first candidate scenarios are ranked separately, and the scenarios with operating cost sensitivity below the set sensitivity threshold and carbon emission impact above the set impact threshold are selected as key scenarios from the ranking results.

[0039] In one embodiment, calculating the probability of occurrence of the scenario corresponding to the scene includes: Calculate the probability density of the scenario in the joint distribution and normalize it to obtain the probability of the scenario occurring.

[0040] In one embodiment, calculating the operating cost sensitivity corresponding to the scenario includes: Obtain the operating cost corresponding to this scenario.

[0041] The difference between the operating cost and the baseline cost is calculated as the operating cost sensitivity for this scenario.

[0042] In one embodiment, calculating the carbon emission impact corresponding to this scenario includes: Calculate the sum of the carbon emissions corresponding to the electricity purchased in this scenario and the carbon emissions corresponding to the output of the micro gas turbine to obtain the total carbon emissions in this scenario.

[0043] The difference between the total carbon emissions and the baseline carbon emissions is calculated as the degree of carbon emission impact corresponding to this scenario.

[0044] In this embodiment, considering that multiple uncertainties in the carbon-source-load process can lead to a surge in the number of scenarios, directly solving all scenarios during day-ahead optimization would not only be computationally intensive but could also dilute the optimization efforts for key risk scenarios. Therefore, a scenario screening mechanism is introduced during the scenario generation stage. By analyzing the probability of scenario occurrence, sensitivity to operating costs, and the degree of carbon emission impact, a set of high-impact, high-probability key scenarios is selected, while low-impact, low-probability scenarios are eliminated to reduce the computational scale.

[0045] The probability of a scenario occurring is calculated by the probability density of the scenario in the joint distribution within the candidate day and then normalized. The operating cost sensitivity is determined by running the optimization model under each candidate day's scenario to obtain the corresponding operating cost, comparing it with the baseline cost, and taking the relative difference as the cost sensitivity for that scenario. The carbon emission impact is determined by calculating the carbon emissions corresponding to the electricity purchased and the micro-gas turbine output under each candidate day's scenario, comparing it with the baseline emissions, and obtaining the relative increase as the degree of carbon emission impact.

[0046] Then, based on the large sample of candidate scenarios (i.e., intraday candidate scenarios) formed by the characterization of correlation and time series features, the subsequent screening is based on three types of indicators: the probability of scenario occurrence, the sensitivity to operating costs, and the degree of carbon emission impact. Regarding the screening rules, extremely low-probability samples that almost never occur can be eliminated first based on a set probability threshold. In the remaining set (i.e., the first candidate scenarios), the scenarios are then sorted and pruned according to the level of operating cost sensitivity and the degree of carbon emission impact, thus obtaining a representative set of "high probability + high impact." This set is considered the key scenario set, used to control the computational scale and focus on a few scenarios that substantially drive the optimization results.

[0047] This embodiment defines the form of the "scenario space" through a set of correlation and temporal features. A large number of candidate intraday scenarios are then obtained through sampling, followed by physical consistency screening and high-impact screening to select key scenarios for optimization. During scenario construction, priority is given to retaining combined scenarios that are likely to occur simultaneously in actual operation and have a significant impact, such as a situation where low sunlight leads to a decrease in photovoltaic output while the output of high-carbon emission power sources on the grid increases and building loads rise. Spurious extreme combinations that are physically unlikely to occur simultaneously are eliminated. This correlation-driven scenario construction method makes the generated worst-case scenarios closer to actual operational risks, avoiding overly conservative approaches caused by irrelevant extreme combinations.

[0048] Step 104: Perform two-stage robust optimization based on the carbon emission intensity level and key scenarios corresponding to the day-ahead predicted carbon emission intensity of the photovoltaic-storage building cluster to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster.

[0049] In one embodiment, step 104 includes: Based on the day-ahead forecasts of carbon emission intensity, photovoltaic output, and electricity load of the photovoltaic-storage building cluster, a first-stage robust optimization is performed to determine the optimization results of the status of each device in the photovoltaic-storage building cluster.

[0050] The constraints are determined based on the carbon emission intensity level corresponding to the day-ahead forecast of the carbon emission intensity of the photovoltaic-storage building cluster.

[0051] Based on the constraints, the second stage of robust optimization is carried out according to the optimization results of each device status and key scenarios to obtain the day-ahead optimized scheduling results of the photovoltaic-storage building cluster.

[0052] For example, an optimization model is established with the objective of minimizing the operating costs and carbon emission costs of four photovoltaic-storage building clusters. This model not only considers the uncertainty of indirect carbon emission intensity at the grid connection point but also comprehensively considers the randomness and volatility of renewable energy output and load power, thus deriving the optimization strategy for the next 24 hours (energy storage charging / discharging state, electricity purchase / sale state, and micro gas turbine start / stop state). The objective function of the day-ahead optimization model is: ; In the formula: This represents the total cost of the control cycle; 、 , ,and These represent the operating costs of micro gas turbines, the cost of purchasing and selling electricity from the distribution network, the cost of time-shiftable load control, the cost of charging and discharging energy storage, and the cost of carbon emissions, respectively. This indicates the control cycle, with a daily control cycle of 24 hours and a time interval of 1 hour. Uncertain variables representing carbon emission intensity This represents the uncertain variables of photovoltaic power output and electricity load.

[0053] Since the uncertain parameters in the model are only related to equipment output and not to equipment start-up and shutdown parameters, a two-stage robust optimization approach can be used to handle these two types of decision variables separately. The first stage deals with equipment start-up and shutdown variables, disregarding the uncertainty of equipment output and making decisions based on predicted values ​​or deterministic parameters. After determining the equipment's start-up and shutdown states, these decision variables are substituted into the second-stage optimization model. In this stage, a robust optimization method is used to handle uncertainty, constructing a polyhedral uncertainty set to describe the range of variation of uncertain parameters, and considering the worst-case equipment output scheme during the optimization process.

[0054] In the "two-stage robust optimization" process, the previously constructed scenario set is not simply thrown in for complete iteration. Instead, it serves two purposes: first, it provides a realistic and credible boundary for the uncertain set; second, it serves as candidate samples for sub-problems to progressively select key scenarios. Specifically, the basic framework of the model remains "the first stage defines discrete states such as equipment start-up and shutdown, and the second stage adjusts continuous operating variables under uncertain disturbances." The source of these uncertain disturbances is the fluctuation patterns of photovoltaic, load, and carbon intensity represented by these candidate scenarios. In terms of solution logic, unlike traditional methods, the worst-case scenario is not arbitrarily selected from hypothetical independent extreme conditions, but must be generated from key scenarios that conform to physical laws and statistical correlations, as produced in the previous scenario construction steps. This makes the meaning of "worst-case" closer to real operational risks, rather than being overly conservative.

[0055] The standard form of the two-stage robust optimization model is as follows: ; , The optimization variable is represented by the following expression: ; In the formula: The first-stage decision variables represent all 0-1 variables in the equivalent model, i.e., the start-stop state of the micro gas turbine. Distribution network power purchase and sale variables and energy storage charging and discharging states; For the second-stage decision variables, represents all continuous variables in the equivalent model. and Both represent column vectors; and This represents the column vector corresponding to the objective function; These represent the coefficient matrices of the constraints; This represents a constant column vector.

[0056] Combination Figure 2 Within the framework of two-stage robust optimization, the core objective is to minimize the total cost within a control cycle (typically 24 hours). The total cost includes the operating cost of the micro gas turbine, the cost of purchasing and selling electricity from the grid, the cost of adjusting mobile loads, the operating cost of energy storage, and the cost of carbon emissions. The first stage is responsible for determining the discrete states of equipment (such as generator start-up and shutdown, energy storage charging and discharging direction, and grid power purchase and sales direction). The second stage, in the face of uncertainty, determines the specific continuous output and energy flow, ensuring feasibility and optimal economics under all possible disturbances.

[0057] In this embodiment of the invention, the predicted carbon emission intensity (i.e., the day-ahead predicted value of the carbon emission intensity of the photovoltaic-storage building cluster) is divided into three levels—low, medium, and high—under a two-stage robust optimization framework. Its function is to transform the predicted carbon emission intensity into "scenario triggering conditions." Thus, during the two-stage robust optimization process, when the predicted carbon emission intensity falls into a certain level, a preset differentiated operation strategy is automatically activated. For example: during low-carbon level periods, low-carbon electricity is prioritized for charging energy storage devices, and the proportion of mobile loads is moderately increased to improve clean energy utilization; during medium-carbon level periods, energy storage discharge and grid power purchase are combined to balance operational economy and low-carbon characteristics; during high-carbon level periods, grid power purchase is strictly limited, and the proportion of high-carbon electricity use is reduced by releasing energy storage and cutting mobile loads. This tiered strategy can automatically switch according to changes in the carbon emission environment during execution, ensuring optimal operation under different uncertain scenarios.

[0058] The calculation of predicted carbon emission intensity is typically completed before the optimization solution. That is, in the input phase, information such as the power system's generation structure, grid losses, and energy composition is used to predict the carbon emission intensity for the next 24 hours, which is then categorized into three levels based on thresholds. The carbon emission intensity classification and two-stage robust optimization operate within the same two-stage robust framework. The carbon emission intensity level determines the constraints and objective weights in the second stage. The optimization form remains unchanged, but the constraint boundaries and operating strategies switch according to different carbon intensity levels, allowing the model to automatically adopt different operating modes under different carbon emission scenarios.

[0059] In one embodiment, after step 104, the method further includes: The misjudgment rate of constraint default rate and carbon intensity classification is statistically analyzed based on external samples or actual operating data.

[0060] If the default rate exceeds the set upper limit, the uncertainty boundary corresponding to the key scenario is expanded.

[0061] If the default rate is constrained to be lower than the set lower limit for default rate, then the uncertainty boundary corresponding to the key scenario is narrowed.

[0062] If the misjudgment rate of carbon intensity classification is higher than the set upper limit of misjudgment rate, the threshold for carbon emission intensity classification will be relaxed.

[0063] If the misjudgment rate of carbon intensity classification is lower than the set lower limit of misjudgment rate, the threshold for carbon emission intensity classification will be tightened.

[0064] For example, the constraint default rate is determined based on the proportion of violations of operational constraints in external samples or actual operational data; the misjudgment rate of carbon intensity classification is determined based on the proportion of inconsistencies between the carbon intensity level corresponding to the day-ahead forecast value of carbon emission intensity corresponding to external samples or actual operational data and the actual carbon emission intensity level.

[0065] Continue to refer to Figure 2 A round-by-round self-calibration mechanism is introduced in the two-stage robust optimization process to dynamically adjust the uncertainty boundary and carbon emission intensity level classification threshold based on the operational performance under each key scenario. For example, if the operating cost or carbon emission level is found to be too high under a certain scenario, the protection level for that scenario is increased in the next round of optimization; conversely, the constraints are appropriately relaxed to improve economic efficiency. This mechanism enables the model to gradually approach the optimal equilibrium point under actual operating conditions with limited computing resources.

[0066] Specifically, after entering the solution and operation phase, the model drives two-stage robust optimization using these key scenarios, while simultaneously initiating round-by-round self-calibration: On the one hand, it statistically analyzes the constraint default rate and the misclassification rate of carbon intensity classification on external samples or actual operating data. If the constraint default rate or the misclassification rate of carbon intensity classification is higher than the predetermined target (e.g., setting an upper limit for the default rate and an upper limit for the misclassification rate), it indicates that the uncertainty boundary or classification threshold is mismatched with the true distribution, and it is necessary to expand the uncertainty boundary or adjust the classification threshold accordingly to improve coverage and robustness. On the other hand, when the constraint default rate is significantly lower than the target (i.e., setting a lower limit for the default rate) and economic efficiency is unnecessarily sacrificed, it indicates overprotection, and the uncertainty boundary should be tightened and the classification threshold relaxed to recover economic efficiency. At the same time, if it is observed that the operating cost or carbon emissions under a certain type of key scenario are consistently high, the model will be "tightened" in the next round for this type of scenario (e.g., stronger protection and stricter policy constraints), and vice versa. Thus, in the cycle of "screening-verification-calibration", the model gradually approaches the balance point that satisfies both reliability objectives and economic efficiency, with key scenarios as the core carrier.

[0067] The constraint default rate is determined by testing the optimized scheme against external samples or actual operational data to see the proportion of time periods or scenarios where hard constraints such as power balance, charge / discharge limits, and load boundaries are violated. The carbon intensity classification misclassification rate measures the accuracy of the carbon intensity classification triggering mechanism, specifically the percentage of instances where carbon emission intensity is classified as low, medium, or high during the prediction phase, but actual operational data may fall into other ranges. Expanding the uncertainty boundary follows a pattern: when the constraint default rate is high, it indicates that the originally defined uncertainty set is too narrow, resulting in insufficient protection. In this case, the allowable range of prediction error should be widened or the uncertainty adjustment parameter increased to maintain model feasibility within a wider fluctuation range. Conversely, tightening the uncertainty boundary indicates a pattern: when the constraint default rate is very low, even far below the target requirement, it suggests that the set is too wide, leading to an overly conservative model and compromised economic efficiency. In this case, the boundary can be appropriately tightened to reduce protection in extreme scenarios, thereby improving profitability. In this process, the determination of objectives typically involves operators balancing reliability and economic efficiency: setting a target default rate (e.g., no more than 5%) and a lower limit for carbon classification accuracy. These two indicators serve as constraints or adjustment standards, ensuring the solution is robust enough to avoid frequent failures during iterations while avoiding excessive costs for a few rare scenarios. Ultimately, the model gradually converges to a balance point that satisfies both the target reliability and maintains optimal economic efficiency by adjusting the boundaries and thresholds round by round.

[0068] Furthermore, embodiments of the present invention may also include collaborative optimization of multiple regions and multiple uncertainties, namely: For photovoltaic-storage building clusters under different geographical and meteorological conditions, regional uncertainty sets are constructed during the modeling phase to reflect the characteristic differences in photovoltaic output, load variation, and carbon emission intensity in each region. In the recently optimized upper-level decision-making, cross-regional power exchange constraints are introduced. This allows power to be transferred from another region with low carbon intensity and power surplus when a region experiences a combination of low sunlight, high carbon intensity, and high load, reducing the use of high-carbon electricity and ensuring local load supply. This cross-regional collaborative mechanism improves the system's flexibility and resource utilization efficiency in multi-regional and multi-scenario environments, while enhancing overall low-carbon characteristics and operational resilience.

[0069] Cross-regional power exchange constraints refer to the power mutual assistance conditions established for multiple photovoltaic-storage building clusters at the upper-level decision-making level of day-ahead optimization. Due to significant differences in the fluctuation characteristics of photovoltaic output, building load, and carbon emission intensity across regions under varying geographical and meteorological conditions, regional uncertainty sets are constructed during the modeling phase, allowing power exchange between regions during system optimization. Essentially, power exchange constraints are a coupling relationship: they require that the power output from one region must equal the total power absorbed by other regions, while also ensuring that the exchanged power does not exceed the transmission capacity limit and does not disrupt the supply-demand balance within each region. The purpose of these constraints is to connect regions, enabling optimization not only to self-regulate within a single region but also to leverage resource complementarity globally, improving overall low-carbon sustainability and flexibility. Whether cross-regional power exchange is necessary depends on the operational status of each region under key scenarios. When a region experiences a typical unfavorable combination (such as insufficient sunlight leading to a sharp drop in photovoltaic output, while simultaneously experiencing high load and a high level of grid carbon intensity), the local supply-demand balance of that region may depend on high-carbon electricity to maintain. In this case, the optimization model compares the cost of purchasing electricity and carbon emissions within the region with the cost of inter-regional power transfer. If it is more economical and lower-carbon to introduce electricity from low-carbon, photovoltaic-surplus regions, then the inter-regional power exchange constraint will be activated, and the corresponding power exchange variable will take a positive value, indicating that inter-regional support has occurred. Conversely, if local conditions in each region are balanced, there are no obvious unfavorable scenarios, or the cost of inter-regional power transfer is higher than local adjustment, the model will not trigger inter-regional exchange, and the power exchange variable will remain at zero or a low level. In other words, inter-regional exchange is not mandatory, but rather an adaptive decision driven by the supply-demand and carbon emission conditions under different regional uncertainties.

[0070] In this robust optimization process, the first stage determines discrete "state-related" decisions (such as micro-turbine start-up and shutdown, energy storage charging and discharging direction, and electricity purchase and sale direction). Then, the three types of uncertainties related to continuous operational quantities (photovoltaic output, building load, and grid carbon emission intensity) are merged into a "carbon-source-load" uncertainty set, which is then placed into the adversarial subproblem of the second stage to "find the worst-case scenario." In practice, this set is characterized by a polyhedron, originating from the linear boundaries of the source / load uncertainty set and the carbon intensity uncertainty set. After entering the subproblem, it acts as a perturbation of the parameters of the equilibrium constraints, operational boundaries, and cost terms (especially carbon costs). Thus, given the decisions of the first stage, the inner optimization must remain feasible under all possible perturbations and minimize the operating cost in the worst-case scenario. This is the two-stage structure of "first determine the state, then resist uncertainty." To avoid the unrealistic and overly conservative nature of traditional "worst-case" scenarios, this embodiment first uses historical "prediction-actual" pairwise data to construct correlation-driven scenarios (including linkage patterns and physical consistency constraints). It merges the synchronous fluctuations of photovoltaics, load, and carbon intensity into a single model, eliminating spurious extremes that are physically difficult to co-occur. The resulting key scenarios are then used as representative points of the uncertainty set for adversarial sub-problems to search. To complement this, an iterative mechanism of "key scenario screening + round-by-round self-calibration" is proposed: the main problem only carries a small number of high-impact / high-probability scenarios for solution. After the solutions are obtained, adversarial sub-problems then "pick worse" scenarios from the set to supplement them. Based on the constraint default rate and carbon intensity grading misjudgment rate from external sample statistics, the uncertainty boundary and carbon emission intensity grading threshold are dynamically relaxed or tightened, gradually achieving a balance between robustness and economy. The carbon intensity grading strategy runs throughout the second phase: when carbon intensity disturbances fall into different levels, corresponding operational templates are automatically triggered (such as restricting electricity purchases, raising the lower limit of discharge, and expanding the amount of movable load transfer). This is equivalent to "contextualizing" the feasible region and marginal cost of the sub-problem, tightening or loosening them, so that carbon emissions and operating costs under worst-case scenarios are under control. Finally, at the system level, cross-regional power exchange is added to the upper-level decision-making, which is equivalent to providing "external flexibility" for the second phase. When a region falls into the adverse scenario of "low sunlight-high carbon-high load," surplus low-carbon electricity from other regions can offset the impact, further reducing the worst-case cost of the sub-problem and enhancing overall resilience. The above link, from top to bottom, is: joint uncertainty set model → fidelity construction of relevant scenarios → key scenario screening and self-calibration control conservatism → contextualizing constraints and costs of carbon intensity grading in sub-problems → multi-regional collaborative expansion of feasible region and reduction of worst-case cost.

[0071] For example, the embodiments of the present invention are experimentally verified using an IEEE 33-node system. The topology diagram of the IEEE 33-node system is shown below. Figure 3As shown, the system has a total of 33 nodes. Assuming nodes 14, 18, 25, and 32 form a photovoltaic-storage building cluster, the day-ahead forecast of the photovoltaic output of the photovoltaic-storage building cluster and the day-ahead forecast of the total load of the IEEE 33 nodes are as follows: Figure 4 and Figure 5 As shown, the actual value fluctuates around the predicted value, with a maximum prediction deviation of ±15%. The upper and lower limits represent the uncertainty range of the parameters considered in this embodiment. The uncertainty adjustment parameter Γ is 10 for all parameters.

[0072] The day-ahead optimization results obtained after considering the triple uncertainty of carbon-source-charge are as follows: Figure 6 As shown in the figure, Grid represents the power purchased and sold by the grid, BT represents the power of energy storage charging and discharging, PV represents the total photovoltaic output of the photovoltaic-storage cluster, and Load represents the total load power of the IEEE 33 nodes.

[0073] Depend on Figure 6 It can be concluded that during the off-peak electricity price period from 01:00 to 7:00, the day-ahead trading price of the grid is lower than the unit generation cost of the micro gas turbine. During this period, the system mainly relies on purchasing electricity from the grid to meet load demand, with the output of the micro gas turbine accounting for only a small portion. From 08:00 to 12:00, the system transitions from the normal electricity price period to the peak electricity price period. During this period, the first peak demand occurs, and the day-ahead trading price of the grid is higher than the unit generation cost of the micro gas turbine. Furthermore, due to the limitations imposed by the generator ramp-up, when the photovoltaic output cannot meet the system load demand, in order to maximize profits, the system controls the generator to increase its power output to reduce the grid's power purchase capacity. During peak hours, the energy storage begins to discharge. From 13:00 to 18:00, when electricity prices are relatively high, the micro gas turbine operates at maximum output power. From 19:00 to 22:00, the peak electricity price period arrives, marking the second peak in electricity demand. However, there is no output from photovoltaic units. To meet load demand, the micro gas turbine operates at maximum power, and the grid's electricity purchase is also at its peak, causing the energy storage to discharge for the second time. From 23:00 to 24:00, during the off-peak electricity price period, due to ramp-up constraints, the power of the micro gas turbine gradually decreases, and the electricity purchased from the grid also begins to decrease. The excess electricity charges the energy storage device, achieving peak shaving and valley filling.

[0074] Time-shiftable load optimization scenarios, such as Figure 7 and Figure 8 As shown, flexible adjustments to time-shiftable loads can be made to adapt to economic differences and supply-demand balance requirements during different electricity price periods.

[0075] from Figure 7 and Figure 8As can be seen, by adjusting the time-shiftable load, some of the electricity demand that would otherwise occur during peak hours is transferred to periods with lower electricity prices, such as 1-7 hours. This transfer not only reduces the amount of electricity that needs to be purchased during peak hours, but also reduces the power supply pressure on the system during peak hours, thus helping to balance the supply and demand relationship of the power system.

[0076] The energy storage charging and discharging states are shown in Table 1-4. When the charging state variable is 1 and the discharging state variable is 0, it indicates that the system is in a charging state during that time period. When the charging state variable is 0 and the discharging state variable is 1, it indicates that the system is in a discharging state during that time period. If both the charging and discharging state variables are zero, it means that the energy storage device does not perform charging or discharging operations during that time period. For ease of identification, A represents charging, B represents discharging, and C represents no charging or discharging operations. Analysis based on Table 1-4 shows that the energy storage system discharges and charges according to demand at different control times. Although the charging and discharging times of the energy storage devices at the four nodes are not exactly the same, the general trend is that charging occurs at night when electricity prices are lower and load demand is lower, while discharging occurs during the day when electricity prices are higher and load demand is higher.

[0077]

[0078]

[0079]

[0080]

[0081] To verify the robust optimization method and its practical effect after introducing uncertainty in the intensity of indirect carbon emissions from the power industry, this embodiment compares two different day-ahead optimization methods and analyzes the actual total operating cost and carbon emissions of the system under different methods. By comparing the actual total operating cost and carbon emissions of the system under these two methods, the effectiveness and advantages of the robust optimization method in improving system performance, reducing operating costs, and reducing carbon emissions are evaluated. The results obtained by the two optimization methods are shown in Table 5:

[0082] After considering robust optimization operation with triple uncertainties in carbon-source-load, the day-ahead forecasting cost of the method in this embodiment is indeed relatively high. This cost increase is mainly due to the fact that the optimization method considers the worst-case scenario of all uncertain parameters in advance when formulating the plan. Through this optimization strategy, the system can still maintain stable operation when facing various adverse conditions in actual operation. In contrast, traditional deterministic methods have lower forecasting costs because they do not consider these uncertainties, but this does not mean that deterministic methods are superior. In the day-ahead electricity trading market, forecasting errors may lead to an imbalance between the power generation plan and actual demand, which needs to be compensated in the real-time market. However, the electricity price in the real-time market is often higher or lower than that in the day-ahead market, thereby increasing transaction costs. It is against this background that the robust optimization method considering multiple uncertainties in carbon-source-load proposed in this embodiment demonstrates its advantages. This method can formulate more robust control schemes, effectively resist the risks brought by real-time market electricity price fluctuations, and enhance the robustness of the system. The final results for different loads and wind and solar output forecasting errors are shown in Table 6.

[0083]

[0084] Prediction error represents the range of fluctuations in the predicted values ​​of load and photovoltaic output at different time periods compared to the actual values. It can be seen that the robust optimization method considering multiple uncertainties in the carbon-source-load system proposed in this embodiment of the invention handles the imbalance caused by prediction error less than the deterministic optimization method. This is because it considers the uncertainties in the carbon-source-load system, has the ability to resist the risk of fluctuations in uncertain parameters, and only requires minimal cost to compensate for the error between the predicted and actual values ​​of uncertain parameters.

[0085] The proposed day-ahead optimization method for photovoltaic-storage buildings, considering the triple uncertainties of carbon, source, and load, comprehensively utilizes innovative technologies such as multi-uncertainty correlation modeling, carbon intensity grading adaptive strategy, key scenario screening and round-by-round self-calibration solution, and multi-regional collaborative optimization. It demonstrates significant advantages in coping with complex and ever-changing operating environments. First, by introducing a multi-uncertainty scenario construction method based on physical correlation and statistical dependence, it effectively avoids spurious extreme scenarios arising from the independent modeling of photovoltaic output, building load, and carbon emission intensity in traditional methods. This not only improves the realism and effectiveness of scenario predictions but also reduces economic losses caused by overly conservative models. Second, this invention innovatively introduces a carbon intensity grading trigger mechanism, dividing predicted carbon emission intensity into low, medium, and high levels, and formulating differentiated energy storage operation, load adjustment, and power purchase and sale strategies for different levels. This allows the system to automatically switch to the optimal control scheme based on predicted carbon emission intensity during the day-ahead scheduling phase, reducing the use of high-carbon electricity at the source and improving low-carbon operation while maintaining economic efficiency. Furthermore, the critical scenario screening and round-by-round self-calibration solution method of this invention significantly reduces the number of scenarios to be considered and lowers computational complexity without compromising robustness. Simultaneously, by dynamically adjusting the uncertainty boundary and classification threshold through iterative optimization, the final optimization result balances safety and profitability during multiple rounds of convergence, avoiding the excessive computational burden and insufficient response to some critical scenarios inherent in traditional full-scenario traversal methods. Finally, this invention introduces a multi-regional collaborative optimization mechanism, constructing regional uncertainty sets for photovoltaic-storage building clusters under different geographical and meteorological conditions, and achieving power surplus / deficit complementarity through cross-regional power exchange in day-ahead optimization. When a region encounters an unfavorable combination of low sunlight, high carbon intensity, and high load, power can be transferred from regions with low carbon intensity and sufficient photovoltaic output, effectively alleviating regional operational pressure and reducing the use of high-carbon electricity. This mechanism significantly improves the overall low-carbon nature, energy utilization efficiency, and operational resilience of the system. In summary, the method of this invention effectively improves economy, low carbon emissions, and scalability while ensuring system robustness. It can meet the day-ahead optimization operation requirements of large-scale photovoltaic-storage building clusters under multiple uncertainties, and has strong engineering applicability and promotion value.

[0086] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0087] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0088] Figure 9A schematic diagram of the day-ahead optimization device for photovoltaic-storage buildings considering multiple uncertainties provided in an embodiment of the present invention is shown. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown, and are described in detail below: like Figure 9 As shown, the day-ahead optimization device for photovoltaic-storage buildings considering multiple uncertainties includes: Model building module 91 is used to establish uncertain sets of carbon emission intensity, photovoltaic power output, and electricity load for photovoltaic-storage building clusters, and incorporate the uncertain sets of photovoltaic power output, electricity load, and carbon emission intensity into the same joint distribution to construct a carbon-source-load uncertain set.

[0089] The scene generation module 92 is used to perform correlation and time series feature analysis on the carbon-source-load uncertainty set, and to sample and obtain candidate intraday scenes based on the analysis results.

[0090] The scenario screening module 93 is used to screen candidate intraday scenarios based on the probability of scenario occurrence, sensitivity to operating costs, and degree of carbon emission impact to obtain key scenarios.

[0091] The day-ahead optimization module 94 is used to perform two-stage robust optimization based on the carbon emission intensity level and key scenarios corresponding to the day-ahead predicted value of the carbon emission intensity of the photovoltaic-storage building cluster, and to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster.

[0092] In one possible implementation, the scene filtering module 93 is specifically used for: For each scenario in the candidate day, calculate the probability of occurrence, operating cost sensitivity, and carbon emission impact of that scenario.

[0093] Scenarios with a probability of occurrence lower than a set probability threshold are removed from the candidate scenarios within the candidate day to obtain the first candidate scenario.

[0094] The operating cost sensitivity and carbon emission impact of the first candidate scenarios are ranked separately, and the scenarios with operating cost sensitivity below the set sensitivity threshold and carbon emission impact above the set impact threshold are selected as key scenarios from the ranking results.

[0095] In one possible implementation, the scene filtering module 93 is specifically used for: Calculate the probability density of the scenario in the joint distribution and normalize it to obtain the probability of the scenario occurring.

[0096] In one possible implementation, the scene filtering module 93 is specifically used for: Obtain the operating cost corresponding to this scenario.

[0097] The difference between the operating cost and the baseline cost is calculated as the operating cost sensitivity for this scenario.

[0098] In one possible implementation, the scene filtering module 93 is specifically used for: Calculate the sum of the carbon emissions corresponding to the electricity purchased in this scenario and the carbon emissions corresponding to the output of the micro gas turbine to obtain the total carbon emissions in this scenario.

[0099] The difference between the total carbon emissions and the baseline carbon emissions is calculated as the degree of carbon emission impact corresponding to this scenario.

[0100] In one possible implementation, the current optimization module 94 is specifically used for: Based on the day-ahead forecasts of carbon emission intensity, photovoltaic output, and electricity load of the photovoltaic-storage building cluster, a first-stage robust optimization is performed to determine the optimization results of the status of each device in the photovoltaic-storage building cluster.

[0101] The constraints are determined based on the carbon emission intensity level corresponding to the day-ahead forecast of the carbon emission intensity of the photovoltaic-storage building cluster.

[0102] Based on the constraints, the second stage of robust optimization is carried out according to the optimization results of each device status and key scenarios to obtain the day-ahead optimized scheduling results of the photovoltaic-storage building cluster.

[0103] In one possible implementation, the current optimization module 94 can also be used for: The misjudgment rate of constraint default rate and carbon intensity classification is statistically analyzed based on external samples or actual operating data.

[0104] If the default rate exceeds the set upper limit, the uncertainty boundary corresponding to the key scenario is expanded.

[0105] If the default rate is constrained to be lower than the set lower limit for default rate, then the uncertainty boundary corresponding to the key scenario is narrowed.

[0106] If the misjudgment rate of carbon intensity classification is higher than the set upper limit of misjudgment rate, the threshold for carbon emission intensity classification will be relaxed.

[0107] If the misjudgment rate of carbon intensity classification is lower than the set lower limit of misjudgment rate, the threshold for carbon emission intensity classification will be tightened.

[0108] In one possible implementation, The constraint default rate is determined based on the proportion of violations of operational constraints from external samples or actual operational data.

[0109] The misclassification rate of carbon intensity classification is determined based on the proportion of discrepancies between the carbon intensity level corresponding to the day-ahead predicted value of carbon emission intensity from external samples or actual operating data and the actual carbon emission intensity level.

[0110] Figure 10 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 10 As shown, the electronic device 10 of this embodiment includes a processor 100 and a memory 101. The memory 101 stores a computer program 102. When the processor 100 executes the computer program 102, it implements the steps in the various method embodiments described above. Alternatively, when the processor 100 executes the computer program 102, it implements the functions of each module / unit in the various device embodiments described above.

[0111] For example, computer program 102 may be divided into one or more modules / units, which are stored in memory 101 and executed by processor 100 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 102 in electronic device 10.

[0112] Electronic device 10 may include, but is not limited to, processor 100 and memory 101. Those skilled in the art will understand that... Figure 10 This is merely an example of electronic device 10 and does not constitute a limitation on electronic device 10. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 10 may also include input / output devices, network access devices, buses, etc.

[0113] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0114] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0115] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties, characterized in that, include: For photovoltaic-storage building clusters, uncertain sets of carbon emission intensity, photovoltaic output, and electricity load are established, and these sets are incorporated into the same joint distribution to construct a carbon-source-load uncertainty set. Correlation and temporal characteristic analysis were performed on the carbon-source-load uncertainty set, and candidate intraday scenes were obtained by sampling based on the analysis results; The candidate intraday scenarios were screened based on the probability of occurrence, sensitivity to operating costs, and the degree of impact on carbon emissions to obtain key scenarios; Based on the carbon emission intensity level corresponding to the day-ahead forecast of the carbon emission intensity of the photovoltaic-storage building cluster and the key scenarios, a two-stage robust optimization is performed to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster.

2. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 1, characterized in that, The process of filtering candidate intraday scenarios based on scenario occurrence probability, operational cost sensitivity, and carbon emission impact to obtain key scenarios includes: For each of the candidate intraday scenarios, calculate the probability of occurrence, operating cost sensitivity, and carbon emission impact of that scenario. Eliminate scenarios whose occurrence probability is lower than a set probability threshold from the candidate day scenarios to obtain the first candidate scenario; The operating cost sensitivity and carbon emission impact in the first candidate scenarios are sorted, and scenarios with operating cost sensitivity lower than a set sensitivity threshold and carbon emission impact higher than a set impact threshold are selected as key scenarios from the sorting results.

3. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 2, characterized in that, Calculate the probability of the scenario occurring, including: Calculate the probability density of the scenario in the joint distribution and normalize it to obtain the probability of the scenario occurring.

4. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 2, characterized in that, Calculate the operational cost sensitivity for this scenario, including: Obtain the operating cost corresponding to this scenario; The difference between the operating cost and the baseline cost is calculated as the operating cost sensitivity for this scenario.

5. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 2, characterized in that, Calculate the carbon emission impact of this scenario, including: Calculate the sum of the carbon emissions corresponding to the electricity purchased in this scenario and the carbon emissions corresponding to the output of the micro gas turbine to obtain the total carbon emissions in this scenario; The difference between the total carbon emissions and the baseline carbon emissions is calculated as the degree of carbon emission impact corresponding to this scenario.

6. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 1, characterized in that, Based on the day-ahead forecast of carbon emission intensity corresponding to the carbon emission intensity level of the photovoltaic-storage building cluster and the aforementioned key scenarios, a two-stage robust optimization is performed to obtain the day-ahead optimized scheduling results of the photovoltaic-storage building cluster, including: Based on the day-ahead forecasts of carbon emission intensity, photovoltaic output, and electricity load of the photovoltaic-storage building cluster, the first stage of robust optimization is carried out to determine the optimization results of the status of each device in the photovoltaic-storage building cluster. The constraints are determined based on the carbon emission intensity level corresponding to the day-ahead forecast of the carbon emission intensity of the photovoltaic-storage building cluster. Based on the constraints, the second stage of robust optimization is performed according to the optimization results of the states of each device and the key scenarios to obtain the day-ahead optimized scheduling results of the photovoltaic-storage building cluster.

7. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 1, characterized in that, After performing two-stage robust optimization based on the carbon emission intensity level corresponding to the day-ahead forecast of the carbon emission intensity of the photovoltaic-storage building cluster and the aforementioned key scenarios to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster, the process also includes: The misjudgment rate of constraint default rate and carbon intensity classification is statistically analyzed based on external samples or actual operating data. If the default rate of the constraint is higher than the set upper limit of the default rate, then the uncertainty boundary corresponding to the key scenario is expanded; If the default rate of the constraint is lower than the set lower limit of the default rate, then the uncertainty boundary corresponding to the key scenario is reduced; If the misjudgment rate of the carbon intensity classification is higher than the set upper limit of the misjudgment rate, the threshold for carbon emission intensity classification will be relaxed. If the misjudgment rate of the carbon intensity classification is lower than the set lower limit of the misjudgment rate, then the threshold of the carbon emission intensity classification is tightened.

8. The day-ahead optimization method for photovoltaic-storage buildings considering multiple uncertainties according to claim 7, characterized in that: The constraint violation rate is determined based on the proportion of violations of operational constraints by the external samples or the actual operational data. The misjudgment rate of the carbon intensity classification is determined based on the proportion of inconsistencies between the carbon intensity level corresponding to the day-ahead predicted value of the carbon emission intensity of the external sample or the actual operating data and the actual carbon emission intensity level.

9. A day-ahead optimization device for photovoltaic-storage buildings considering multiple uncertainties, characterized in that, include: The model building module is used to establish uncertain sets of carbon emission intensity, photovoltaic output, and electricity load for photovoltaic-storage building clusters, and to incorporate the uncertain sets of photovoltaic output, electricity load, and carbon emission intensity into the same joint distribution to construct a carbon-source-load uncertain set. The scene generation module is used to perform correlation and time series feature analysis on the carbon-source-load uncertainty set, and to sample and obtain candidate intraday scenes based on the analysis results; The scenario screening module is used to screen the candidate intraday scenarios based on the probability of scenario occurrence, sensitivity to operating costs, and degree of carbon emission impact to obtain key scenarios; The day-ahead optimization module is used to perform two-stage robust optimization based on the carbon emission intensity level corresponding to the day-ahead predicted value of the carbon emission intensity of the photovoltaic-storage building cluster and the key scenario, so as to obtain the day-ahead optimized scheduling result of the photovoltaic-storage building cluster.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.

Images