A method and system for determining the time sequence feasible region of net purchase power of a cold-heat-electricity micro-energy network

Abstract

A method and system for determining the time-series feasible domain of net purchased electricity in a CHP microgrid. The method involves acquiring equipment configuration parameters and historical operating data of the CHP microgrid system to construct a physical model of the system; constructing a set of joint uncertainty scenarios involving spot electricity prices, output from various renewable energy sources, and CHP loads; extracting empirical boundary features of net purchased electricity from historical operating data using a data-driven algorithm to construct a prior knowledge base of the feasible domain; establishing an optimization model of the time-series feasible domain considering equipment operating constraints, multi-energy coupling balance constraints, and uncertainty propagation; solving for the feasible intervals of net purchased electricity in each time period using a combined numerical and analog algorithm; and outputting and verifying the characterization results of the time-series feasible domain of net purchased electricity. This invention characterizes the time-series feasible domain of net purchased electricity for CHP microgrids participating in grid interaction under the dual uncertainty of spot electricity price fluctuations and source-load conditions, providing decision support for electricity spot market application and dispatch.

Description

Technical Field

[0001] This invention belongs to the field of power system dispatching and energy management, and specifically relates to a method and system for determining the time-series feasible domain of net power purchase for a microgrid with cooling, heating and electricity. Background Technology

[0002] The spot electricity pricing mechanism has been gradually implemented in several provincial power grids. As an important market participant, combined cooling, heating, and power (CCHP) microgrids face the critical technical challenge of scientifically assessing their claimable net power purchase range in the spot market. CCHP microgrids integrate various devices such as gas-fired power generation, waste heat utilization, and energy storage. Their net power purchase is influenced by complex multi-energy coupling constraints and uncertainties. Traditional methods have the following limitations in this scenario: First, spot electricity prices are highly random and time-dependent, with daily price fluctuations reaching several times or even tens of times. This makes the optimal response strategies of microgrids differ significantly under different price scenarios. Furthermore, the feasible region obtained under the fixed price assumption cannot reflect the impact of price uncertainty on the net power purchase boundary.

[0003] Secondly, the CHP microgrid simultaneously accepts distributed renewable energy sources such as photovoltaics and wind power. The strong fluctuations in power output on the source side and the time-varying characteristics of cooling, heating and electricity demand on the load side are superimposed to form a double uncertainty, which causes the boundary of the feasible domain of the system to change dynamically with the operating conditions, making it difficult to describe accurately with static intervals.

[0004] Furthermore, there are multiple energy flow coupling relationships within the microgrid: electrical power indirectly affects the supply of heat and cold through waste heat boilers and lithium bromide units, and the introduction of inter-time state coupling by thermal storage devices and energy storage devices makes the temporal feasible domain of net purchased power have high-dimensional constrained coupling characteristics. Traditional enumeration or linearization methods are difficult to meet the requirements of online characterization in terms of both accuracy and efficiency.

[0005] Finally, while existing data-driven methods have advantages in handling high-dimensional uncertainty, they lack the embedding of physical constraints and are prone to generating pseudo-feasible solutions that violate energy balance or equipment operating limits; while pure model-driven methods face the serious curse of dimensionality due to the large number of scenarios and multi-time period coupling.

[0006] Therefore, how to quickly and accurately characterize the time-series feasible domain of net purchased power of microgrids for cooling, heating, and power generation under the dual uncertainties of spot electricity price fluctuations and source and load is an important technical problem that urgently needs to be solved. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method and system for determining the time-series feasible domain of net power purchase by a microgrid (cooling, heating, and electricity), thereby solving the technical problem of rapidly and accurately characterizing the time-series feasible domain of net power purchase by a microgrid participating in the electricity spot market under complex and uncertain environments.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution.

[0009] This invention first discloses a method for determining the time-series feasible region of net purchased power in a microgrid (cooling, heating, and electricity), the method comprising the following steps: The system acquires equipment configuration parameters and historical operating data from the energy management system and equipment data acquisition system of the microgrid (cooling, heating, and power), and constructs a system physical model containing topology information for the conversion of electrical, thermal, and cooling energy based on these parameters and data. The equipment configuration parameters include the rated capacity, conversion efficiency, and operating limits of gas turbines, waste heat boilers, lithium bromide absorption chillers, electric chillers, thermal storage devices, and energy storage devices. The historical operating data includes historical time-series records of net purchased power, equipment output, cooling, heating, and power loads, and renewable energy output. Based on the historical operating data, a set of joint uncertainty scenarios covering spot electricity prices, output of various renewable energy sources, and cooling, heating, and power loads is constructed to characterize the time-series correlation structure of multi-source uncertainty. A nonparametric kernel density estimation algorithm based on Gaussian kernel function is used to perform probability density modeling on the net power purchase of the historical operating data. Combined with data-driven methods of quantile extraction and time-series clustering, empirical feasible boundary features of net power purchase are extracted to form a prior knowledge base. Based on the physical model of the system, equipment operation constraints and multi-energy coupling balance constraints are established, and a feasible region optimization characterization model that satisfies the equipment operation constraints and multi-energy coupling balance constraints is constructed. The feasible range of net purchased power is characterized as a set of constraints coupled between time periods. The algorithm of digital-analog joint drive is adopted to retrieve the empirical boundary of net purchased power from the prior knowledge base. Based on the constraint propagation reasoning mechanism and lightweight neural network, the temporal feasible domain of net purchased power in each time period is solved, and the multi-energy coupling balance constraint and equipment operation constraint are substituted for verification. Output the time-series feasible domain representation results of the net purchased power, and perform feasibility verification under the joint uncertainty scenario set.

[0010] The present invention further includes the following preferred embodiments: The step of constructing a joint uncertainty scenario set covering spot electricity prices, output of various renewable energy sources, and cooling, heating, and power loads based on the historical operating data further includes: Statistical modeling was performed on historical spot electricity price time series data to train a conditional probability model, identify the intraday peak-valley distribution characteristics of electricity prices, the occurrence pattern of peak periods and the probability of cross-period shifts, and generate multiple electricity price scenario curves covering typical high, medium and low price patterns using a conditional probability sequence sampling method. A Beta distribution model was established for the historical prediction error of photovoltaic power output according to weather classification; a Weibull distribution model was established for the prediction error of wind power output; the rank correlation coefficient between the power output of each energy source within the same period was fitted using historical data from the same period, and the Copula function was used to synthesize joint error scenarios to generate a set of power output error scenarios that reflect the spatiotemporal correlation of multiple energy sources. By combining electricity price scenarios and power output error scenarios, a fast forward reduction algorithm is used to compress the joint scenario set to a preset size, while retaining typical scenarios that represent the characteristics of uncertainty distribution.

[0011] The method employs a nonparametric kernel density estimation algorithm based on Gaussian kernel function to perform probability density modeling on the net power purchase data of the historical operating data. Combined with data-driven methods of quantile extraction and time-series clustering, empirically feasible boundary features of the net power purchase are extracted to form a prior knowledge base. This further includes: Historical operational data are categorized by season, date type, and weather conditions to construct a categorized historical sample set; Kernel density estimation is performed on the historical distribution of net power purchased in each time period under each scenario, the marginal distribution is fitted, the empirical density function of net power purchased in each time period is obtained, and then the quantile boundary and the historical maximum and minimum values ​​are calculated to form a hierarchical empirical boundary description vector to identify the dense and sparse regions of the historical feasible interval. Extract the daily curve shape feature vector of the average net power purchase for each time period to form the time series feature of the empirical feasible boundary of net power purchase.

[0012] The construction of the feasible region optimization characterization model that satisfies the device operation constraints and multi-energy coupling balance constraints further includes: Establish multi-energy coupling balance constraints for cold, heat and electricity microgrids, including power balance constraints, thermal balance constraints and cold balance constraints; Establish equipment operation constraints, including upper and lower limits of gas turbine output, ramp rate constraints, recursive relationship constraints of the state of charge of energy storage devices between adjacent time periods, upper and lower limits of state of charge constraints, and logical mutual exclusion constraints that charging and discharging cannot be carried out simultaneously; similar state recursion and power limit constraints of thermal storage devices; minimum operating time constraints for the switching between hot and cold modes of lithium bromide units; and cumulative operating time constraints for each piece of equipment within the entire scheduling cycle. Construct upper and lower bound subproblems for solving the feasible region of net purchased power time series. Obtain the boundary of the feasible region by maximizing and minimizing the net purchased power in each time period, respectively. That is, the feasible region of net purchased power time series is defined as the set of net purchased power time series that the system can satisfy the above-mentioned full-time multi-energy coupling balance constraint and time-by-time equipment operation constraint of the cold, heat and power microgrid by reasonably allocating the output of each device under all typical uncertainty scenarios.

[0013] The method employs a combined numerical and analog-driven algorithm to retrieve the empirical boundary of net purchased power from the prior knowledge base. Based on a constraint propagation reasoning mechanism and a lightweight neural network, it solves the temporal feasible domain of net purchased power for each time period and substitutes it into the multi-energy coupling balance constraints and equipment operation constraints for verification. This further includes: The empirical boundary of the most similar historical category is retrieved from the prior knowledge base and used as the initial value for the hot start of net power purchase in each time period; A constraint propagation reasoning mechanism is adopted, and the energy storage state coupling relationship between adjacent time periods is used to tighten the feasible range of net power purchase for each time period through forward and backward propagation. Taking the daily feature vector as input, which includes load forecast curve features, various renewable energy output forecast features, electricity price scenario features and prior boundary features, a lightweight fully connected neural network pre-trained on historical data is invoked to output the prediction results of the upper and lower bounds of net power purchase in the entire time domain. The net power purchase boundary values ​​predicted by the neural network for each time period are substituted into the multi-energy coupling balance constraints and equipment operation constraints for verification. For time periods where constraints are found to be violated, local precise optimization is initiated to correct the violated boundary values ​​with the minimum adjustment range.

[0014] After outputting the characterization results of the time-series feasible domain of the net purchased power, the method further includes: The upper and lower bound sequences of net power purchases for each time period are used as boundary representations of the time-series feasible domain. The cross-time period coupling constraint description between key time periods is output. Combined with the power spot market declaration rules, the recommended declaration range of net power purchases is generated. Based on the aforementioned temporal feasible domain, feasibility verification is carried out under multiple uncertainty scenarios, and the probability of the system operating within the feasible domain under each scenario is statistically analyzed. When spot electricity prices or source-load forecast information are updated, a rolling correction mechanism for the feasible region is triggered, enabling online dynamic characterization of the time-series feasible region.

[0015] This invention also discloses a system for determining the time-series feasible region of net purchased power for a cooling, heating, and power (CHP) microgrid using the aforementioned method, further comprising: It includes a data modeling module, an uncertainty characterization module, a data-driven boundary extraction module, a feasible region optimization module, a mathematical-model co-solution module, and a result output verification module; The data modeling module is used to acquire the equipment configuration parameters and historical operating data of the energy management system and equipment data acquisition system of the microgrid (CHP), and to construct a system physical model containing topology information for the conversion of electrical, thermal, and cooling energy based on the equipment configuration parameters and historical operating data. The equipment configuration parameters include the rated capacity, conversion efficiency, and operating limits of gas turbines, waste heat boilers, lithium bromide absorption chillers, electric chillers, thermal storage devices, and energy storage devices. The historical operating data includes historical time-series records of net purchased power, equipment output, CHP load, and renewable energy output. The uncertainty characterization module is used to construct a set of joint uncertainty scenarios covering spot electricity prices, output of multiple renewable energy sources, and cooling, heating, and power loads based on the historical operating data, which is used to characterize the time-series correlation structure of multi-source uncertainty. The data-driven boundary extraction module is used to perform probability density modeling of the net power purchased from the historical operating data using a nonparametric kernel density estimation algorithm based on Gaussian kernel function, and to extract empirical feasible boundary features of net power purchased by combining quantile extraction and time-series clustering data-driven methods to form a prior knowledge base. The feasible region optimization module is used to establish equipment operation constraints and multi-energy coupling balance constraints based on the system physical model, construct a feasible region optimization characterization model that satisfies the equipment operation constraints and multi-energy coupling balance constraints, and characterize the feasible range of net purchased power as a set of constraints coupled between time periods. The digital-analog collaborative solution module is used to retrieve the empirical boundary of net purchased power from the prior knowledge base using a digital-analog collaborative driving algorithm, solve the temporal feasible domain of net purchased power for each time period based on the constraint propagation reasoning mechanism and lightweight neural network, and verify it by substituting it into the multi-energy coupling balance constraint and equipment operation constraint. The result output verification module is used to output the characterization results of the time-series feasible domain of the net purchased power and to perform feasibility verification under the joint uncertainty scenario set.

[0016] Accordingly, this application also discloses a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the aforementioned method for determining the timing feasible domain of net purchased power in a microgrid.

[0017] Accordingly, this application also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned method for determining the timing feasible domain of net purchased power in a cold, hot, and electric microgrid.

[0018] The beneficial effects of this invention are that, compared with the prior art, this invention provides a method and system for determining the time-series feasible domain of net purchased power in a cooling, heating, and electric microgrid, which has the following advantages: 1. Comprehensive Uncertainty Characterization: By using a joint scenario set method to simultaneously consider three types of uncertainties and their temporal correlations—spot electricity prices, output from multiple renewable energy sources, and cooling, heating, and power loads—the robustness of the feasible region characterization results is significantly enhanced, enabling the safe operation of microgrids in complex market environments.

[0019] 2. Efficient Solution Through Collaborative Mathematical Modeling: This approach integrates data-driven prior boundary features with precise constraints from a physical model. Prior warm-start and constraint propagation significantly compress the search space, while lightweight neural networks accelerate inference. This reduces the online solution time for full-time feasible domain characterization to minutes. Simultaneously, a physical correction mechanism ensures physical consistency of the results, overcoming the limitations of purely data-driven and purely model-driven methods. Tests show that the solution efficiency is improved by over 80% compared to purely physical model methods, while avoiding the risk of pseudo-feasible solutions generated by purely data-driven methods.

[0020] 3. Complete characterization of temporal coupling: Through the cross-time period constraint propagation of the charge state of the energy storage device, the coupling relationship between each time period in the temporal feasible domain of net purchased power is fully preserved. The output results not only include the boundary intervals of each time period, but also the joint feasible set information between time periods, that is, the constraint relationship of feasible combinations. It can be directly used to formulate multi-time period collaborative scheduling strategies, providing a complete constraint description basis for formulating multi-time period collaborative optimization scheduling strategies.

[0021] 4. Online rolling adaptive: The feasible region rolling correction mechanism enables the characterization results to be dynamically adjusted as spot prices are updated in real time and source-load forecast deviations are corrected, meeting the online characterization requirements for multiple rolling optimizations within the day. The inference latency reaches the minute level, which meets the real-time requirements of power spot market dispatch and improves the adaptability of microgrid dispatch decisions in complex market environments.

[0022] 5. Direct support for market decision-making: The output time-series feasible domain can be directly connected to the application process of the electricity spot market, providing microgrid operators with a quantified net power purchase application range, effectively reducing the risk of market clearing deviation caused by improper application strategies. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for determining the time-series feasible domain of net purchased power in a microgrid for cooling, heating, and electricity, provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the time-series feasible domain characterization framework of multi-energy coupled topology and digital-analog collaborative driving of microgrids provided in the embodiments of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0026] The embodiments described in this application are merely some, not all, embodiments of the present invention. Based on the spirit of the present invention, other embodiments obtained by those skilled in the art without inventive effort are all within the protection scope of the present invention.

[0027] See Figure 1 As shown, the method for determining the time-series feasible domain of net purchased power in a microgrid for cooling, heating, and electricity disclosed in this invention includes the following steps: Step S101: Obtain the equipment configuration parameters and historical operating data of the energy management system and equipment data acquisition system of the microgrid for cooling, heating and electricity, and construct a system physical model containing topology information of electric, heat and cold energy conversion based on the equipment configuration parameters and historical operating data.

[0028] Specifically, the configuration parameters of various equipment obtained from the microgrid energy management system (EMS) and equipment data acquisition system include: the rated power generation, power generation efficiency, and waste heat output coefficient of the gas turbine; the thermal efficiency and maximum heat output of the waste heat boiler; the coefficient of performance (COP) and coefficient of heating of the lithium bromide absorption chiller, as well as its rated cooling / heating power limits; the cooling COP and power limits of the electric chiller; and the rated capacity, maximum charge / discharge power, charge / discharge efficiency, and initial state of charge of the thermal storage device and the electric storage device.

[0029] Specifically, time-by-time operation data within the historical scheduling cycle is collected, including net purchased power time-series records, output sequences of each device, measured values ​​of cooling, heating and power loads, measured values ​​of photovoltaic and wind power outputs, and spot electricity price records for the corresponding time periods, forming training and validation datasets.

[0030] Specifically, the equipment configuration parameters and historical operating data are used to construct a system physical model that includes topological information on the conversion of electrical, thermal, and cold energy. This model is described in detail below.

[0031] (1) Model function: to characterize the conversion path and coupling constraint relationship between the three types of energy flow (electric, heat and cold) in the microgrid, and to give the energy balance equation of each energy node and the energy conversion equation at the device level, which serve as the physical basis for constructing the feasible region optimization model in step S104 and the physical basis for constraint propagation reasoning in step S105.

[0032] (2) Model input: Equipment configuration parameter set Θ={gas turbine electrical efficiency, gas turbine heat output coefficient, waste heat boiler, bromine cooling, bromine heating, electric cooling, electric storage charging, electric storage discharging, heat storage charging, heat storage discharging, heat storage heat dissipation}; rated capacity and power limit of each equipment; output variables of each equipment during time period; external input energy (natural gas flow, photovoltaic output, wind power output).

[0033] (3) Model output: Node-device association matrix, representing the connection relationship between 3 energy nodes (electric node, heat node, cold node) and 8 energy conversion devices (gas turbine, waste heat boiler, lithium bromide unit, electric chiller, energy storage device, thermal storage device, photovoltaic, wind power); multi-energy coupling coefficient matrix, quantifying the conversion contribution coefficient of each device to the 3 types of energy nodes; device-level energy conversion equation set; net purchased power, the exchange interface with the power grid.

[0034] (4) Model structure: The model consists of a three-layer structure: node layer, device layer, and edge layer. The node layer contains three energy nodes: electrical, thermal, and cold; the device layer contains the above eight energy conversion / storage devices; and the edge layer consists of directed weighted edges with weights of conversion efficiency or COP coefficient.

[0035] (5) Algorithm employed: The topology construction adopts the incidence matrix method in graph theory, and the incidence matrix A and coupling matrix C are automatically generated from the device configuration parameters; the conversion efficiency and COP parameter identification adopt the least squares method, and the objective function for parameter identification is:

[0036] T The total number of historical time periods used to identify the parameters; t For time period numbering, t ∈{1, 2, …, T}; Θ is the vector of equipment parameters to be identified, including gas turbine power generation efficiency, gas turbine waste heat production coefficient, waste heat boiler thermal efficiency, lithium bromide absorption chiller cooling COP, lithium bromide absorption chiller heating COP, electric chiller COP, energy storage device charging efficiency, energy storage device discharging efficiency, thermal storage device charging efficiency, thermal storage device releasing efficiency, and thermal storage device heat dissipation coefficient. For the first t The time-period equipment-level input vector includes natural gas flow, net power purchase from the grid, and measured values ​​of photovoltaic and wind power output; The actual output measurement vector of the equipment collected from the energy management system in time period t includes the gas turbine power generation, waste heat boiler heat generation, lithium bromide unit cooling and heating power, electric refrigeration unit cooling power, and the measured values ​​of the charging and discharging power of the energy storage device and the heat storage device. The device output prediction vector is calculated from the device-level energy conversion equations; ||·||2 is the Euclidean norm (norm 2). The optimal estimate of the device parameter vector Θ is obtained by regressing historical operating data to solve the above least squares problem; the model consistency verification adopts node energy balance verification based on Kirchhoff's energy conservation law to ensure that the deviation between the multi-energy coupled topology model and the actual operating data is less than 5%.

[0037] The correlation matrix A, coupling matrix C, and equipment-level energy conversion equations output by the model serve as the direct source for establishing power balance, thermal balance, cold balance, and equipment operation constraints in step S104. Meanwhile, the cross-period recursive relationship between SOC energy storage and SOC heat storage given by the model is the physical basis for the forward propagation and reverse tightening of the constraint propagation inference unit in step S105.

[0038] Step S102: Based on the historical operating data, construct a set of joint uncertainty scenarios covering spot electricity prices, output of multiple renewable energy sources, and cooling, heating, and power loads, which is used to characterize the time-series correlation structure of multi-source uncertainty.

[0039] First, statistical modeling is performed on historical spot electricity price time-series data to train a conditional probability model, identifying the intraday peak-valley distribution characteristics of electricity prices, the pattern of peak periods, and the probability of cross-period shifts. Specifically, K-means clustering is first used to analyze historical electricity price λ. t The price is divided into three categories: high price, medium price, and low price. The state variables are... s t ∈{1, 2, 3} correspond to the three types of cluster centers respectively; then estimate the state transition probability between adjacent time periods, the th t The time period transitions from state i to the 1st state. t The conditional probability of state j in time interval +1 is defined as follows: a ij , t = P(s t+1 = j | s t = i) = N ij , t Σ j N ij , t (1) In the formula: s t Let s be the state of electricity price during time period t. t∈ {1, 2, 3}; i and j are the state values ​​for time periods t and t+1, respectively, i, j ∈ {1, 2, 3}; a ij , t N represents the conditional probability of transitioning from state i to state j in time period t+1; ij , t Σ represents the historical frequency of transitions from state i to state j during time period t; j N ij , t Let be the sum of the historical frequencies of transitions from state i to all possible states during time interval t. In each state s... t Under the condition of =i, the electricity price λ t It follows a Gaussian mixture emission distribution: f(λ t | s t = i) = Σ m=1 w im · (λ t μ im ,σ² im (2) In the formula: λ t Let t be the electricity price for time period t; M be the number of Gaussian components; m be the component number, m ∈ {1, 2, ..., M}; w im μ im σ im These are the weights, mean, and standard deviation of the m-th Gaussian component in state i, respectively. The parameters are estimated from historical data using the EM algorithm. (μ, σ²) represents the normal distribution density function with mean μ and variance σ². Based on the above model, the conditional probability sequence sampling method is used to extract the state sequence s time-by-time. t The electrical value λ is sampled from the corresponding Gaussian mixture distribution. t This generates multiple electricity price scenario curves covering typical forms of high, medium, and low prices, enabling the scenario set to represent the main characteristics of electricity price distribution.

[0040] Then, parameterized probability modeling was performed on the prediction errors of photovoltaic and wind power output respectively, and the Copula function was used to synthesize the joint error scenario.

[0041] (1) Beta distribution modeling of photovoltaic power output prediction error. Considering that photovoltaic power output is significantly affected by weather type, the prediction error of photovoltaic power output is normalized under three weather categories: sunny, cloudy, and overcast. (After linear transformation to the [0, 1] interval) Establish a Beta distribution model, whose probability density function is shown in the following equation: (3) In the formula: For the first t The time-normalized photovoltaic (PV) forecast error, ranging from [0, 1], is obtained by linear transformation of the original PV output forecast error; α and β are two shape parameters of the Beta distribution, α>0 and β>0, respectively calculated by the method of moments from the mean and variance of historical samples under the corresponding weather type; B(α, β) is the Beta function, defined as... It plays a normalization role in the probability density function. The bounded nature of the Beta distribution can accurately characterize the actual distribution range of photovoltaic power output error, avoiding non-physical out-of-bounds scenarios.

[0042] (2) Weibull distribution modeling of wind power output prediction error. Considering the skewed distribution characteristics of wind power output error, the modeling of wind power output prediction error is modeled using the Weibull distribution model. The Weibull distribution model is established, and its probability density function is shown in the following equation: (4) In the formula: Let t be the wind power forecast error for time period t. >0; k is the shape parameter, which describes the skewness of the distribution; c is the scale parameter, which describes the broadening of the distribution; parameters k and c are obtained from historical samples using the maximum likelihood estimation method.

[0043] (3) Synthesis of Copula Function for Multi-Energy Correlation. Considering the correlation between photovoltaic and wind power output errors caused by common meteorological conditions, the Spearman rank correlation coefficient ρ between the two is fitted using historical data from the same period, and the joint distribution is synthesized using the Gaussian Copula function: C(u, v; ρ) = Φ2(Φ - ¹(u), Φ - ¹(v); ρ)(5) In the formula: u and v are the cumulative probability values ​​of the marginal distributions of photovoltaic and wind power errors, respectively, with values ​​ranging from [0, 1]; ρ is the Spearman rank correlation coefficient of photovoltaic and wind power errors, with values ​​ranging from [-1, 1]; Φ - ¹(·) is the inverse function of the standard normal distribution; Φ2 is the cumulative distribution function of the bivariate standard normal distribution with a correlation coefficient of ρ. Based on the above Gaussian Copula function, bivariate correlated samples with a specified correlation structure are sampled, and then the correlation error scenarios of photovoltaic and wind power are obtained by inverse function transformation of the Beta distribution and Weibull distribution, respectively. These are then superimposed on the point prediction values ​​to obtain a set of output error scenarios reflecting the spatiotemporal correlation of multiple energy sources.

[0044] Furthermore, a joint modeling of the demand uncertainties for cooling, heating, and electricity loads is conducted. First, the prediction residual sequences for cooling, heating, and electricity loads are extracted from historical load data and stratified by season and intraday time period to construct corresponding prediction error sample sets. Then, considering the coupling relationship between the three load types in terms of building thermal inertia, production process linkage, and user behavior, a ternary normal distribution is used to jointly model the prediction errors for cooling, heating, and electricity. The mean and covariance matrix are estimated from historical samples, where the diagonal elements of the covariance matrix reflect the uncertainty intensity of each type of load, and the off-diagonal elements reflect the correlation between cooling, heating, and electricity. Finally, based on the estimated covariance matrix, the Cholesky decomposition method is used to generate load error scenarios with a specified correlation structure, which are then superimposed on the predicted load values ​​to obtain a joint scenario set for cooling, heating, and electricity loads. This scenario set, together with the aforementioned electricity price scenario set and renewable energy output scenario set, constitutes a multi-source uncertainty joint scenario set, serving as input for subsequent scenario reduction and feasible region characterization.

[0045] Combining the aforementioned electricity price and power output error scenarios, a Fast Forward Reduction (FFR) algorithm is employed to compress the initial large-scale joint scenario set to a preset size. This algorithm aims to minimize the Wasserstein distance and iteratively selects the most representative scenarios: in each iteration, the Kantorovich distance between each candidate scenario and the retained scenario set is calculated sequentially, and the scenario with the smallest distance is added to the retention set. This process is repeated until the number of retained scenarios reaches the preset size. Finally, the probabilities of deleted scenarios are merged into the retained scenario with the closest Euclidean distance, ensuring that the reduced scenario set approximates the original set as closely as possible in terms of probability distribution. This algorithm has a time complexity of O(N²), capable of compressing tens of thousands of scenarios to hundreds within seconds, ensuring a reasonable balance between accuracy and efficiency in subsequent calculations.

[0046] Step S103: The nonparametric kernel density estimation algorithm based on Gaussian kernel function is used to perform probability density modeling on the net power purchased from the historical operating data. Combined with the data-driven method of quantile extraction and time-series clustering, the empirical feasible boundary features of the net power purchased are extracted to form a prior knowledge base.

[0047] The data-driven algorithm comprises three parts: first, a Gaussian kernel density estimation method that uses the Silverman empirical rule to determine the bandwidth parameter, used to fit the empirical probability density of net power purchases in each time period; second, a quantile extraction algorithm based on the empirical density function, used to characterize the hierarchical empirical boundary; and third, a time-series clustering algorithm based on the dynamic time warping distance (DTW), used to classify historical operating days according to the shape of the net power purchase curve.

[0048] Historical operational data is categorized into multiple dimensions based on date type (weekday, weekend, holiday), season (spring, summer, autumn, winter), and weather conditions (sunny, fair, severe). For each category, a historical sample set of net power purchase for each scheduling period is extracted.

[0049] Kernel density estimation is performed on the historical distribution of net electricity purchases for each time period under each scenario, and the marginal distribution is fitted to obtain the empirical density function of net electricity purchases for each time period. Specifically, for the historical samples of net electricity purchases for time period t in category c, an empirical probability density function is constructed using a Gaussian kernel function: f(p) = (1 / (N·h)) · Σ K((p - P i ) / h)(6) Where: p is the net purchased power value; N is the number of historical samples for this category during this period; P i Let be the i-th historical sample value; K(·) be the standard Gaussian kernel function; h be the kernel bandwidth, selected according to the Silverman rule h = 1.06·σ·N^(-1 / 5), where σ is the standard deviation of the historical samples. Then, the quantile boundaries are calculated using the empirical cumulative distribution function: q(α) = inf{p : F(p) ≥ α} (7) In the formula: F(p) is the empirical cumulative distribution function obtained by integrating f(p); α takes values ​​of 0.05, 0.25, 0.50, 0.75, and 0.95, corresponding to the 5%, 25%, 50%, 75%, and 95% quantile boundaries; inf{·} is the infimum operator. Simultaneously, the maximum value Pmax and minimum value Pmin of the net power purchased during this period are directly obtained from historical samples.

[0050] Arrange the above 7 boundary values ​​from low to high to form a hierarchical empirical boundary description vector for each time period: b = [Pmin, q(0.05), q(0.25), q(0.50), q(0.75), q(0.95), Pmax] (8) In the formula: b is a 7-dimensional column vector, which describes the multi-level boundaries of the historical distribution of net power purchases during the period. Based on the spacing between adjacent quantiles in the description vector b, dense regions (segments with spacing less than 0.5σ) and sparse regions (segments with spacing greater than 0.5σ) of the historical feasible intervals are identified.

[0051] The daily curve shape feature vector of the average net power purchase for each time period is extracted. Specifically, based on historical operating data, the time-series samples of net power purchase are first classified according to season, date type, and weather conditions. Then, the average value of the net power purchase curve for each operating day under the same category is calculated for each time period to obtain the average daily curve of net power purchase for that category. Finally, curve shape indicators such as peak-to-valley ratio, load factor, average rate of change of the climbing segment, average rate of change of the falling segment, and distribution of the transfer amplitude of net power purchase in adjacent time periods are extracted from the average daily curve, and these indicators are combined to form the daily curve shape feature vector. The daily curve shape feature vector is used to characterize the time-series variation characteristics of the empirically feasible boundary of net power purchase and is used for similar scenario matching and initial value selection for hot start in the subsequent numerical simulation co-solution stage.

[0052] Step S104: Based on the physical model of the system, establish equipment operation constraints and multi-energy coupling balance constraints, and construct a feasible domain optimization characterization model that satisfies the equipment operation constraints and multi-energy coupling balance constraints.

[0053] Establish full-time-domain multi-energy coupling balance constraints and time-period equipment operation constraints for the cooling, heating, and power microgrid.

[0054] The multi-energy coupling balance constraints include power balance constraints, thermal balance constraints, and cold balance constraints. Specifically, in the power balance constraint, the net purchased power, gas turbine power generation, energy storage device discharge power, and renewable energy output in each time period jointly meet the power demand of the electrical load, the power consumption of the electric chiller, and the charging power of the energy storage device; in the thermal balance constraint, the heat supply of the waste heat boiler, the waste heat recovery of the gas turbine, and the heat release of the energy storage device in each time period jointly meet the heat demand of the heat load, the driving heat of the lithium bromide unit, and the charging heat of the energy storage device; in the cold balance constraint, the cooling capacity of the lithium bromide unit and the cooling capacity of the electric chiller in each time period jointly meet the cooling load demand.

[0055] The equipment operation constraints include gas turbine operation constraints, energy storage device operation constraints, thermal storage device operation constraints, lithium bromide unit operation constraints, and cumulative equipment operation constraints. Specifically, gas turbine operation constraints limit the gas turbine output to between the minimum stable output and the rated output, and limit the output variation between adjacent time periods to not exceed its ramping capacity; energy storage device operation constraints limit the energy storage device's state of charge to be continuously progressive between adjacent time periods, keeping the state of charge within the allowable upper and lower limits, with charging and discharging power not exceeding the corresponding power limits, and restricting simultaneous charging and discharging within the same time period; thermal storage device operation constraints limit the thermal storage state to be continuously progressive between adjacent time periods, keeping the thermal storage state within the allowable upper and lower limits, with charging and discharging power not exceeding the corresponding power limits, and considering heat loss during thermal storage; lithium bromide unit operation constraints limit its cooling or heating capacity to be within the equipment's allowable output range, and require that switching between cooling and heating modes meet the minimum continuous operating time requirement; cumulative equipment operation constraints limit the cumulative operating time, start-stop count, or continuous operating time of gas turbines, lithium bromide units, electric chillers, energy storage devices, and thermal storage devices within the entire scheduling cycle to meet the equipment's safe operation requirements.

[0056] A robust optimization framework is adopted to handle uncertainties: an upper bound subproblem and a lower bound subproblem are constructed to solve the feasible region of net purchased power time series. The feasible region boundary is obtained by maximizing and minimizing the net purchased power in each time period. That is, the feasible region of net purchased power time series is defined as the set of net purchased power time series that the system can satisfy the above-mentioned full-time-domain multi-energy coupling balance constraint of the cold, heat and electricity microgrid and the equipment operation constraint of each time period by reasonably allocating the output of each device under all typical uncertainty scenarios. This ensures the sufficiency of the coverage of the output feasible region in actual operation.

[0057] Step S105: Using the digital-analog collaborative driving algorithm, the empirical boundary of net purchased power is retrieved from the prior knowledge base. Based on the constraint propagation reasoning mechanism and lightweight neural network, the temporal feasible domain of net purchased power for each time period is solved, and the multi-energy coupling balance constraint and equipment operation constraint are substituted for verification.

[0058] Based on the type label of the operating scenario on the day, the empirical boundary of the most similar historical category is retrieved from the prior knowledge base. The 75th and 25th percentiles are used as the initial estimates of the upper and lower bounds of net power purchase for each time period, and are passed to the subsequent optimization solver as the initial value for hot start, avoiding inefficient iteration starting from a large search range.

[0059] Based on the continuity of the state of charge of energy storage and thermal storage devices, a feasible region propagation rule is established between time periods. Specifically, the state recursion relationship of the energy storage device is expressed as: (9) in, Indicates the first The status of the time-storage energy storage device; Indicates the first The status of the time-storage energy storage device; and They represent the first The charging and discharging power of the time-storage energy storage device; and These represent charging efficiency and discharging efficiency, respectively. Indicates the length of the scheduling period; Indicates the first Energy loss during a given period. When the energy storage device is an electrical storage device. Indicates the state of charge; when the energy storage device is a thermal storage device. This indicates a state of heat storage.

[0060] Based on the above recursive relationship of states, if the first state is known... The upper limit of net power purchase for a given period is then used to deduce the first [period] by combining the equipment output limit, cooling, heating and power load, and renewable energy output for that period. The achievable state range of time-limited energy storage devices and thermal storage devices, thereby determining and tightening the first... Feasible range of net power purchase during the time period.

[0061] The reverse tightening refers to starting from the end of the scheduling cycle or the known requirements of subsequent periods, and reversing the judgment to determine the energy storage margin and heat storage margin that should be retained in the preceding periods. For example, when the first... When a period of time requires meeting higher electrical, thermal, or cooling loads, it is necessary to [further details needed]. At the end of the time period, the energy storage device or thermal storage device retains sufficient usable energy; if the first The value of net power purchase at a certain time period will lead to an excessively low state of energy storage, causing the first... If the load demand or equipment operating constraints cannot be met during a certain period, then this value will be changed from the previous value. The feasible range of the time period is eliminated. By alternating between forward derivation and reverse tightening, net power purchase values ​​that would lead to energy storage exceeding limits, insufficient energy supply, or equipment power exceeding limits in subsequent time periods are gradually eliminated, so that the feasible range of net power purchase for each time period converges and stabilizes.

[0062] Using a comprehensive daily feature vector as input, the comprehensive daily feature vector includes load forecast curve features, multiple renewable energy output forecast features, electricity price scenario features, historical boundary features of net purchased power, and intraday variation features of net purchased power. Specifically, the load forecast curve features are extracted from the forecast curves of cooling load, heating load, and electrical load and their combined scenarios in step S102; the multiple renewable energy output forecast features are extracted from the output forecast curves of photovoltaic, wind power, and other renewable energy sources and their error scenarios in step S102; the electricity price scenario features are extracted from the spot electricity price scenario curve generated in step S102; the historical boundary features of net purchased power are obtained from the empirical boundary description of net purchased power for each time period in step S103; and the intraday variation features of net purchased power are obtained from the peak-to-valley ratio, load factor, average rate of change in the climbing segment, average rate of change in the descending segment, and the distribution of the transfer amplitude between adjacent time periods of the average daily curve of net purchased power in step S103. These features are input into a lightweight fully connected neural network pre-trained on historical data, outputting a fast approximate prediction result of the upper and lower bounds of net purchased power across the entire time domain.

[0063] The net power purchase boundary values ​​predicted by the neural network for each time period are substituted into the multi-energy coupling balance constraints and equipment operation constraints for verification. For time periods where constraints are found to be violated, local exact optimization is initiated in the vicinity of the neural network prediction value to correct the violated boundary value with the minimum adjustment range, ensuring that the final output full-time feasible domain boundary satisfies all physical constraints.

[0064] Step S106: Output the characterization results of the time-series feasible domain of the net purchased power, and perform multi-scenario feasibility verification under typical uncertainty scenarios.

[0065] The upper and lower bound sequences of net power purchases for each time period obtained through the combined numerical simulation and computational modeling solution are used as boundary representations of the temporal feasible region. At the same time, the cross-time period coupling constraint description between key time periods is output. Combined with the electricity spot market declaration rules, a recommended declaration range for net power purchases is generated.

[0066] To facilitate direct invocation by the downstream scheduling decision-making module, the temporal feasible domain is compactly represented. Specifically, firstly, the upper and lower bounds of net purchased power for each time period are retained to describe the range of net purchased power within a single time period; secondly, the maximum allowable variation in net purchased power between adjacent time periods is retained to describe the limits on the rise and fall of the net purchased power curve between time periods; thirdly, the cross-time period coupling constraints formed by the continuity of the states of energy storage and thermal storage devices are retained to describe the impact of the operating state of the previous time period on the feasible range of the next time period; finally, the key operating limitations of gas turbines, lithium bromide units, electric chillers, energy storage devices, and thermal storage devices are retained to ensure that the recommended application interval meets the equipment safety operation requirements.

[0067] After forming the above constraint set, duplicate constraints, constraints completely contained by other constraints, and redundant constraints that do not affect the feasible region boundary in all typical scenarios are deleted. Only the necessary constraints that limit the net power purchase boundary and cross-time period feasibility are retained. Thus, the original high-dimensional constraint set is reorganized into a simplified constraint set including "time period upper and lower bounds, adjacent time period change restrictions, cross-time period energy storage state restrictions, and key equipment operation restrictions". The feasible region representation results are stored in this simplified constraint set, thereby achieving the compact representation.

[0068] Several test scenarios are extracted from the uncertain scenarios. For each test scenario, it is verified whether the system can keep the net purchased power within the defined feasible region at all times by reasonably allocating the output of the equipment, while satisfying all multi-energy balance and equipment operation constraints. The probability of operation within the feasible region is statistically analyzed and a feasibility verification report is generated.

[0069] Furthermore, this application establishes an online rolling correction mechanism for the feasible domain. It monitors real-time updates of spot electricity prices, measured deviations in the output of various renewable energy sources, and measured deviations in load. When changes in key parameters exceed preset thresholds, it automatically triggers local corrections of the feasible domain for the affected period. This incremental update maintains the real-time accuracy of the feasible domain representation across the entire time domain, meeting the online application requirements for intraday rolling optimization.

[0070] The beneficial effects of this invention are that, compared with the prior art, this invention provides a method and system for determining the time-series feasible domain of net purchased power in a cooling, heating, and electric microgrid, which has the following advantages: 1. Comprehensive Uncertainty Characterization: By using a joint scenario set method to simultaneously consider three types of uncertainties and their temporal correlations—spot electricity prices, output from multiple renewable energy sources, and cooling, heating, and power loads—the robustness of the feasible region characterization results is significantly enhanced, enabling the safe operation of microgrids in complex market environments.

[0071] 2. Efficient Solution Through Collaborative Mathematical Modeling: This approach integrates data-driven prior boundary features with precise constraints from a physical model. Prior warm-start and constraint propagation significantly compress the search space, while lightweight neural networks accelerate inference. This reduces the online solution time for full-time feasible domain characterization to minutes. Simultaneously, a physical correction mechanism ensures physical consistency of the results, overcoming the limitations of purely data-driven and purely model-driven methods. Tests show that the solution efficiency is improved by over 80% compared to purely physical model methods, while avoiding the risk of pseudo-feasible solutions generated by purely data-driven methods.

[0072] 3. Complete characterization of temporal coupling: Through the cross-time period constraint propagation of the charge state of the energy storage device, the coupling relationship between each time period in the temporal feasible domain of net purchased power is fully preserved. The output results not only include the boundary intervals of each time period, but also the joint feasible set information between time periods, that is, the constraint relationship of feasible combinations. It can be directly used to formulate multi-time period collaborative scheduling strategies, providing a complete constraint description basis for formulating multi-time period collaborative optimization scheduling strategies.

[0073] 4. Online rolling adaptive: The feasible region rolling correction mechanism enables the characterization results to be dynamically adjusted as spot prices are updated in real time and source-load forecast deviations are corrected, meeting the online characterization requirements for multiple rolling optimizations within the day. The inference latency reaches the minute level, which meets the real-time requirements of power spot market dispatch and improves the adaptability of microgrid dispatch decisions in complex market environments.

[0074] 5. Direct support for market decision-making: The output time-series feasible domain can be directly connected to the application process of the electricity spot market, providing microgrid operators with a quantified net power purchase application range, effectively reducing the risk of market clearing deviation caused by improper application strategies.

[0075] Simulation results show that, compared with traditional methods, the present invention improves the accuracy of feasible domain boundary characterization by 75% in typical uncertainty scenarios, increases the temporal feasible domain coverage to over 98%, and shortens the online solution time by 82%, effectively supporting the operational decision-making of CHP microgrids participating in the electricity spot market.

[0076] This invention can be a system, method, and / or computer program product. This invention also discloses a system for determining the feasible region of net purchased power time series in a cooling, heating, and electric microgrid based on the aforementioned method for determining the feasible region of net purchased power time series in a cooling, heating, and electric microgrid. The system includes a data modeling module, an uncertainty characterization module, a data-driven boundary extraction module, a feasible region optimization module, a mathematical-analog co-solution module, and a result output verification module. The data modeling module is used to acquire the equipment configuration parameters and historical operating data of the energy management system and equipment data acquisition system of the microgrid (CHP), and to construct a system physical model containing topology information for the conversion of electrical, thermal, and cooling energy based on the equipment configuration parameters and historical operating data. The equipment configuration parameters include the rated capacity, conversion efficiency, and operating limits of gas turbines, waste heat boilers, lithium bromide absorption chillers, electric chillers, thermal storage devices, and energy storage devices. The historical operating data includes historical time-series records of net purchased power, equipment output, CHP load, and renewable energy output. The uncertainty characterization module is used to construct a set of joint uncertainty scenarios covering spot electricity prices, output of multiple renewable energy sources, and cooling, heating, and power loads based on the historical operating data, which is used to characterize the time-series correlation structure of multi-source uncertainty. The data-driven boundary extraction module is used to perform probability density modeling of the net power purchased from the historical operating data using a nonparametric kernel density estimation algorithm based on Gaussian kernel function, and to extract empirical feasible boundary features of net power purchased by combining quantile extraction and time-series clustering data-driven methods to form a prior knowledge base. The feasible region optimization module is used to establish equipment operation constraints and multi-energy coupling balance constraints based on the system physical model, construct a feasible region optimization characterization model that satisfies the equipment operation constraints and multi-energy coupling balance constraints, and characterize the feasible range of net purchased power as a set of constraints coupled between time periods. The digital-analog collaborative solution module is used to retrieve the empirical boundary of net purchased power from the prior knowledge base using a digital-analog collaborative driving algorithm, solve the temporal feasible domain of net purchased power for each time period based on the constraint propagation reasoning mechanism and lightweight neural network, and verify it by substituting it into the multi-energy coupling balance constraint and equipment operation constraint. The result output verification module is used to output the characterization results of the time-series feasible domain of the net purchased power and to perform feasibility verification under the joint uncertainty scenario set.

[0077] In a preferred embodiment, the above-mentioned mathematical-analog co-solution module includes: The prior hot-start unit is used to retrieve the empirical boundary of the most similar historical category from the prior knowledge base as the initial value for the hot start of net power purchase in each time period; The constraint propagation reasoning unit is used to tighten the feasible range of net power purchase for each time period by using the constraint propagation reasoning mechanism and utilizing the coupling relationship of energy storage state between adjacent time periods; The neural network acceleration unit is used to take the daily feature vector as input, which includes load forecast curve features, various renewable energy output forecast features, electricity price scenario features and prior boundary features, and calls a lightweight fully connected neural network pre-trained on historical data to output the prediction results of the upper and lower bounds of net power purchase in the entire time domain. The physical correction unit is used to substitute the net power purchase boundary values ​​predicted by the neural network for each time period into the multi-energy coupling balance constraints and equipment operation constraints for verification. For time periods where constraints are found to be violated, local precise optimization is initiated to correct the violation boundary values ​​with the minimum adjustment magnitude.

[0078] Based on the spirit of this invention, those skilled in the art will readily conceive of a computer program product that can be obtained based on the aforementioned method for determining the timing feasible domain of net purchased electricity in a microgrid (cooling, heating, and power) system. The computer program product may include a computer-readable storage medium on which computer-readable program instructions are loaded to enable a processor to implement various aspects of this disclosure. That is, this application also includes a terminal comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the aforementioned method for determining the timing feasible domain of net purchased electricity in a microgrid (cooling, heating, and power) system.

[0079] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example, but not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0080] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0081] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for determining the time-series feasible domain of net purchased power in a microgrid with cooling, heating, and electricity, characterized in that, Includes the following steps: The system acquires equipment configuration parameters and historical operating data from the energy management system and equipment data acquisition system of the microgrid (cooling, heating, and power), and constructs a system physical model containing topology information for the conversion of electrical, thermal, and cooling energy based on these parameters and data. The equipment configuration parameters include the rated capacity, conversion efficiency, and operating limits of gas turbines, waste heat boilers, lithium bromide absorption chillers, electric chillers, thermal storage devices, and energy storage devices. The historical operating data includes historical time-series records of net purchased power, equipment output, cooling, heating, and power loads, and renewable energy output. Based on the historical operating data, a set of joint uncertainty scenarios covering spot electricity prices, output of various renewable energy sources, and cooling, heating, and power loads is constructed to characterize the time-series correlation structure of multi-source uncertainty. A nonparametric kernel density estimation algorithm based on Gaussian kernel function is used to perform probability density modeling on the net power purchase of the historical operating data. Combined with data-driven methods of quantile extraction and time-series clustering, empirical feasible boundary features of net power purchase are extracted to form a prior knowledge base. Based on the physical model of the system, equipment operation constraints and multi-energy coupling balance constraints are established, and a feasible region optimization characterization model that satisfies the equipment operation constraints and multi-energy coupling balance constraints is constructed. The feasible range of net purchased power is characterized as a set of constraints coupled between time periods. The algorithm of digital-analog joint drive is adopted to retrieve the empirical boundary of net purchased power from the prior knowledge base. Based on the constraint propagation reasoning mechanism and lightweight neural network, the temporal feasible domain of net purchased power in each time period is solved, and the multi-energy coupling balance constraint and equipment operation constraint are substituted for verification. Output the time-series feasible domain representation results of the net purchased power, and perform feasibility verification under the joint uncertainty scenario set.

2. The method for determining the time-series feasible domain of net purchased power in a microgrid (cooling, heating, and electricity) according to claim 1, characterized in that, The step of constructing a joint uncertainty scenario set covering spot electricity prices, output of various renewable energy sources, and cooling, heating, and power loads based on the historical operating data further includes: Statistical modeling was performed on historical spot electricity price time series data to train a conditional probability model, identify the intraday peak-valley distribution characteristics of electricity prices, the occurrence pattern of peak periods and the probability of cross-period shifts, and generate multiple electricity price scenario curves covering typical high, medium and low price patterns using a conditional probability sequence sampling method. A Beta distribution model was established for the historical prediction error of photovoltaic power output according to weather classification; a Weibull distribution model was established for the prediction error of wind power output; the rank correlation coefficient between the power output of each energy source within the same period was fitted using historical data from the same period, and the Copula function was used to synthesize joint error scenarios to generate a set of power output error scenarios that reflect the spatiotemporal correlation of multiple energy sources. By combining electricity price scenarios and power output error scenarios, a fast forward reduction algorithm is used to compress the joint scenario set to a preset size, while retaining typical scenarios that represent the characteristics of uncertainty distribution.

3. The method for determining the time-series feasible domain of net purchased power in a microgrid (cooling, heating, and power) according to claim 2, characterized in that, The method employs a nonparametric kernel density estimation algorithm based on Gaussian kernel function to perform probability density modeling on the net power purchase data of the historical operating data. Combined with data-driven methods of quantile extraction and time-series clustering, empirically feasible boundary features of the net power purchase are extracted to form a prior knowledge base. This further includes: Historical operational data are categorized by season, date type, and weather conditions to construct a categorized historical sample set; Kernel density estimation is performed on the historical distribution of net power purchased in each time period under each scenario, the marginal distribution is fitted, the empirical density function of net power purchased in each time period is obtained, and then the quantile boundary and the historical maximum and minimum values ​​are calculated to form a hierarchical empirical boundary description vector to identify the dense and sparse regions of the historical feasible interval. Extract the daily curve shape feature vector of the average net power purchase for each time period to form the time series feature of the empirical feasible boundary of net power purchase.

4. The method for determining the time-series feasible domain of net purchased power in a microgrid (cooling, heating, and electricity) according to claim 3, characterized in that, The construction of the feasible region optimization characterization model that satisfies the device operation constraints and multi-energy coupling balance constraints further includes: Establish multi-energy coupling balance constraints for cold, heat and electricity microgrids, including power balance constraints, thermal balance constraints and cold balance constraints; Establish equipment operation constraints, including upper and lower limits of gas turbine output, ramp rate constraints, recursive relationship constraints of the state of charge of energy storage devices between adjacent time periods, upper and lower limits of state of charge constraints, and logical mutual exclusion constraints that charging and discharging cannot be carried out simultaneously; similar state recursion and power limit constraints of thermal storage devices; minimum operating time constraints for the switching between hot and cold modes of lithium bromide units; and cumulative operating time constraints for each piece of equipment within the entire scheduling cycle. Construct upper and lower bound subproblems for solving the feasible region of net purchased power time series. Obtain the boundary of the feasible region by maximizing and minimizing the net purchased power in each time period, respectively. That is, the feasible region of net purchased power time series is defined as the set of net purchased power time series that the system can satisfy the above-mentioned full-time multi-energy coupling balance constraint and time-by-time equipment operation constraint of the cold, heat and power microgrid by reasonably allocating the output of each device under all typical uncertainty scenarios.

5. The method for determining the time-series feasible domain of net purchased power in a microgrid (cooling, heating, and electricity) according to claim 4, characterized in that, The method employs a combined numerical and analog-driven algorithm to retrieve the empirical boundary of net purchased power from the prior knowledge base. Based on a constraint propagation reasoning mechanism and a lightweight neural network, it solves the temporal feasible domain of net purchased power for each time period and substitutes it into the multi-energy coupling balance constraints and equipment operation constraints for verification. This further includes: The empirical boundary of the most similar historical category is retrieved from the prior knowledge base and used as the initial value for the hot start of net power purchase in each time period; A constraint propagation reasoning mechanism is adopted, and the energy storage state coupling relationship between adjacent time periods is used to tighten the feasible range of net power purchase for each time period through forward and backward propagation. Taking the daily feature vector as input, which includes load forecast curve features, various renewable energy output forecast features, electricity price scenario features and prior boundary features, a lightweight fully connected neural network pre-trained on historical data is invoked to output the prediction results of the upper and lower bounds of net power purchase in the entire time domain. The net power purchase boundary values ​​predicted by the neural network for each time period are substituted into the multi-energy coupling balance constraints and equipment operation constraints for verification. For time periods where constraints are found to be violated, local precise optimization is initiated to correct the violated boundary values ​​with the minimum adjustment range.

6. The method for determining the time-series feasible domain of net purchased power in a microgrid (cooling, heating, and power) according to claim 5, characterized in that, After outputting the characterization results of the time-series feasible domain of the net purchased power, the method further includes: The upper and lower bound sequences of net power purchases for each time period are used as boundary representations of the time-series feasible domain. The cross-time period coupling constraint description between key time periods is output. Combined with the power spot market declaration rules, the recommended declaration range of net power purchases is generated. Based on the aforementioned temporal feasible domain, feasibility verification is carried out under multiple uncertainty scenarios, and the probability of the system operating within the feasible domain under each scenario is statistically analyzed. When spot electricity prices or source-load forecast information are updated, a rolling correction mechanism for the feasible region is triggered, enabling online dynamic characterization of the time-series feasible region.

7. A system for determining the time-series feasible domain of net purchased power in a microgrid (cooling, heating, and electricity), characterized in that, It includes a data modeling module, an uncertainty characterization module, a data-driven boundary extraction module, a feasible region optimization module, a mathematical-model co-solution module, and a result output verification module; The data modeling module is used to acquire the equipment configuration parameters and historical operating data of the energy management system and equipment data acquisition system of the microgrid (CHP), and to construct a system physical model containing topology information for the conversion of electrical, thermal, and cooling energy based on the equipment configuration parameters and historical operating data. The equipment configuration parameters include the rated capacity, conversion efficiency, and operating limits of gas turbines, waste heat boilers, lithium bromide absorption chillers, electric chillers, thermal storage devices, and energy storage devices. The historical operating data includes historical time-series records of net purchased power, equipment output, CHP load, and renewable energy output. The uncertainty characterization module is used to construct a set of joint uncertainty scenarios covering spot electricity prices, output of multiple renewable energy sources, and cooling, heating, and power loads based on the historical operating data, which is used to characterize the time-series correlation structure of multi-source uncertainty. The data-driven boundary extraction module is used to perform probability density modeling of the net power purchased from the historical operating data using a nonparametric kernel density estimation algorithm based on Gaussian kernel function, and to extract empirical feasible boundary features of net power purchased by combining quantile extraction and time-series clustering data-driven methods to form a prior knowledge base. The feasible region optimization module is used to establish equipment operation constraints and multi-energy coupling balance constraints based on the system physical model, construct a feasible region optimization characterization model that satisfies the equipment operation constraints and multi-energy coupling balance constraints, and characterize the feasible range of net purchased power as a set of constraints coupled between time periods. The digital-analog collaborative solution module is used to retrieve the empirical boundary of net purchased power from the prior knowledge base using a digital-analog collaborative driving algorithm, solve the temporal feasible domain of net purchased power for each time period based on the constraint propagation reasoning mechanism and lightweight neural network, and verify it by substituting it into the multi-energy coupling balance constraint and equipment operation constraint. The result output verification module is used to output the characterization results of the time-series feasible domain of the net purchased power and to perform feasibility verification under the joint uncertainty scenario set.

8. The system for determining the time-series feasible domain of net purchased power in a microgrid with cooling, heating, and electricity according to claim 7, characterized in that, The digital-analog collaborative solving module further includes: The prior hot-start unit is used to retrieve the empirical boundary of the most similar historical category from the prior knowledge base as the initial value for the hot start of net power purchase in each time period; The constraint propagation reasoning unit is used to tighten the feasible range of net power purchase for each time period by using the constraint propagation reasoning mechanism and utilizing the coupling relationship of energy storage state between adjacent time periods; The neural network acceleration unit is used to take the daily feature vector as input, which includes load forecast curve features, various renewable energy output forecast features, electricity price scenario features and prior boundary features, and calls a lightweight fully connected neural network pre-trained on historical data to output the prediction results of the upper and lower bounds of net power purchase in the entire time domain. The physical correction unit is used to substitute the net power purchase boundary values ​​predicted by the neural network for each time period into the multi-energy coupling balance constraints and equipment operation constraints for verification. For time periods where constraints are found to be violated, local precise optimization is initiated to correct the violation boundary values ​​with the minimum adjustment magnitude.

9. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method for determining the time-series feasible domain of net power purchase for a microgrid as described in any one of claims 1 to 6.

10. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute the time-series feasible domain determination method for net power purchase of a microgrid as described in any one of claims 1 to 6.

Images