A meteorological adaptive energy dispatching method, device, equipment, and storage medium
By clustering meteorological situation characteristic data and using the R-Vine Copula model, a multi-energy flow dependent structure scenario that conforms to actual meteorological conditions is generated, which solves the problem of the scenario deviating from the actual physical law in the existing technology and realizes precise day-ahead optimization control of gas turbines and energy storage equipment.
Patent Information
- Application Number
- CN202611253564.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies cannot effectively capture the dynamic characteristics of multi-energy flow dependent structures as meteorological conditions change, resulting in deviations between the generated wind and solar power output and multi-load scenarios and actual physical laws, and failing to provide accurate day-ahead optimization control commands for hardware equipment such as gas turbines and energy storage.
By clustering meteorological situation characteristic data, a Markov state transition probability matrix and an R-Vine Copula multi-energy flow dependency model of meteorological situation are constructed. A joint scenario set considering meteorological time series evolution constraints is generated and solved in a day-ahead stochastic optimization scheduling model to obtain the optimal control command under each typical operation scenario.
It ensures that the generated wind and solar power output and diverse load scenarios conform to actual physical laws, provides accurate day-ahead optimization control commands, and ensures optimal control of the system under the power balance of multiple energy buses and equipment operation constraints.
Smart Images

Figure CN123073791A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of energy dispatching technology, and in particular to an energy dispatching method, apparatus, equipment and storage medium based on weather adaptation. Background Technology
[0002] With the continuous increase in the proportion of renewable energy installed capacity, regional integrated energy systems (RIES) with a high proportion of intermittent power sources such as wind power and photovoltaics have become an important direction for energy transformation. Since wind and solar power output and cooling, heating and power loads are random and fluctuating, characterizing their joint probability distribution and generating typical operating scenarios is an important prerequisite for carrying out day-ahead optimization scheduling of the system.
[0003] Currently, for the joint generation of wind and solar power output and multiple load scenarios, existing technologies mainly adopt two types of methods: one is the multivariate dependency modeling method based on static Copula functions, which uses global historical data to fit a fixed Copula topology to describe the correlation between wind and solar power output and electricity, heat and cooling loads; the other is based on pure data-driven deep learning or conventional random sampling methods, which directly learns the data distribution from historical samples and generates scenarios to serve scheduling decisions.
[0004] However, the global static Copula structure cannot capture the dynamic characteristics of multi-energy flow dependent structures as meteorological conditions change. Pure data-driven sampling lacks macroscopic constraints of meteorological evolution laws, which will cause the generated wind and solar power output and multi-load scenarios to deviate from actual physical laws. Using this distorted scenario as input to the day-ahead scheduling model will not be able to generate accurate day-ahead optimized control commands for hardware equipment such as gas turbines and energy storage. Summary of the Invention
[0005] This application provides a weather-adaptive energy dispatching method, device, equipment, and storage medium that can capture the dynamic characteristics of multi-energy flow dependent structures as weather conditions change, and introduces the temporal evolution constraints of the weather situation to make the generated wind and solar power output and multi-load scenarios conform to actual physical laws, thereby providing accurate input for the day-ahead dispatching model, and then generating precise day-ahead optimized control commands for hardware equipment such as gas turbines and energy storage.
[0006] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, this application provides a meteorological adaptive energy dispatching method, characterized in that the method includes: Historical operational data of the region where the integrated energy system is located is obtained. The historical operational data includes meteorological situation characteristic data and various energy measurement data. The meteorological situation characteristic data includes total solar irradiance, wind speed and ambient temperature. The various energy measurement data includes photovoltaic power output, wind power output, and terminal electrical load, heat load and cooling load. The meteorological situation feature data is subjected to meteorological situation clustering to obtain K meteorological situation categories. Each time segment in the historical operation data is labeled with the meteorological situation label corresponding to the meteorological situation category, and a historical meteorological database with meteorological situation labels is constructed. K is a positive integer greater than or equal to 2. Based on the historical meteorological database with meteorological situation labels, the temporal transition frequency between various meteorological situation categories is statistically analyzed, and a Markov state transition probability matrix of meteorological situation is constructed. For each meteorological situation category, the marginal distribution of each energy measurement data is fitted to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, the original measurement data is subjected to probability integral transformation and mapped to a uniform probability space to obtain a standardized probability sequence. For each meteorological situation category, based on the standardized probability sequence, an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category is constructed. The R-Vine Copula multi-energy flow dependency model includes the optimal topological edge set, the optimal Pair-Copula function type, and the optimal parameter set. Based on the Markov state transition probability matrix and the R-Vine Copula multi-energy flow dependency model, a joint scenario set considering meteorological time series evolution constraints and multi-dimensional source-load dependency characteristics is generated. The joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence of each typical operating scenario, where N is a positive integer greater than or equal to 2. Using the N typical operating scenarios and the probability of occurrence of each typical operating scenario as input boundary conditions, a day-ahead stochastic optimization scheduling model is constructed with the objective function of minimizing the comprehensive expected operating cost. Under the conditions of satisfying the multi-energy bus power balance constraints, energy conversion equipment operation constraints, energy storage equipment operation constraints, and flexible load response constraints of the integrated energy system, the objective function is solved to obtain the day-ahead optimal control command for each controllable device in the integrated energy system under each typical operating scenario.
[0007] Optionally, the meteorological situation feature data is clustered to obtain K meteorological situation categories, and a meteorological situation label corresponding to the meteorological situation category is marked for each time segment in the historical operational data to construct a historical meteorological database with meteorological situation labels, including: A set of meteorological feature vectors is constructed based on the total solar irradiance, wind speed, and ambient temperature. The K-means clustering algorithm was used to cluster the meteorological feature vector set, and the optimal number of clusters K was determined by the sum of squared errors and the silhouette coefficient, resulting in K meteorological situation categories. The meteorological situation label corresponding to the meteorological situation category of each time segment in the meteorological feature vector set is synchronously bound to each energy measurement data at the same time to construct a historical meteorological database with meteorological situation labels.
[0008] Optionally, the step of constructing a Markov state transition probability matrix for the meteorological situation based on the historical meteorological database with meteorological situation labels includes: (The text abruptly ends here, so the translation stops as well.) The K typical meteorological situation categories are defined as the discrete state space of a Markov chain. ; Extract the meteorological situation label sequence arranged in time series from the historical meteorological database with meteorological situation labels; The maximum likelihood estimation method is used to statistically analyze the meteorological situation label sequence consisting of states S. i Transition to state S j frequency N ij and state S i Total frequency N i Calculate the single-step state transition probability P ij =N ij / N i ; With the single-step state transition probability P ij Construct a Markov state transition probability matrix of dimension K×K.
[0009] Optionally, for each of the aforementioned meteorological situation categories, marginal distribution fitting is performed on each of the energy measurement data to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, a probability integral transformation is performed on the original measurement data to map it to a uniform probability space, resulting in a standardized probability sequence, including: For each meteorological situation category, historical observations of photovoltaic power output, wind power output, electrical load, heat load, and cooling load are extracted from the historical meteorological database labeled with meteorological situation to construct a five-dimensional source-load sample set specific to that meteorological situation category. For each source load variable under each meteorological situation category, a nonparametric kernel density estimation is used to fit the probability density function of the source load variable, and the optimal smoothing parameter is determined by minimizing the asymptotic integral mean square error. Based on the probability density function obtained from the fitting, the original measurement data of each source load variable are subjected to probability integral transformation using the empirical cumulative distribution function, and mapped to the uniform probability space of [0,1] to obtain the standardized probability sequence.
[0010] Optionally, for each of the aforementioned meteorological situation categories, based on the standardized probability sequence, an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category is constructed. The R-Vine Copula multi-energy flow dependency model includes an optimal topological edge set, an optimal Pair-Copula function type, and an optimal parameter set, including: For each of the aforementioned meteorological situation categories, the standardized probability sequence is used as a five-dimensional training sample set to construct the R-Vine Copula joint probability density function; The maximum spanning tree algorithm is adopted, with the optimization objective of maximizing the sum of the absolute values of Kendall's rank correlation coefficients, to determine the optimal set of acyclic connected edges for the current layer of the tree layer by layer; For each edge in the optimal acyclic connection edge set, traverse the candidate Pair-Copula function family, use maximum likelihood estimation to fit the parameters and minimize the Akaike information criterion as the evaluation index, and select the optimal Pair-Copula function type and its optimal parameter set. Based on the determined results of the current layer tree, conditional cumulative distribution probability data is generated through H function transformation. The topology optimization of the next layer tree is iteratively performed until all layer tree structures are completed. The optimal topology edge set, the optimal Pair-Copula function type, and the optimal parameter set are then output.
[0011] Optionally, based on the Markov state transition probability matrix and the R-Vine Copula multi-energy flow dependency model, a joint scenario set considering meteorological time-series evolution constraints and multi-dimensional source-load dependency characteristics is generated. This joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence for each typical operating scenario, including: The Markov state transition probability matrix is invoked, and the Markov chain Monte Carlo random walk algorithm is used to generate the meteorological situation evolution sequence for the next M consecutive natural days, wherein the meteorological situation of any natural day is taken from the K typical meteorological situation categories, and M is a positive integer greater than or equal to 2. Based on the meteorological situation category of each natural day, match the R-VineCopula multi-energy flow dependency model corresponding to that meteorological situation category; For multiple scheduling sections within each natural day, based on the R-Vine Copula multi-energy flow dependency model, a multidimensional independent white noise sequence is generated using the standard Monte Carlo algorithm, and the independent white noise sequence is mapped to a simulated probability sequence with multi-energy flow dependency relationship through the inverse Rosenblatt transform. The simulated probability sequence is inversely transformed into source-load power scenario data with actual dimensions using an empirical inverse cumulative distribution function, generating multiple joint scenarios; Clustering algorithms are used to reduce the number of joint scenarios. The number of scenarios N to be retained is determined based on the sum of squared errors and the silhouette coefficient. N typical operating scenarios are extracted, and the probability of occurrence of each typical operating scenario is calculated based on the proportion of samples in the cluster to which each typical operating scenario belongs.
[0012] Optionally, the objective function is expressed as:
[0013] in, Let be the probability of the occurrence of the s-th typical operating scenario. Let the external energy purchase cost be the cost of the s-th typical operating scenario. Let be the equipment operation and maintenance cost under the s-th typical operating scenario. The carbon emission trading cost under the s-th typical operating scenario is... This represents the cost of flexible load demand response compensation under the s-th typical operating scenario.
[0014] Secondly, this application provides a weather-adaptive energy dispatching device, comprising: The acquisition module is used to acquire historical operating data of the area where the integrated energy system is located. The historical operating data includes meteorological situation characteristic data and various energy measurement data. The meteorological situation characteristic data includes total solar irradiance, wind speed and ambient temperature. The various energy measurement data includes photovoltaic power output, wind power output, and terminal electrical load, heat load and cooling load. The processing module is used to perform meteorological situation clustering on the meteorological situation feature data to obtain K meteorological situation categories, and to label each time segment in the historical operational data with meteorological situation labels corresponding to the meteorological situation categories, constructing a historical meteorological database with meteorological situation labels, where K is a positive integer greater than or equal to 2; based on the historical meteorological database with meteorological situation labels, to count the time series transition frequency between each meteorological situation category, and to construct a Markov state transition probability matrix for the meteorological situation; for each meteorological situation category, to perform marginal distribution fitting on each energy measurement data to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, to perform probability integral transformation on the original measurement data, mapping it to a uniform probability space to obtain a standardized probability sequence; for each meteorological situation category, based on the standardized probability sequence, to construct an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category, wherein the R-Vine Copula multi-energy flow dependency model includes the optimal topological edge set, the optimal Pair-Copula function type, and the optimal parameter set; based on the Markov state transition probability matrix and the R-Vine... The Copula multi-energy flow dependency model generates a joint scenario set that takes into account meteorological time series evolution constraints and multi-dimensional source-load dependency characteristics. The joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence of each typical operating scenario, where N is a positive integer greater than or equal to 2. The scheduling module is used to construct a day-ahead stochastic optimization scheduling model with the N typical operating scenarios and the probability of occurrence of each typical operating scenario as input boundary conditions, with the objective function of minimizing the comprehensive expected operating cost. Under the conditions of satisfying the multi-energy bus power balance constraints, energy conversion equipment operation constraints, energy storage equipment operation constraints and flexible load response constraints of the integrated energy system, the module solves the objective function to obtain the day-ahead optimal control command for each controllable device in the integrated energy system under each typical operating scenario.
[0015] Thirdly, this application provides a computing device, including a memory and a processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of the first aspects.
[0016] Fourthly, this application provides a computer-readable storage medium for storing a computer program for performing the method as described in any one of the first aspects.
[0017] As can be seen from the above technical solution, this application has at least the following beneficial effects: This application clusters meteorological situation feature data, enabling the topological edge set, Pair-Copula function type, and parameter set of the R-Vine Copula to adaptively change with the meteorological situation. This ensures that a specific R-Vine Copula model exists for any given meteorological situation, and the dependency structure used in scene generation has a one-to-one correspondence with the actual weather conditions, solving the problem of distortion in multidimensional dependency relationships under non-matching weather conditions in static Copulas. This application defines meteorological situation categories as the discrete state space of Markov chains and constructs a Markov state transition probability matrix by statistically analyzing the temporal transition frequencies between different situations. This matrix drives the diurnal evolution of meteorological situations during scene generation, eliminating physically improbable jump paths between meteorological states from the perspective of transition probability. This ensures that the generated scene conforms to the evolution of actual meteorological conditions across daily time spans, solving the problem of scene temporal distortion caused by meteorological jumps. This application also includes scene generation with meteorological situation labels. The process flow and the day-ahead stochastic optimization scheduling model are placed under the same meteorological situation classification system for data transfer. The dimensions and time sections of the generated scenarios are directly mapped to the wind and solar power output boundaries and load boundaries in the scheduling model. This allows the scheduling optimization solution to reflect the physical operating boundaries under various meteorological situations, solving the problem of inaccurate day-ahead scheduling instructions caused by the separation of scenarios and scheduling. This application uses a multi-scenario probabilistic expectation optimization scheduling framework to incorporate the operating costs corresponding to each typical scenario into the objective function according to the probability of occurrence. As a result, under the premise of satisfying the multi-energy bus power balance constraints, equipment operation constraints, and flexible load response constraints, the overall expected cost of scheduling decisions under each scenario is minimized, realizing a systematic response to source-load uncertainty.
[0018] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description
[0019] Figure 1 A flowchart illustrating a weather-adaptive energy dispatching method provided in this application embodiment; Figure 2 A schematic diagram of a weather-adaptive energy dispatching device provided in an embodiment of this application; Figure 3 This is a schematic diagram of a computing device provided in an embodiment of this application. Detailed Implementation
[0020] The terms "first," "second," and "third," etc., used in this application specification and accompanying drawings are used to distinguish different objects, not to limit a specific order.
[0021] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.
[0022] The fundamental reason why existing technical solutions cannot provide accurate inputs for day-ahead scheduling of integrated energy systems while ensuring the quality of scenario generation is that the accuracy of day-ahead optimized scheduling results highly depends on the true reflection of actual physical laws by wind and solar power output and multi-load scenario sets. Specifically, the multi-energy flow dependency structure is significantly affected by changes in meteorological conditions. The dependency relationships between photovoltaic power output, wind power output, and electricity / heat / cooling loads vary considerably under different meteorological conditions. Existing solutions use a global static Copula structure, failing to consider the dynamic changes in the multi-energy flow dependency structure with meteorological conditions, leading to a systematic deviation between the generated joint scenarios and actual physical laws. Simultaneously, existing pure data-driven sampling methods lack macroscopic constraints on the temporal evolution of meteorological conditions, causing the temporal evolution patterns of wind and solar power output and loads in the generated scenarios to deviate from actual meteorological processes. This deviation directly transmits to the day-ahead scheduling optimization results and cannot be eliminated through subsequent scenario reduction or correction. Ultimately, while existing scheduling solutions can complete the solution, their output day-ahead control commands are difficult to accurately match actual operating conditions, failing to provide precise day-ahead optimized control commands for hardware devices such as gas turbines and energy storage.
[0023] In view of this, embodiments of this application provide a weather-adaptive energy scheduling method, which can be executed by a processing device, which can be a terminal or a server.
[0024] The processing equipment extracts meteorological situation features from historical operational data and performs clustering to label each time segment with meteorological situation tags. Based on the labeled database, it statistically analyzes the temporal transition frequency between meteorological situations to construct a Markov state transition probability matrix, thereby transforming the macroscopic inertia of meteorological evolution into microscopic sampling temporal boundary conditions. Simultaneously, it constructs a dedicated R-Vine Copula multi-energy flow dependency model for each meteorological situation category, enabling the topology of the Copula, the type of the Pair-Copula function, and its parameters to adaptively change with the meteorological situation, capturing the dynamic evolution of the multi-energy flow dependency structure under meteorological conditions. On this basis, the Markov state transition probability matrix drives the daily-scale meteorological situation evolution, and the R-Vine Copula model corresponding to the meteorological situation drives the multi-dimensional source-load joint sampling of intraday scheduling sections, generating a joint scenario set of timing and time-series evolution constraints and multi-dimensional dependency features. After clustering and reduction, this set is input into the day-ahead stochastic optimization scheduling model to solve for the controllable equipment controllable commands under each typical scenario. This application introduces the driving relationship between meteorological conditions and multi-energy flow dependent structures into the scenario generation process, and establishes a physical consistency association between scenario generation and scheduling optimization, which solves the problem of inaccurate day-ahead scheduling instructions caused by scenario distortion, and ensures that the generated typical scenarios and scheduling optimizations are consistent at the meteorological condition level.
[0025] To make the technical solution of this application clearer and easier to understand, the following describes a weather-adaptive energy dispatching method provided by an embodiment of this application, in conjunction with the above application scenarios. Figure 2 As shown, this figure is a flowchart of a weather-adaptive energy dispatching method provided in an embodiment of this application. In this embodiment, the method includes: S1. The processing equipment acquires historical operating data of the area where the integrated energy system is located.
[0026] An integrated energy system refers to a system that coordinates and optimizes the supply, conversion, storage, and consumption of multiple energy forms such as electricity, heat, and natural gas. In this application, the integrated energy system includes photovoltaic power generation arrays, wind turbine generators, combined cooling, heating, and power (CCHP) systems, gas-fired boilers, electric chillers, and equipment such as electric energy storage, thermal energy storage, and cold energy storage, supplying energy to end users through a multi-energy busbar for electricity, heat, and cooling.
[0027] The historical operational data includes meteorological characteristics data and various energy measurement data. The meteorological characteristics data includes total solar irradiance, wind speed, and ambient temperature. Total solar irradiance refers to the solar radiation power received per unit area, measured in W / m²; wind speed refers to the airflow speed at the hub at a height of 50 meters, measured in m / s; and ambient temperature refers to the atmospheric temperature, measured in °C. The various energy measurement data include photovoltaic output, wind power output, and terminal electrical load, thermal load, and cooling load. Photovoltaic output refers to the active power output of the photovoltaic array; wind power output refers to the active power output of the wind turbine generator; and terminal electrical load, thermal load, and cooling load refer to the electrical power, thermal power, and cooling power consumed by end users, respectively. All power units are in MW.
[0028] The processing equipment synchronously collects total solar irradiance, wind speed, and ambient temperature data from meteorological monitoring stations deployed within the region. It also collects data on photovoltaic output, wind power output, electrical load, heat load, and cooling load from photovoltaic arrays, wind turbines, and terminal load measurement devices deployed within the integrated energy system. The processing equipment organizes the collected data into a time series format at 15-minute intervals, resulting in a historical operational dataset. This historical operational dataset covers the continuous changes in meteorological conditions and various energy measurement data over time.
[0029] S2. The processing equipment performs meteorological situation clustering on the meteorological situation characteristic data to obtain K meteorological situation categories, and labels the meteorological situation category corresponding to each time segment in the historical operation data to construct a historical meteorological database with meteorological situation labels, where K is a positive integer greater than or equal to 2.
[0030] The processing equipment extracts three-dimensional meteorological time-series data of total solar irradiance, wind speed, and temperature from historical operational data, forming a set of meteorological feature vectors. Each meteorological feature vector in this set corresponds to a time segment, containing the total solar irradiance, wind speed, and temperature values at that moment.
[0031] In some embodiments, the processing device constructs a set of meteorological feature vectors based on the total solar irradiance, wind speed, and ambient temperature; it uses the K-means clustering algorithm to cluster the set of meteorological feature vectors, and uses the sum of squared errors and silhouette coefficients to determine the optimal number of clusters K, thus obtaining K meteorological situation categories; it synchronously binds the meteorological situation label corresponding to each time segment meteorological situation category in the set of meteorological feature vectors to each energy measurement data at the same time, thereby constructing a historical meteorological database with meteorological situation labels.
[0032] Specifically, the processing equipment uses the K-means unsupervised clustering algorithm to classify the meteorological feature vector set into different clusters. K-means clustering is a method that divides data samples into K clusters, assigning each sample to the cluster corresponding to the nearest cluster center. The cluster centers are iteratively updated to optimize the density of samples within each cluster. During the clustering process, the processing equipment uses both the sum of squared errors (SSE) and the silhouette coefficient to determine the optimal number of clusters K. The sum of squared errors (SSE) measures the density of data within each cluster; a smaller SSE value indicates that the data points within the cluster are more concentrated towards the cluster center, resulting in better cluster cohesion. The mathematical formula for SSE is:
[0033] in, The total number of clusters is given. Let represent the sample set of the k-th cluster; Belongs to cluster The meteorological feature vector of the i-th time section; Let be the cluster center of the k-th cluster; It represents the square of the spatial Euclidean distance from the meteorological feature vector of the i-th time section to its cluster center.
[0034] The silhouette coefficient is used to characterize the cohesion of samples within a cluster and the separation of samples between clusters. For any i-th time segment, its silhouette coefficient is... Defined as:
[0035] in, The distance between the meteorological feature vector of the i-th time segment and the vectors of all other time segments in the same cluster is denoted as . Let be the average distance from the meteorological feature vector of the i-th time segment to the vectors of all other time segments in the nearest cluster.
[0036] The processing device sets a traversal interval for the number of candidate clusters. For each given number of candidate clusters K within this traversal interval, K-means clustering is performed independently, and its corresponding SSE value and silhouette coefficient are calculated. The processing device comprehensively compares the SSE inflection points and silhouette coefficient peak values under all candidate cluster numbers K, and, combined with the actual engineering requirements of integrated energy system scheduling, determines the optimal number of clusters. .
[0037] Specifically, the processing device first observes the inflection point where the rate of decrease in the SSE curve slows down (e.g., the rate of change is less than the rate of change threshold) to preliminarily determine the number of candidate clusters. If the granularity of the clusters corresponding to the preliminarily determined optimal SSE inflection point is too coarse and cannot effectively distinguish the multi-energy flow dependence characteristics of the transition season, an adaptive translation correction mechanism is triggered. Based on the candidate clusters, the device searches for the next K value that satisfies the lower limit constraint of engineering physical resolution in the direction of increasing the number of clusters. The processing device simultaneously verifies whether the profile coefficient at this point meets the lower limit threshold condition of separation, that is, the global average profile coefficient must be greater than the preset effective separation threshold (e.g., set to 0.40), and the decrease in profile coefficient from the initial inflection point must be within the set tolerance range (e.g., within 10%). If the above dual threshold verification is passed, it proves that the number of clusters still has intra-cluster separation while improving resolution, and the processing device locks this K value as the final optimal number of clusters. .
[0038] After the above adaptive optimization, the processing equipment divides the meteorological characteristic data for the whole year into... Typical meteorological situations are identified. After situation segmentation, the processing equipment assigns discretized meteorological situation labels to each data segment in the time series. Subsequently, the processing equipment maps and synchronously binds these discrete situation labels on the meteorological time axis to the five-dimensional source-load physical feature vectors of photovoltaic power output, wind power output, electrical load, heat load, and cooling load at the same time, thereby constructing a historical database with meteorological situation labels. Each record in this labeled historical database contains a time segment, a meteorological situation label, and the corresponding five-dimensional source-load data at that time.
[0039] S3. Based on the historical meteorological database with meteorological situation labels, the processing device counts the temporal transition frequency between various meteorological situation categories and constructs a Markov state transition probability matrix for the meteorological situation.
[0040] A Markov chain is a stochastic process in which the state at the next moment depends only on the state at the current moment and is independent of earlier historical states. This application treats the evolution of weather conditions over time as a Markov process.
[0041] In some examples, the processing device will Weather-like situations are defined as the discrete state space of a Markov chain. The processing equipment extracts time-series weather situation label sequences from a historical meteorological database. It employs maximum likelihood estimation, estimating state transition probabilities by statistically analyzing the frequency of state transitions within the historical long-term weather situation label data. Based on the discrete sequences of historical weather situations, the processing equipment statistically analyzes the weather situation from state... Transition to state single-step state transition probability The calculation formula is as follows:
[0042] in, This indicates that in historical time-series data, the meteorological situation... Transformed into meteorological situation Statistical frequency; Representing the meteorological situation in historical data Total frequency of occurrence. Represents the weather situation at time t. Weather conditions up to time t+1 The single-step state transition probability.
[0043] The processing equipment traverses all meteorological situations in the discrete state space, calculates the transition probabilities between each situation, and thus constructs a dimensionless system. Markov state transition probability matrix:
[0044] This matrix satisfies the probability constraint that the sum of its rows is 1, i.e. This state transition probability matrix reflects the transition characteristics of regional climate between different meteorological conditions. Each element in the matrix... This represents the number of days on the current day. Under similar weather conditions, the next day will change to the first The probability of a weather-like situation.
[0045] S4. For each of the meteorological situation categories, the processing device performs marginal distribution fitting on each of the energy measurement data to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, the original measurement data is subjected to probability integral transformation and mapped to a uniform probability space to obtain a standardized probability sequence.
[0046] Marginal probability distribution refers to the probability distribution of a single random variable (without considering other variables). In an integrated energy system, photovoltaic power output, wind power output, electrical load, heat load, and cooling load each have different probability distribution characteristics.
[0047] In some examples, targeting The processing equipment extracts historical observations of photovoltaic power output, wind power output, electrical load, heat load, and cooling load corresponding to various weather conditions from a historical meteorological database tagged with weather conditions, and constructs a system based on these conditions. Each sample set consists of an independent, situation-specific five-dimensional source-load sample set. Each sample set contains historical observation data for each of the five variables under that meteorological situation.
[0048] For each source load variable under each weather condition category, the processing equipment uses nonparametric kernel density estimation to fit the probability density function of that source load variable. Kernel density estimation is a nonparametric probability density estimation method that places a kernel function at each sample data point and then superimposes all kernel functions to form the overall probability density curve. Unlike parametric estimation, kernel density estimation does not presuppose that the data follows a specific parametric distribution; instead, it is driven by the data itself.
[0049] For any source / load variable under a specific meteorological situation Let its sample size be... The processing equipment uses kernel density estimation to fit its probability density function as follows:
[0050] in, The probability density of the specific source / charge continuous physical power independent variable (in MW) is to be determined. The first historical sample set corresponding to the current specific meteorological situation One power observation value; This represents the total number of historical samples, i.e., the sample size mentioned above. This is a smoothing parameter, and its value determines the smoothness of the fitted curve. This is the kernel function.
[0051] The processing device uses a Gaussian kernel function, the formula of which is as follows:
[0052] To avoid the interference of subjective experience on fitting accuracy, the processing device adopts a strategy of minimizing the integral mean square error (MISE) to smooth the parameters. Perform adaptive optimization. The integral mean square error between the kernel estimate and the true value is expressed as:
[0053] in, For mathematical expectation operators; It is the theoretically unknown true probability density function.
[0054] Due to the true distribution Since the optimal smoothing parameters cannot be directly obtained from the above formula, the processing device uses the Silverman rule to perform an asymptotic equivalent substitution for the integral objective function. Given a selected Gaussian kernel function, the optimal smoothing parameters are derived by minimizing the asymptotic mean square error of the integral. The solution formula is as follows:
[0055] in, This represents the standard deviation of the sample data for the source / load variable under the current situation.
[0056] The optimal smoothing parameter is obtained from the above formula. The calculation is transformed into only comparing with the known total number of samples. and sample variance Related direct algebraic operations. Obtained by the processing device. Then, by substituting it back into the Gaussian kernel function formula, the nonparametric marginal probability distribution curve of the source / load variable under the current situation can be obtained.
[0057] After obtaining the marginal probability density curves of each variable using kernel density estimation, in order to satisfy the Copula function's requirement for the domain of the input variables... To meet the requirement of uniform distribution, the processing equipment further utilizes the empirical cumulative distribution function to perform a probability integral transformation on the original measurement data. The probability integral transformation maps a continuous random variable to its cumulative distribution function. Transformation method for a uniformly distributed random variable over an interval. The transformation formula is:
[0058] in, The first historical sample set corresponding to the current specific meteorological situation One power observation value; This is the cumulative distribution function obtained based on kernel density estimation; To map to The uniform distribution probability value of the interval.
[0059] After the above probability integral transformation, the five-dimensional source-load data (photovoltaic output, wind power output, electrical load, heat load, and cooling load) with different dimensions are uniformly mapped to a dimensionless [database / load]. A standardized historical probability sequence is obtained from a uniform probability space. This standardized probability series eliminates the dimensional differences between variables, allowing data from different variables to be analyzed on a uniform probability scale.
[0060] S5. For each meteorological situation category, the processing device constructs an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category based on the standardized probability sequence. The R-Vine Copula multi-energy flow dependency model includes the optimal topological edge set, the optimal Pair-Copula function type, and the optimal parameter set.
[0061] A copula function is a function that connects the joint distribution of multiple random variables to their respective marginal distributions. R-Vine Copula is a high-dimensional extension of copula theory. It characterizes the complex dependencies between high-dimensional random variables by nesting multiple binary copula functions (called pair-copulas) in a specific tree structure. In R-Vine Copula, the "vine" is a tree structure composed of multiple layers. Each layer consists of nodes and edges. Nodes represent random variables or their condition variables, and edges represent pair-copula functions between two nodes. For five-dimensional random variables, the tree structure of R-Vine Copula contains four layers.
[0062] The processing equipment retrieves standardized probability sequences output from various weather conditions after probability integral transformation. Construct a five-dimensional training sample set ,from to This corresponds to the standardized probability space of five source-load characteristics: photovoltaic power output, wind power output, electrical load, heat load, and cooling load. The corresponding R-Vine Copula joint probability density function is... for:
[0063] in, is the subscript for the layer number of the R-Vine Copula tree network. For the five-dimensional source-load feature space, its bottom-level connection tree structure contains a total of 4 layers, i.e. ; For the first The set of all connecting edges in a layered tree structure; For the first Layered tree edge set Any specific connecting edge in; and For connecting edges The two nodes connected at both ends; For connecting edges The corresponding set of condition variables; For connecting nodes and The density function of the binary pair-copula; This is the Copula parameter set corresponding to the binary node; and For a given set of condition variables The conditional cumulative distribution probability value is calculated from the input probability data using the conditional probability distribution formula.
[0064] against For weather-related scenarios, the processing device independently executes weather-driven adaptive topology optimization. For any five-dimensional uniformly distributed sample set corresponding to a weather-related scenario, the processing device uses a Dißmann layer-by-layer construction strategy, starting from the first layer of the tree... Up to the 4th level tree Perform the following steps: S51, Optimal selection of topology connections.
[0065] Regarding the current first The input variables of the layered tree are processed by the processing device to calculate the empirical Kendall's equations between each pair of variables. Absolute value of the rank correlation coefficient. Kendall's A correlation coefficient is a nonparametric statistic that measures the degree of monotonic correlation between two variables; the larger its absolute value, the stronger the correlation. The processing device optimizes by maximizing the sum of the absolute values of the correlation coefficients, and uses the maximum spanning tree algorithm to determine the optimal set of acyclic edges for the current layer. .
[0066] S52, Parameter Fitting and Copula Type Optimization.
[0067] The processing device provides a family of alternative Pair-Copula functions, including Gaussian, t, Gumbel, Clayton, and Frank, in its model library as binary connection node models for constructing the underlying tree-like network.
[0068] The Gumbel function is used to capture "upper-tail heavy" features and is suitable for characterizing concurrent dependencies with extremely high power values; the Clayton function is used to capture "lower-tail heavy" features and is suitable for characterizing concurrent dependencies with extremely low power values; while the Gaussian, t, and Frank functions are used to characterize symmetric correlations under normal conditions.
[0069] For each optimized edge established based on the maximum spanning tree algorithm The processing device iterates through all binary Pair-Copula function families in the candidate library and performs the following calculations: First, fitting calculation. For each candidate binary Pair-Copula function, the processing equipment calls the maximum likelihood estimation method, substituting the actual observed probability sample data under the current meteorological situation into the function to obtain the optimal parameter set corresponding to the function. And calculate the maximum likelihood function value at this time. .
[0070] Second, the evaluation calculation. This involves obtaining the maximum likelihood value. The processing device then substitutes this value into the Akaike Information Criterion (AIC) formula to calculate the evaluation score of the candidate function. The AIC criterion is an evaluation standard for selecting the optimal model from multiple candidate models; a smaller value indicates a better balance between goodness of fit and model complexity. The AIC formula is as follows:
[0071] in, The maximum likelihood function value is obtained by the maximum likelihood estimation method for the processing device for the current candidate binary Pair-Copula function; This represents the number of model parameters.
[0072] Third, the optimal output is selected. The processing device compares the AIC scores of all candidate functions and selects the family of candidate functions that minimizes the AIC value, along with their corresponding maximum likelihood estimation parameters, as the optimal function type for that specific optimization edge. With the optimal parameter set .
[0073] S53, Variable Dimensionality Reduction and Iteration.
[0074] After determining the optimal Pair-Copula function for the current layer, the processing device uses this function to calculate the partial derivatives (i.e., H-function transformation) with respect to the two node variables of the current edge, generating the function for the next layer (the...). The conditional cumulative distribution probability data for the (layer) tree fitting is processed. The processing device repeats the above steps until the entire 4-layer tree structure is constructed.
[0075] Through the above objective optimization, the processing equipment outputs a set of corresponding optimal structural parameters for different meteorological conditions: optimal topological edge set. Optimal function type With the optimal parameter set .
[0076] Optimal topological edge set The most crucial interdependent transmission paths among the five physical quantities—photovoltaic power output, wind power output, electrical load, heat load, and cooling load—under this meteorological condition are depicted using a network connection model. (Optimal function type) The types of tail-related features between dependent nodes were identified. Optimal parameter set. Specific coefficient values are given to quantify the degree of dependence between two physical quantities.
[0077] Outputting this set of parameters is equivalent to generating the R-Vine Copula joint probability density function specific to this weather situation. .
[0078] S6. Based on the Markov state transition probability matrix and the R-Vine Copula multi-energy flow dependency model, the processing device generates a joint scenario set that takes into account meteorological time series evolution constraints and multi-dimensional source-load dependency characteristics, and performs clustering reduction on the joint scenario set to obtain N typical operating scenarios and the occurrence probability of each typical operating scenario, where N is a positive integer greater than or equal to 2.
[0079] An operational scenario refers to the time-series power sequence of a comprehensive energy system within a specific time frame (such as a calendar day), including photovoltaic output, wind power output, electrical load, thermal load, and cooling load. A combined scenario set refers to a collection of multiple operational scenarios.
[0080] Processing device calls Markov state transition probability matrix The Markov chain Monte Carlo random walk algorithm is used to generate future... A sequence of meteorological trends over consecutive natural days Among them, any natural day Corresponding specific meteorological situation From the discrete meteorological situation space The transition value is, i.e. .
[0081] The basic logic of the Markov chain Monte Carlo random walk algorithm is as follows: starting from an initial state, at each step, the next state is determined by randomly sampling based on the transition probability vector corresponding to the current state, and so on, to recursively generate a complete state sequence. In this application, the processing device determines the state sequence by... When determining the daily weather conditions, call the matrix. The middle corresponds to the first The row probability vector of the daily situation is used as the probability distribution for sampling.
[0082] Processing equipment according to the current number Specific weather status labels for each day In the model library, perform targeted matching and load the optimal topology edge set of the R-Vine Copula tree network specific to this situation. Optimal Pair-Copula Function Type and its optimal parameter set , which serves as the multi-energy flow joint dependent topology that drives the generation of joint scenarios throughout the day.
[0083] In automatic matching and loading of the first After implementing a daily-specific R-Vine tree-like dependency topology, the processing equipment independently performs dependency sampling on 96 scheduling sections (in 15-minute increments) every hour of the day. This sampling is applied to any specific time point of the day. The processing equipment uses the standard Monte Carlo algorithm to generate 5-dimensional independent... Uniformly distributed white noise sequence .
[0084] The processing device invokes the inverse Rosenblatt transform algorithm based on this specific weather situation. The corresponding inverse function of the binary Pair-Copula conditional cumulative distribution (i.e. Operator), according to the top-down hierarchical logic of the R-Vine tree network (from the 1st to the 4th level of the tree), converts independent white noise sequences The layer-by-layer conditional mapping is a simulated probability sequence with dependencies at that moment. .
[0085] The Rosenblatt transform is a method for converting independent, uniformly distributed random variables into random variables with a specified dependency structure. The inverse Rosenblatt transform performs the reverse mapping process, and its recursive logic is as follows: Direct mapping of first-dimensional variables: .
[0086] Layer-by-layer conditional mapping of variables from the second to the fifth dimension: .
[0087] in, Indicates the first Time of the first Simulated probability values of dimensional variables; For the conditional cumulative distribution function of the bivariate pair-copula (i.e. The inverse operator of the function. In mathematics, a function is defined as the partial derivative of a bivariate Copula joint probability distribution with respect to its moderating variable.
[0088] Through the above recursive calculations, the independent white noise sequence It is mapped layer by layer into a simulated probability sequence with multi-energy flow dependencies at that moment. .
[0089] The processing equipment needs to The inverse transformation maps back to the power space with actual dimensions. Here, the inverse transformation is a quantile mapping of historical discrete samples.
[0090] For the 96 dispatch sections in a single day, the processing equipment extracts historical real data under specific meteorological conditions and at specific times. , No. Let the total number of power observation samples in the set be denoted as . This The samples are arranged in ascending order from smallest to largest to obtain an ordered sample sequence: The processing equipment utilizes the empirical inverse cumulative distribution function to simulate probabilities. Mapped to real analog power Its mathematical equation is:
[0091] in, This is the empirical inverse cumulative distribution function of the variable under specific meteorological conditions and specific time sections; This represents the final simulated power.
[0092] The actual calculation method for the empirical inverse cumulative distribution function is as follows: ,in , The first in the historical data sample sequence Each observed power value, This is the floor function. When To prevent array out-of-bounds errors, the index is forced to... .
[0093] Through the aforementioned quantile mapping and index addressing, the probability sequence of independent white noise after Copula dependency reconstruction is restored to source-load power scene data with real physical dimensions (MW) and retaining the original nonlinear distribution characteristics.
[0094] The above process is achieved through "Markov transition matrix". The evolution of daily meteorological conditions is driven by a dual-constraint control mechanism: multi-dimensional source-load joint sampling of intraday 15-minute cross-sections using a dedicated R-VineCopula tree topology. The generated scene sequences conform to the transition patterns of natural meteorological conditions across the daily time span, while preserving the nonlinear correlation and asymmetric tail dependence characteristics of five-dimensional heterogeneous energy sources (PV output, wind power output, electrical load, thermal load, and cooling load) at the intraday 15-minute scheduling scale.
[0095] The processing equipment uses the K-means clustering algorithm to reduce the number of generated joint scenes (e.g., 100 groups).
[0096] To determine the optimal number of scenes to retain The processing device reuses the aforementioned sum of squared errors (SSE) "elbow rule" and contour coefficient dual-index evaluation mechanism to optimize the number of scene clusters. In a specific embodiment of this application, after the above dual-index evaluation and optimization, five of the most representative typical daily operation scenes are extracted.
[0097] The processing device counts the number of samples within each cluster of typical scenarios, and calculates the probability of occurrence for each typical scenario by determining the proportion of samples in each cluster to the total number of initial scenarios. This probability is denoted as... .this A typical operating scenario carrying the probability of occurrence will be used as the boundary condition for discretization.
[0098] S7. The processing device uses the N typical operating scenarios and the probability of occurrence of each typical operating scenario as input boundary conditions to construct a day-ahead stochastic optimization scheduling model with the objective function of minimizing the comprehensive expected operating cost. Under the conditions of satisfying the multi-energy bus power balance constraints, energy conversion equipment operation constraints, energy storage equipment operation constraints, and flexible load response constraints of the integrated energy system, the objective function is solved to obtain the day-ahead optimal control command for each controllable device in the integrated energy system under each typical operating scenario.
[0099] The processing equipment integrates the above typical scenarios and their probabilities with the constraints and demand responses of various equipment in the integrated energy system to construct a day-ahead stochastic optimization scheduling model based on multi-scenario expectations.
[0100] The processing equipment determines the set of decision variables.
[0101] The reduction generated by the aforementioned steps Joint stochastic optimization is performed on several typical scenarios, based on time... The device constraint variables driven are uniformly extended to include scene-dimensional subscripts. The combined control commands.
[0102] In a single scheduling cycle 15-minute time slots within To control the step size, for each typical scenario The following four types of joint decision variables are centrally set for the processing equipment: (1) External network interaction decision variables: continuous active power of the system purchasing and selling electricity to the external power grid , And 0-1 Boolean state variables for simultaneous purchase and sale. .
[0103] (2) Decision variables for energy conversion unit output: gas turbine power generation Heat output of gas-fired boilers Electric chiller power consumption Absorption chiller output and waste heat recovery heat power .
[0104] (3) Decision variables for multi-energy storage scheduling: Taking electric energy storage as an example, it includes continuous charging and discharging power. , Actual energy storage status and 0-1 Boolean variables for mutual exclusion control , .
[0105] (4) Flexible demand response decision variables: three types of flexible loads: electricity, heat, and cooling. Actual transfer power Power removed and power reduction .
[0106] Each iteration of the aforementioned decision variables in the multidimensional solution space will directly change the power throughput state of each term in the multi-energy bus power balance equation set.
[0107] The processing equipment constructs multi-energy coupling constraints.
[0108] The processing equipment construction incorporates RIES multi-energy coupling constraints for demand response and energy hubs. This step, based on the physical power of the multi-dimensional dependent scenario of photovoltaic output, wind power output, electrical load, thermal load, and cooling load output from the previous steps, introduces a flexible demand response mechanism for electricity, heat, and cold on the load side, and constructs the mixed-integer linear programming operating boundary for each energy conversion and storage device within the system.
[0109] First, constraints on energy conversion equipment.
[0110] Constraints on energy conversion equipment include wind and solar power output boundaries, combined cooling, heating and power (CCHP) conversion constraints, gas boiler conversion constraints, and electric chiller conversion constraints, which will be introduced separately below.
[0111] Wind and light output limits: The wind turbine generators and photovoltaic arrays configured within the system are no longer subject to repeated calculations of the underlying meteorological-physical mechanisms in the scheduling model. At any given time... active power boundary and The multi-scenario physical power sequence is given by the output of the aforementioned steps.
[0112] Photovoltaic power output boundary Equivalent to assigning values to the first-dimensional power variable ,Right now Wind turbine output boundary Equivalent to assigning values to the second-dimensional power variable ,Right now .
[0113] Constraints of Combined Cooling, Heating and Power (CCHP) Conversion: The combined cooling, heating, and power (CCHP) unit consists of a gas turbine, a waste heat boiler, and an absorption chiller coupled together. The gas turbine consumes natural gas to generate electricity, and the recoverable waste heat generated is absorbed by the waste heat boiler for heating, which is then distributed to the absorption chiller for refrigeration or directly supplied to the heat bus.
[0114] Power generated by gas turbines consuming natural gas With recoverable high-temperature waste heat power The transformation constraints are:
[0115]
[0116] in, For gas turbines Power generation at any given time (kW); for Input the natural gas consumption (m³ / h) of the gas turbine at all times. For the lower heating value of natural gas, take ; , These are the power generation efficiency and non-recoverable heat loss rate of the gas turbine, respectively. for The recoverable waste heat power (kW) generated by the gas turbine at any given time.
[0117] Waste heat boilers recover flue gas and convert it into usable heat power. And absorption chillers utilize the allocated heat power Provide cooling output The transformation constraints are:
[0118]
[0119] in, For waste heat boilers in The power (kW) that is actually converted into usable heat energy at any given moment. The waste heat recovery efficiency of the waste heat boiler; For absorption chillers Cooling power (kW) at any given time; for The heat power (kW) allocated to the absorption chiller at all times; This refers to the coefficient of performance (COP) of the absorption chiller. The heat generated by the waste heat boiler, after deducting the heat required to drive the absorption chiller, is used for the remaining portion... It is directly connected to the heat bus for heating, that is... .
[0120] Conversion constraints for gas-fired boilers: When the waste heat generated by the combined cooling, heating, and power (CCHP) unit cannot meet the system's peak heat load, the gas-fired boiler is deployed independently as an auxiliary heat source, directly burning natural gas for heating. Its gas-heat conversion formula is:
[0121] in, For gas boilers Heat production power (kW) at any given time; This is the input natural gas consumption (m³ / h) of the gas-fired boiler. This refers to the heat conversion efficiency of a gas-fired boiler.
[0122] Electric chiller conversion constraints: The electric chiller consumes the combined electrical energy from the power bus for cooling, working in conjunction with the absorption chiller to share the park's cooling load demand. Its electricity-to-cooling conversion formula is:
[0123] in, For electric refrigeration units Cooling power (kW) at any given time; For electric refrigeration units Power consumed at any given time (kW); This is the coefficient of performance (COP) of the electric refrigeration unit.
[0124] Second, constraints on energy storage equipment.
[0125] The system is configured with electrical energy storage, thermal energy storage, and cold energy storage to mitigate the spatiotemporal mismatch between energy flows and supply and demand. The physical state evolution of these three types of energy storage devices and their charge / discharge mutual exclusion mechanisms exhibit isomorphic characteristics in mixed-integer linear programming. Taking the electrical energy storage device as an example, its constraints are:
[0126]
[0127]
[0128]
[0129]
[0130]
[0131] in, , These respectively represent the energy storage devices in and Actual energy storage capacity at any given time (kWh); The self-discharge rate of the energy storage device; , In order to be in Charging and discharging power (kW) at any given time; , For charging efficiency and discharging efficiency; The scheduling time step; , The lower and upper limits (kWh) of the safe storage capacity allowed for energy storage devices. , Maximum allowable charging and discharging power (kW); , These are 0-1 Boolean variables used to represent charging and discharging states. The model enforces constraints on the start and end times of the scheduling. and All of their energy storage capacity has been restored to maximum capacity. 20%.
[0132] For the thermal and cold energy storage configured within the system, the energy evolution and constraint state equations are consistent with the aforementioned energy storage model. Only the stored energy in the equation needs to be considered. Replace each with the actual heat storage With actual cold storage capacity ; charging and discharging power , Replace each with a charge / discharge heat power , and charging / discharging cooling power , Corresponding subscript Similarly, replace with and .
[0133] Third, flexible load response constraints.
[0134] In regional integrated energy systems, the introduction of demand-side response can guide users to change their energy consumption habits to cooperate with the system's peak shaving and valley filling and the consumption of new energy sources. This application classifies multi-energy loads into three categories: rigid baseline loads, transferable loads, and loads that can be reduced.
[0135] Rigid loads refer to the basic loads that must be met at all times to satisfy the basic production and living needs of users in the park and cannot be subject to regulation.
[0136] After introducing demand response, the system in Actual multi-energy load demand at any given time Can be derived from the original predicted load The sum of load transfer and load reduction results in:
[0137] To ensure the linearity and efficiency of the planning solution, this application further decouples the bidirectional translation load transfer into input and output variables:
[0138] Among them, subscript These represent three different types of energy: electricity, heat, and cold. for The original base load forecast value (kW) is assigned from the power sequence of the 3rd dimension (electrical load), 4th dimension (heat load), and 5th dimension (cooling load) output in the previous steps, i.e. , , ; for Net load transferred in after the transfer operation (kW) at any time. and These represent the actual power transferred in and out (kW), respectively. for The amount of load (kW) that is reduced at any given time.
[0139] To ensure a basic energy experience for users and maintain industrial production standards, the power transfer and reduction within a single scheduling period cannot be infinitely amplified. The constraint model is as follows:
[0140]
[0141] in, The maximum response ratio coefficient that allows users to perform load transfer in a single time period; the second equation represents the response ratio over the entire scheduling cycle. Within the system, the total load transferred out must equal the total load transferred in to ensure energy conservation.
[0142] The constraints that allow for load reduction are:
[0143] in, for The amount of load to be reduced at any time; This is the maximum allowable reduction ratio.
[0144] Fourth, energy balance constraints for multi-energy busbars.
[0145] Based on the modeling concept of energy hubs, the regional integrated energy system is abstracted as a multi-input, multi-output energy coupling node. The processing equipment adopts the power balance equation of multi-energy bus node to describe the energy flow and distribution relationship between "source-grid-load-storage".
[0146] Power balance constraints of the power bus: The terminal electrical load is jointly met by the power purchased from the upstream grid, wind and solar renewable energy generation, gas turbine generation, and energy storage discharge. At the same time, the power consumption allocated to the electric chiller and the charging amount of the energy storage must be deducted.
[0147] Thermal bus power balance constraints: The terminal heat load is met by heat generated by the waste heat boiler (minus the heat driving the absorption chiller), heat generated by the gas boiler, and heat released by the thermal energy storage, while deducting the heat charge of the thermal energy storage:
[0148] Cold bus power balance constraints: The terminal cooling load is supplied by the combined cooling of absorption chillers, electric chillers, and cold energy storage, while deducting the cooling charge of cold energy storage.
[0149] The multi-energy demand variable on the left side of the above equation , , All values are assigned from the actual net load variable after taking into account the aforementioned demand-side response reduction and transfer operations.
[0150] The total input to the natural gas network also follows the principle of node flow balance:
[0151] The processing equipment integrates the four categories of decision variables—external interaction, unit output, energy storage scheduling, and flexible response—into a unified control decision set. The day-ahead optimal scheduling of the integrated energy system seeks an optimal set of equipment action commands in a multi-dimensional solution space, with the optimization objective being to minimize the overall expected operating cost of the system within the day-ahead scheduling period. The objective function is:
[0152] in, Let be the probability of the occurrence of the s-th typical operating scenario. Let the external energy purchase cost be the cost of the s-th typical operating scenario. Let be the equipment operation and maintenance cost under the s-th typical operating scenario. The carbon emission trading cost under the s-th typical operating scenario is... This represents the cost of flexible load demand response compensation under the s-th typical operating scenario.
[0153] The external energy purchase cost includes the net cost of purchasing / selling electricity with the external power grid under the time-of-use pricing mechanism, as well as the cost of purchasing natural gas from the natural gas grid:
[0154] in, This represents the total number of cross sections within the current day's scheduling cycle. and They represent The grid purchase price and sales price of electricity during different time periods; express The unit price of natural gas purchased during the specified time period; and These represent the system in the scenario. Down The actual power purchased from and sold to the external power grid during a given period; Indicates the system in the scenario Down The amount of natural gas purchased from the natural gas network during a given period.
[0155] Equipment operation and maintenance costs include the regular maintenance costs of wind turbines and photovoltaic systems, the additional depreciation costs caused by frequent start-ups and shutdowns of gas turbines and gas boilers, and the energy consumption costs of multi-dimensional energy storage equipment during energy flow.
[0156] in, and The system in the scene The total cost of depreciation during startup and the total cost of depreciation during shutdown caused by frequent operation under the drive. The cost of routine operation and maintenance of each piece of equipment; This refers to the equivalent energy consumption cost of electric / heating / cold energy storage equipment during the charging and discharging process.
[0157] The processing equipment incorporates a carbon trading mechanism based on life cycle assessment. Using life cycle assessment energy chain analysis, the total life cycle carbon emission coefficients for each type of energy covered by the system are calculated. :
[0158] in, , , They represent energy. Carbon emission coefficient components in the manufacturing, storage, transportation, and end-use stages (unit: ).
[0159] Scenario based on carbon trading market mechanism The following is the formula for calculating carbon trading costs:
[0160] in, This is the benchmark price for the carbon trading market during the current scheduling cycle; For the system in the scene The actual total carbon emissions driven by this scenario are determined by the actual output power of each energy conversion device and its corresponding life-cycle carbon emission coefficient. The sum of the products is obtained; This refers to the total amount of free carbon emission allowances issued to the system by regulatory authorities based on the baseline method.
[0161] The system in specific typical scenarios When a flexible load is deployed and the original energy consumption habits are changed, price compensation must be paid to the user side as agreed:
[0162] in, and They represent in The compensation electricity price standard given to the user side when the system calls the unit of transferable load and the unit of load that can be reduced during the time period (unit: yuan / kWh).
[0163] The processing device integrates the low-carbon economy objective function based on multi-scenario probability expectation constructed above with various constraints. This is achieved by introducing a large... The method transforms the nonlinear logic coupling links in the original system into a standard mixed-integer linear programming problem and performs global optimization.
[0164] When interacting with the external power grid, the system at the same time The act of both purchasing and selling electricity must be prohibited. Sufficiently large positive real constants should be introduced into the processing equipment. (e.g., set to 100000) and electricity purchase status variable Construct a reverse reconstruction of pure linear inequalities:
[0165]
[0166] The above conversion ensures that when At that time, the system can only purchase electricity; when At that time, the system could only sell electricity, thus achieving linear decoupling between discrete logic and continuous power.
[0167] Similarly, the mutual exclusion logic of charging and discharging of multi-dimensional energy storage devices has also been strictly linearized through a similar 0-1 Boolean variable boundary contraction concept.
[0168] After the above transformation, the original stochastic optimization model is reconstructed into a standard mixed-integer linear programming model.
[0169] The processing equipment relies on a computer processing platform, uses Python as the underlying algorithm running and data compilation environment, and adopts the PuLP high-level mathematical modeling library as the model building and calling interface to standardize and transform the aforementioned mixed integer linear programming model into a standard coefficient matrix and a multi-dimensional decision variable search space that can be recognized by the solver.
[0170] The model inputs include: time-series-multidimensional dependent scenario physical power data generated in the aforementioned steps, market economic signals of the system (time-of-use electricity price, carbon price, gas price), and rated parameters of each physical device and established operating boundaries in the aforementioned steps.
[0171] The processing equipment uses a CBC (COIN-OR Branch and Cut) mixed-integer programming solver to perform iterative optimization on the mixed-integer linear programming model. Relying on the convergence gap tolerance of the solver, the optimal solution is found in the search space that minimizes the overall expected operating cost of the system in multiple scenarios.
[0172] After the solution is completed, the processing equipment outputs the optimal set of decision variables (i.e., the four types of decision variable sets defined above). In practical engineering applications, these optimal decision variables will be transformed into control commands issued to the underlying devices of the integrated energy system. For example: based on Set the active power output valves of the gas turbine for each time period; based on Trigger the charge / discharge relay of the energy storage system to operate; based on This includes sending flexible load switching control signals to terminal intelligent devices. Ultimately, by executing these control commands, the optimized scheduling and operation of the regional integrated energy system is achieved.
[0173] The above text combined Figures 1 to 2 This application provides a detailed description of a weather-adaptive energy dispatching method. The apparatus and equipment provided in this application will be described below with reference to the accompanying drawings.
[0174] like Figure 3 As shown in the figure, this is a schematic diagram of a weather-adaptive energy dispatching device provided in an embodiment of this application. The device includes: The acquisition module 301 is used to acquire historical operating data of the area where the integrated energy system is located. The historical operating data includes meteorological situation characteristic data and various energy measurement data. The meteorological situation characteristic data includes total solar irradiance, wind speed and ambient temperature. The various energy measurement data includes photovoltaic power output, wind power output, and terminal electrical load, heat load and cooling load.
[0175] Processing module 302 is used to perform meteorological situation clustering on the meteorological situation feature data to obtain K meteorological situation categories, and to label each time segment in the historical operational data with meteorological situation labels corresponding to the meteorological situation categories, constructing a historical meteorological database with meteorological situation labels, where K is a positive integer greater than or equal to 2; based on the historical meteorological database with meteorological situation labels, to count the time series transition frequency between each meteorological situation category, and to construct a Markov state transition probability matrix of the meteorological situation; for each meteorological situation category, to perform marginal distribution fitting on each energy measurement data to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, to perform probability integral transformation on the original measurement data, mapping it to a uniform probability space to obtain a standardized probability sequence; for each meteorological situation category, based on the standardized probability sequence, to construct an R-VineCopula multi-energy flow dependency model corresponding to that meteorological situation category, wherein the R-VineCopula multi-energy flow dependency model is... The Copula multi-energy flow dependency model includes the optimal topological edge set, the optimal Pair-Copula function type, and the optimal parameter set. Based on the Markov state transition probability matrix and the R-VineCopula multi-energy flow dependency model, a joint scenario set considering meteorological time-series evolution constraints and multi-dimensional source-load dependency characteristics is generated. The joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence of each typical operating scenario, where N is a positive integer greater than or equal to 2.
[0176] Optionally, the processing module 302 is specifically used to perform meteorological situation clustering on the meteorological situation feature data to obtain K meteorological situation categories, and to label each time segment in the historical operational data with a meteorological situation label corresponding to the meteorological situation category, thereby constructing a historical meteorological database with meteorological situation labels, including: A set of meteorological feature vectors is constructed based on the total solar irradiance, wind speed, and ambient temperature. The K-means clustering algorithm was used to cluster the meteorological feature vector set, and the optimal number of clusters K was determined by the sum of squared errors and the silhouette coefficient, resulting in K meteorological situation categories. The meteorological situation label corresponding to the meteorological situation category of each time segment in the meteorological feature vector set is synchronously bound to each energy measurement data at the same time to construct a historical meteorological database with meteorological situation labels.
[0177] Optionally, the processing module 302 is further configured to, based on the historical meteorological database with meteorological situation labels, statistically analyze the temporal transition frequency between various meteorological situation categories and construct a Markov state transition probability matrix for the meteorological situation, including: The K typical meteorological situation categories are defined as the discrete state space of a Markov chain. ; Extract the meteorological situation label sequence arranged in time series from the historical meteorological database with meteorological situation labels; The maximum likelihood estimation method is used to statistically analyze the meteorological situation label sequence consisting of states S. i Transition to state S j frequency N ij and state S i Total frequency N i Calculate the single-step state transition probability P ij =N ij / N i ; With the single-step state transition probability P ij Construct a Markov state transition probability matrix of dimension K×K.
[0178] Optionally, the processing module 302 is specifically used to perform marginal distribution fitting on each of the energy measurement data for each of the meteorological situation categories to obtain the marginal probability distribution of each energy variable under that meteorological situation category; and to perform probability integral transformation on the original measurement data based on the marginal probability distribution, mapping it to a uniform probability space to obtain a standardized probability sequence, including: For each meteorological situation category, historical observations of photovoltaic power output, wind power output, electrical load, heat load, and cooling load are extracted from the historical meteorological database labeled with meteorological situation to construct a five-dimensional source-load sample set specific to that meteorological situation category. For each source load variable under each meteorological situation category, a nonparametric kernel density estimation is used to fit the probability density function of the source load variable, and the optimal smoothing parameter is determined by minimizing the asymptotic integral mean square error. Based on the probability density function obtained from the fitting, the original measurement data of each source load variable are subjected to probability integral transformation using the empirical cumulative distribution function, and mapped to the uniform probability space of [0,1] to obtain the standardized probability sequence.
[0179] Optionally, the processing module 302 is specifically used to construct, for each meteorological situation category, an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category based on the standardized probability sequence. The R-Vine Copula multi-energy flow dependency model includes an optimal topological edge set, an optimal Pair-Copula function type, and an optimal parameter set, including: For each of the aforementioned meteorological situation categories, the standardized probability sequence is used as a five-dimensional training sample set to construct the R-Vine Copula joint probability density function; The maximum spanning tree algorithm is adopted, with the optimization objective of maximizing the sum of the absolute values of Kendall's rank correlation coefficients, to determine the optimal set of acyclic connected edges for the current layer of the tree layer by layer; For each edge in the optimal acyclic connection edge set, traverse the candidate Pair-Copula function family, use maximum likelihood estimation to fit the parameters and minimize the Akaike information criterion as the evaluation index, and select the optimal Pair-Copula function type and its optimal parameter set. Based on the determined results of the current layer tree, conditional cumulative distribution probability data is generated through H function transformation. The topology optimization of the next layer tree is iteratively performed until all layer tree structures are completed. The optimal topology edge set, the optimal Pair-Copula function type, and the optimal parameter set are then output.
[0180] Optionally, the processing module 302 is specifically used to generate a joint scenario set considering meteorological time-series evolution constraints and multidimensional source-load dependence features based on the Markov state transition probability matrix and the R-Vine Copula multi-energy flow dependency model, and to perform clustering and reduction on the joint scenario set to obtain N typical operating scenarios and the occurrence probability of each typical operating scenario, including: The Markov state transition probability matrix is invoked, and the Markov chain Monte Carlo random walk algorithm is used to generate the meteorological situation evolution sequence for the next M consecutive natural days, wherein the meteorological situation of any natural day is taken from the K typical meteorological situation categories, and M is a positive integer greater than or equal to 2. Based on the meteorological situation category of each natural day, match the R-VineCopula multi-energy flow dependency model corresponding to that meteorological situation category; For multiple scheduling sections within each natural day, based on the R-Vine Copula multi-energy flow dependency model, a multidimensional independent white noise sequence is generated using the standard Monte Carlo algorithm, and the independent white noise sequence is mapped to a simulated probability sequence with multi-energy flow dependency relationship through the inverse Rosenblatt transform. The simulated probability sequence is inversely transformed into source-load power scenario data with actual dimensions using an empirical inverse cumulative distribution function, generating multiple joint scenarios; Clustering algorithms are used to reduce the number of joint scenarios. The number of scenarios N to be retained is determined based on the sum of squared errors and the silhouette coefficient. N typical operating scenarios are extracted, and the probability of occurrence of each typical operating scenario is calculated based on the proportion of samples in the cluster to which each typical operating scenario belongs.
[0181] The scheduling module 303 is used to construct a day-ahead stochastic optimization scheduling model with the N typical operating scenarios and the occurrence probability of each typical operating scenario as input boundary conditions, with the objective function being the minimization of the comprehensive expected operating cost. Under the conditions of satisfying the multi-energy bus power balance constraints, energy conversion equipment operation constraints, energy storage equipment operation constraints, and flexible load response constraints of the integrated energy system, the objective function is solved to obtain the day-ahead optimal control command for each controllable device in the integrated energy system under each typical operating scenario.
[0182] Optionally, the scheduling module 303 is specifically used to construct an objective function for a day-ahead stochastic optimization scheduling model based on the N typical operating scenarios and the probability of occurrence of each typical operating scenario. The objective function is expressed as minimizing the overall expected operating cost:
[0183] in, Let be the probability of the occurrence of the s-th typical operating scenario. Let the external energy purchase cost be the cost of the s-th typical operating scenario. Let be the equipment operation and maintenance cost under the s-th typical operating scenario. The carbon emission trading cost under the s-th typical operating scenario is... This represents the cost of flexible load demand response compensation under the s-th typical operating scenario.
[0184] The weather-adaptive energy dispatching device according to the embodiments of this application can correspondingly execute the method described in the embodiments of this application, and the other operations and / or functions of each module / unit of the weather-adaptive energy dispatching device are respectively for implementing Figure 2 For the sake of brevity, the corresponding processes of each method in the illustrated embodiments will not be described in detail here.
[0185] This application also provides a computing device.
[0186] like Figure 3 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 400 includes a bus 401, a processor 402, a communication interface 403, and a memory 404. The processor 402, the memory 404, and the communication interface 403 communicate with each other via the bus 401.
[0187] Bus 401 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0188] Processor 402 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).
[0189] Communication interface 403 is used for communication with external devices.
[0190] Memory 404 may include volatile memory, such as random access memory (RAM). Memory 404 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).
[0191] The memory 404 stores executable code, and the processor 402 executes the executable code to perform the aforementioned non-destructive testing method based on weather-adaptive energy dispatch.
[0192] Specifically, in achieving Figure 3 In the case of the illustrated embodiment, and Figure 3 When the modules or units of the weather-adaptive energy dispatching device described in the embodiments are implemented in software, the following steps are performed: Figure 3 The software or program code required for the functions of each module / unit can be partially or entirely stored in memory 404. Processor 402 executes the program code corresponding to each unit stored in memory 404, and executes the aforementioned weather-adaptive energy dispatching method.
[0193] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the aforementioned weather-adaptive energy scheduling method.
[0194] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.
[0195] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.
[0196] When the computer program product is executed by a computer, the computer performs any of the aforementioned weather-adaptive energy dispatching methods. The computer program product can be a software installation package; when any of the aforementioned weather-adaptive energy dispatching methods needs to be used, the computer program product can be downloaded and executed on the computer.
[0197] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.
[0198] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be covered within the scope of protection of this application.
Claims
1. A meteorological adaptive energy dispatching method, characterized in that, The method includes: Historical operational data of the region where the integrated energy system is located is obtained. The historical operational data includes meteorological situation characteristic data and various energy measurement data. The meteorological situation characteristic data includes total solar irradiance, wind speed and ambient temperature. The various energy measurement data includes photovoltaic power output, wind power output, and terminal electrical load, heat load and cooling load. The meteorological situation feature data is subjected to meteorological situation clustering to obtain K meteorological situation categories. Each time segment in the historical operation data is labeled with the meteorological situation label corresponding to the meteorological situation category, and a historical meteorological database with meteorological situation labels is constructed. K is a positive integer greater than or equal to 2. Based on the historical meteorological database with meteorological situation labels, the temporal transition frequency between various meteorological situation categories is statistically analyzed, and a Markov state transition probability matrix of meteorological situation is constructed. For each meteorological situation category, the marginal distribution of each energy measurement data is fitted to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, the original measurement data is subjected to probability integral transformation and mapped to a uniform probability space to obtain a standardized probability sequence. For each meteorological situation category, based on the standardized probability sequence, an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category is constructed. The R-Vine Copula multi-energy flow dependency model includes the optimal topological edge set, the optimal Pair-Copula function type, and the optimal parameter set. Based on the Markov state transition probability matrix and the R-Vine Copula multi-energy flow dependency model, a joint scenario set considering meteorological time series evolution constraints and multi-dimensional source-load dependency characteristics is generated. The joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence of each typical operating scenario, where N is a positive integer greater than or equal to 2. Using the N typical operating scenarios and the probability of occurrence of each typical operating scenario as input boundary conditions, a day-ahead stochastic optimization scheduling model is constructed with the objective function of minimizing the comprehensive expected operating cost. Under the conditions of satisfying the multi-energy bus power balance constraints, energy conversion equipment operation constraints, energy storage equipment operation constraints, and flexible load response constraints of the integrated energy system, the objective function is solved to obtain the day-ahead optimal control command for each controllable device in the integrated energy system under each typical operating scenario.
2. The method according to claim 1, characterized in that, The meteorological situation feature data is clustered to obtain K meteorological situation categories, and a meteorological situation label corresponding to the meteorological situation category is marked for each time segment in the historical operational data to construct a historical meteorological database with meteorological situation labels, including: A set of meteorological feature vectors is constructed based on the total solar irradiance, wind speed, and ambient temperature. The K-means clustering algorithm was used to cluster the meteorological feature vector set, and the optimal number of clusters K was determined by the sum of squared errors and the silhouette coefficient, resulting in K meteorological situation categories. The meteorological situation label corresponding to the meteorological situation category of each time segment in the meteorological feature vector set is synchronously bound to each energy measurement data at the same time to construct a historical meteorological database with meteorological situation labels.
3. The method according to claim 2, characterized in that, Based on the historical meteorological database with meteorological situation labels, the frequency of temporal transitions between various meteorological situation categories is statistically analyzed to construct a Markov state transition probability matrix for the meteorological situation, including: The K typical meteorological situation categories are defined as the discrete state space of a Markov chain. ; Extract the meteorological situation label sequence from the historical meteorological database with meteorological situation labels, arranged in time series; The maximum likelihood estimation method is used to statistically analyze the meteorological situation label sequence consisting of states S. i Transition to state S j frequency N ij and state S i Total frequency N i Calculate the single-step state transition probability P ij =N ij / N i ; With the single-step state transition probability P ij Construct a Markov state transition probability matrix of dimension K×K.
4. The method according to claim 1, characterized in that, For each meteorological situation category, marginal distribution fitting is performed on each energy measurement data to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, a probability integral transformation is performed on the original measurement data to map it to a uniform probability space, resulting in a standardized probability sequence, including: For each meteorological situation category, historical observations of photovoltaic power output, wind power output, electrical load, heat load, and cooling load are extracted from the historical meteorological database labeled with meteorological situation to construct a five-dimensional source-load sample set specific to that meteorological situation category. For each source load variable under each meteorological situation category, a nonparametric kernel density estimation is used to fit the probability density function of the source load variable, and the optimal smoothing parameter is determined by minimizing the asymptotic integral mean square error. Based on the probability density function obtained from the fitting, the original measurement data of each source load variable are subjected to probability integral transformation using the empirical cumulative distribution function, and mapped to the uniform probability space of [0,1] to obtain the standardized probability sequence.
5. The method according to claim 1, characterized in that, For each meteorological situation category, based on the standardized probability sequence, an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category is constructed. The R-Vine Copula multi-energy flow dependency model includes an optimal topological edge set, an optimal Pair-Copula function type, and an optimal parameter set, including: For each of the aforementioned meteorological situation categories, the standardized probability sequence is used as a five-dimensional training sample set to construct the R-Vine Copula joint probability density function; The maximum spanning tree algorithm is adopted, with the optimization objective of maximizing the sum of the absolute values of Kendall's rank correlation coefficients, to determine the optimal set of acyclic connected edges for the current layer of the tree layer by layer; For each edge in the optimal acyclic connection edge set, traverse the candidate Pair-Copula function family, use maximum likelihood estimation to fit the parameters and minimize the Akaike information criterion as the evaluation index, and select the optimal Pair-Copula function type and its optimal parameter set. Based on the determined results of the current layer tree, conditional cumulative distribution probability data is generated through H function transformation. The topology optimization of the next layer tree is iteratively performed until all layer tree structures are completed. The optimal topology edge set, the optimal Pair-Copula function type, and the optimal parameter set are then output.
6. The method according to claim 1, characterized in that, Based on the Markov state transition probability matrix and the R-Vine Copula multi-energy flow dependency model, a joint scenario set considering meteorological time-series evolution constraints and multi-dimensional source-load dependency characteristics is generated. This joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence for each typical operating scenario, including: The Markov state transition probability matrix is invoked, and the Markov chain Monte Carlo random walk algorithm is used to generate the meteorological situation evolution sequence for the next M consecutive natural days, wherein the meteorological situation of any natural day is taken from the K typical meteorological situation categories, and M is a positive integer greater than or equal to 2. Based on the meteorological situation category of each natural day, match the R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category; For multiple scheduling sections within each natural day, based on the R-Vine Copula multi-energy flow dependency model, a multidimensional independent white noise sequence is generated using the standard Monte Carlo algorithm, and the independent white noise sequence is mapped to a simulated probability sequence with multi-energy flow dependency relationship through the inverse Rosenblatt transform. The simulated probability sequence is inversely transformed into source-load power scenario data with actual dimensions using an empirical inverse cumulative distribution function, generating multiple joint scenarios; Clustering algorithms are used to reduce the number of joint scenarios. The number of scenarios N to be retained is determined based on the sum of squared errors and the silhouette coefficient. N typical operating scenarios are extracted, and the probability of occurrence of each typical operating scenario is calculated based on the proportion of samples in the cluster to which each typical operating scenario belongs.
7. The method according to claim 1, characterized in that, The objective function is expressed as: in, Let be the probability of the occurrence of the s-th typical operating scenario. Let be the external energy purchase cost under the s-th typical operating scenario. Let be the equipment operation and maintenance cost under the s-th typical operating scenario. The carbon emission trading cost under the s-th typical operating scenario is... This represents the cost of flexible load demand response compensation under the s-th typical operating scenario.
8. A weather-adaptive energy dispatching device, characterized in that, The device includes: The acquisition module is used to acquire historical operating data of the area where the integrated energy system is located. The historical operating data includes meteorological situation characteristic data and various energy measurement data. The meteorological situation characteristic data includes total solar irradiance, wind speed and ambient temperature. The various energy measurement data includes photovoltaic power output, wind power output, and terminal electrical load, heat load and cooling load. The processing module is used to perform meteorological situation clustering on the meteorological situation feature data to obtain K meteorological situation categories, and to label each time segment in the historical operational data with meteorological situation labels corresponding to the meteorological situation categories, constructing a historical meteorological database with meteorological situation labels, where K is a positive integer greater than or equal to 2; based on the historical meteorological database with meteorological situation labels, to count the time series transition frequency between each meteorological situation category, and to construct a Markov state transition probability matrix for the meteorological situation; for each meteorological situation category, to perform marginal distribution fitting on each energy measurement data to obtain the marginal probability distribution of each energy variable under that meteorological situation category; based on the marginal probability distribution, to perform probability integral transformation on the original measurement data, mapping it to a uniform probability space to obtain a standardized probability sequence; for each meteorological situation category, based on the standardized probability sequence, to construct an R-Vine Copula multi-energy flow dependency model corresponding to that meteorological situation category, wherein the R-Vine Copula multi-energy flow dependency model includes the optimal topological edge set, the optimal Pair-Copula function type, and the optimal parameter set; based on the Markov state transition probability matrix and the R-Vine... The Copula multi-energy flow dependency model generates a joint scenario set that takes into account meteorological time series evolution constraints and multi-dimensional source-load dependency characteristics. The joint scenario set is then clustered and reduced to obtain N typical operating scenarios and the probability of occurrence of each typical operating scenario, where N is a positive integer greater than or equal to 2. The scheduling module is used to construct a day-ahead stochastic optimization scheduling model with the N typical operating scenarios and the probability of occurrence of each typical operating scenario as input boundary conditions, with the objective function of minimizing the comprehensive expected operating cost. Under the conditions of satisfying the multi-energy bus power balance constraints, energy conversion equipment operation constraints, energy storage equipment operation constraints and flexible load response constraints of the integrated energy system, the module solves the objective function to obtain the day-ahead optimal control command for each controllable device in the integrated energy system under each typical operating scenario.
9. A computing device, characterized in that, Including memory and processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method as described in any one of claims 1 to 7.
