A method and system for optimizing electricity purchase plans on the retail side based on load fractal characteristics

Abstract

This invention provides a method and system for optimizing electricity purchase plans on the retail side based on load fractal characteristics, belonging to the field of electricity market technology. This invention extracts load fractal characteristics reflecting their correlation features by performing modal decomposition on historical load and bidding data; analyzes electricity market rules to form constraint units and identifies risk-sensitive nodes that cause fluctuations in electricity purchase costs; performs perturbation correction on the fractal characteristics based on these sensitive nodes to generate corrected load forecast results; maps the forecast results to the constraint structure to construct the correlation topology between load characteristics and electricity purchase costs; further extracts the cost sensitivity distribution, establishes a hierarchical mapping model, performs structured optimization, and outputs the electricity purchase plan. This invention improves the responsiveness of electricity purchase strategies to market changes and the level of cost control under complex market conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electricity market, and particularly relates to a power purchase plan optimization method and system based on load fractal characteristics. BACKGROUND

[0002] With the deepening of the reform of electricity market, the operation decision mode of the power selling side gradually changes from rule-driven to data-driven as a key participant in the electricity trading system; under the background of multiple trading mechanisms and complex bidding strategies coexisting, the power selling enterprise not only needs to accurately respond to market price fluctuations, but also needs to realize the dual goals of cost control and benefit maximization under various transaction constraints; however, due to the strong nonlinear and multi-scale evolution characteristics between load demand and market price, the traditional power purchase plan compilation method cannot effectively depict the potential correlation between them, resulting in enlarged prediction error and high plan rigidity, which further restricts the flexibility and economy of the power selling strategy.

[0003] On the other hand, the transaction rules, clearing mechanisms and capacity constraints involved in the electricity market have obvious structural coupling characteristics, and such structural constraints often form non-explicit disturbance effects on the cost structure of the power selling behavior; the existing methods generally regard various market constraints as static boundary conditions, lack the ability to model the dynamic transmission path of the constraints in the time sequence evolution, and thus cannot identify the sensitive response relationship of the key nodes in the constraint chain to the cost fluctuations, affecting the forward-looking and robustness of the plan optimization.

[0004] Therefore, an intelligent power purchase optimization method that can integrate the multi-scale response characteristics of load and price and combine the structural characteristics of market constraints is needed to improve the prediction accuracy and strategy adaptability of the power selling side in a complex trading environment. SUMMARY

[0005] The present application aims to at least solve one of the technical problems existing in the prior art; for this purpose, the present application provides a power purchase plan optimization method and system based on load fractal characteristics.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions: In a first aspect, the present application provides a power purchase plan optimization method based on load fractal characteristics, which comprises: modal decomposition processing of historical load data and historical market bidding data to extract load fractal characteristics representing the mapping relationship between load and market bidding; disturbance transmission analysis of the load fractal characteristics according to market transaction constraints to determine risk-sensitive nodes causing power purchase cost fluctuations; disturbance transmission correction of the load fractal characteristics based on the risk-sensitive nodes to generate a disturbed load prediction result. Based on the load forecast results and the risk-sensitive nodes, a correlation topology between load fractal characteristics and electricity purchase costs is established to output the electricity purchase plan on the sales side.

[0007] Furthermore, the extraction of load fractal features representing the mapping relationship between load and market price includes: Based on the time-scale unfolding structure of historical load data, modal decomposition is performed on the historical load data to obtain a set of load modes that characterize the multi-scale fluctuation pattern of load. Based on the temporal evolution structure of historical market price data, modal decomposition is performed on the historical market price data to obtain a set of price modes used to characterize the multi-scale change pattern of prices. Based on the structural correspondence between the load mode set and the price mode set at different time scales, the load fractal characteristics are determined.

[0008] Furthermore, the identification of risk-sensitive nodes that trigger fluctuations in electricity purchase costs includes: The market transaction constraints are parsed into multiple constraint action units, and the dependencies between constraint action units are determined. By incorporating load fractal characteristics into the dependency relationship, the transmission path of load fractal characteristics in the constraint unit is analyzed, and the response characteristic values ​​of load fractal characteristics at each constraint unit node are obtained. Based on the structural turning points of the response characteristic values ​​in the transmission path, risk-sensitive nodes are determined.

[0009] Furthermore, determining the dependencies between the constraint-acting units includes: According to market trading rules, market trading constraints are broken down into multiple constraint units; The dependencies between constraint-acting units are determined based on the logical order of each constraint-acting unit.

[0010] Furthermore, the generation of perturbation-constrained load forecast results includes: Based on the location of the risk-sensitive node in the constraint unit, identify the affected fractal structure in the load fractal characteristics; The affected fractal structure is corrected to form a load fractal feature after disturbance conduction correction; A load prediction model is constructed using the load fractal characteristics after disturbance conduction correction, and load prediction results are generated.

[0011] Furthermore, the identification of affected fractal structures in the load fractal features includes: The scope of the disturbance impact is determined based on the location of the constraint unit corresponding to the risk-sensitive node; Based on the range of the disturbance, the corresponding fractal structure in the load fractal characteristics is identified.

[0012] Furthermore, the process of establishing the associated topology includes: The load forecasting results are structurally mapped to the constraint action units corresponding to the risk-sensitive nodes to determine the forecast-constraint action structure. Based on the prediction-constraint structure, a correlation topology between load fractal characteristics and electricity purchase cost is constructed.

[0013] Further, determining the prediction-constraint action structure includes: Determine the structural location of risk-sensitive nodes in the load forecast results; Based on the corresponding positions of the structure, the load prediction results are mapped to the constraint action units to form a prediction-constraint action structure.

[0014] Furthermore, the power purchase plan on the power sales side includes: Based on the prediction-constraint structure, the cost sensitivity distribution of load fractal characteristics under different constraint units is determined; A hierarchical quantitative mapping relationship between load fractal characteristics and electricity purchase cost is established based on cost sensitivity distribution; The structured optimization calculation is performed using the hierarchical quantization mapping relationship to form a power purchase plan for the electricity sales side.

[0015] Secondly, the present invention provides a power purchase plan optimization system based on load fractal characteristics, implemented based on the aforementioned power purchase plan optimization method based on load fractal characteristics, the system comprising: Fractal extraction module: used to perform modal decomposition processing on historical load data and historical market price data, and extract load fractal features that characterize the mapping relationship between load and market price; Constraint Analysis Module: Used to perform disturbance transmission analysis on the load fractal characteristics based on market transaction constraints, and to identify risk-sensitive nodes that cause fluctuations in electricity purchase costs; Correction and prediction module: used to perform disturbance transmission correction on the load fractal characteristics based on the risk-sensitive nodes, and generate a load prediction result constrained by disturbance; Topology optimization module: used to establish a correlation topology between load fractal characteristics and electricity purchase cost based on the load forecast results and the risk-sensitive nodes, so as to output the electricity purchase plan on the sales side.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a dual-modal feature extraction mechanism for load and price based on modal decomposition, which realizes a structured characterization of load response behavior at multiple time scales, improves the ability of the electricity sales side to model the load-price coupling relationship, and provides more discriminative feature support for electricity purchase plans.

[0017] This invention enhances the ability to identify sources of disturbance and analyze the structure of constraint transmission paths during the power purchase optimization process by resolving market transaction constraints into constraint action units with associated structures and identifying risk-sensitive nodes that have an amplifying effect on power purchase costs.

[0018] This invention improves the cost response accuracy and flexible control capability of electricity purchase strategies on the sales side in complex market environments by establishing a topological mapping relationship between prediction results and cost-sensitive nodes, and by combining hierarchical quantification and structured optimization methods. Attached Figure Description

[0019] Figure 1 This is a flowchart of an optimization method for electricity purchase plans on the retail side based on load fractal characteristics, as described in Example 1.

[0020] Figure 2 This is an architecture diagram of a power purchase plan optimization system based on load fractal characteristics for the power sales side, as shown in Example 2. Detailed Implementation

[0021] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1 Please see Figure 1 This invention provides a method for optimizing electricity purchase plans on the retail side based on load fractal characteristics, comprising: Modal decomposition is performed on historical load data and historical market price data to extract load fractal features that characterize the mapping relationship between load and market price. It should be noted that in this embodiment, historical load data of users in the electricity sales area and historical market price data of the market area are collected first. The historical load data specifically refers to the data continuously recorded and uploaded by smart meters in real time, and the data collection frequency is preferably, but not limited to, once every 15 minutes or 30 minutes. The historical market price data specifically refers to the electricity transaction price data of the electricity market in the historical period, including but not limited to the day-ahead market clearing price, real-time market price and ancillary service price.

[0023] It should be further noted that, in order to ensure the representativeness and validity of the data, the historical load data in this embodiment preferably covers the operating conditions of user loads, including but not limited to different load patterns such as peak load periods on weekdays, off-peak load periods on weekdays, and load periods on holidays; the historical market quotation data should also cover the trading patterns of the electricity market, including but not limited to typical trading periods such as peak electricity consumption periods in summer, peak electricity consumption periods in winter, and off-peak electricity consumption periods on weekdays.

[0024] In a specific implementation, the extraction of load fractal features representing the mapping relationship between load and market price includes: Based on the time-scale unfolding structure of historical load data, modal decomposition is performed on the historical load data to obtain a set of load modes that characterize the multi-scale fluctuation pattern of load. In specific implementation, this embodiment uses the Empirical Mode Decomposition (EMD) method or the Variational Mode Decomposition (VMD) method to perform mode decomposition on the historical load data to obtain a multi-scale mode set that can reflect the fluctuation characteristics of the load data. The specific disclosure is as follows: First, the historical load data sequence is represented as The decomposition algorithm decomposes it into multiple eigenmode functions that can reflect the characteristics of local oscillations. And a residual term representing the overall trend. The total number of intrinsic mode functions is determined based on the data characteristics and ranges from 3 to 8.

[0025] It should be noted that each Modal functions represent the local oscillation characteristics of load data at a certain time scale, specifically including short-cycle load fluctuation characteristics (such as hourly fluctuations), medium-cycle load fluctuation characteristics (such as daily fluctuations), and long-cycle load trend characteristics (such as weekly or monthly fluctuations), etc., with different time scale fluctuation patterns.

[0026] Subsequently, based on the decomposition results of the load data, the load mode set was determined as follows: .

[0027] Based on the temporal evolution structure of historical market price data, modal decomposition is performed on the historical market price data to obtain a set of price modes used to characterize the multi-scale change pattern of prices. In this specific implementation, the historical market price data is also processed using Empirical Mode Decomposition (EMD) or Variational Mode Decomposition (VMD) methods to extract a multi-scale mode set from the price data. Specifically, firstly, the historical market price data sequence is represented as... The decomposition algorithm is used to deconstruct it into multiple pricing mode functions that reflect the evolution characteristics at different time scales. And a trend term or residual term representing the overall price trend. The total number of bidding modal functions is determined by a decomposition algorithm based on the complexity of the bidding data, with a preferred value range of 3 to 6.

[0028] Specifically, each Modal functions correspond to the changing characteristics of price data at different scales, including short-term market fluctuations (such as intraday hourly price fluctuations), medium-term market fluctuations (such as intraday price fluctuations), and long-term market trends (such as weekly or monthly price trend fluctuations).

[0029] Furthermore, the set of pricing modes is defined as follows: .

[0030] Based on the structural correspondence between the load mode set and the price mode set at different time scales, the load fractal characteristics are determined.

[0031] In practical implementation, this embodiment is based on the load mode set. With the pricing modal set The structural correspondence between various modal functions is quantitatively described using fractal theory, as detailed below: First, a structural correspondence analysis is performed on each set of load and pricing mode functions (e.g., short-cycle load mode and short-cycle pricing mode); specifically, this is done by setting a sliding window. (The window length can be one day or one week), and it slides along the time axis with a preset step size to calculate the first time within each window. The load mode function and the first The Pearson correlation coefficient sequence between the pricing modal functions is used to quantitatively describe the degree of structural correlation between them at a specified time scale.

[0032] Then, based on the numerical distribution of the correlation coefficients within each sliding window, the time-varying characteristics are determined. Variational load fractal eigenvector .

[0033] It should be understood that the load fractal feature vector is essentially a quantitative description of the multi-scale self-similar structural characteristics of power load under market disturbances by extracting the cross-correlation statistical evolution law between load modes and bidding modes at different time scales.

[0034] It should be noted that, since the correlation between power load and market price varies non-stationary at different frequency components, this embodiment defines this cross-scale consistency mapping as a fractal structure. The feature vector is composed of the statistical parameters of the correlation coefficient at the corresponding scale of each mode, including the mean, variance, maximum and minimum values ​​of the correlation coefficient.

[0035] The load fractal characteristics Specifically, it is expressed as: ; In the formula, , They represent in The first time in the sliding window corresponding to the time The load mode function and the first The mean and variance of the correlation coefficients corresponding to each pricing modal function. , These represent the maximum and minimum values ​​of the correlation coefficient within the window, respectively.

[0036] Based on market transaction constraints, a disturbance transmission analysis is performed on the load fractal characteristics to identify risk-sensitive nodes that cause fluctuations in electricity purchase costs. It should be noted that the market transaction constraints described in this embodiment specifically refer to various transaction rules and constraints that the electricity retailer must follow in the actual process of participating in the electricity market, including but not limited to day-ahead market price range constraints, real-time market clearing price fluctuation constraints, electricity purchase declaration constraints, and system operation safety constraints. This embodiment first decomposes the above constraints using a structured analysis method to form multiple constraint action units with action logic and action boundaries, and determines the dependencies between each constraint action unit. Based on this, combined with the dynamic transmission process of load fractal characteristics between constraint action units, the risk-sensitive nodes that cause fluctuations in the electricity purchase cost of the electricity retailer are accurately identified.

[0037] In a specific implementation, determining the risk-sensitive nodes that trigger fluctuations in electricity purchase costs includes: The market transaction constraints are parsed into multiple constraint action units, and the dependencies between constraint action units are determined. It is understood that this embodiment is based on the actual trading rules and historical trading data of the electricity market. First, the trading constraints of each market are broken down in detail to obtain multiple basic constraint units with functional roles. Then, the logical relationship between each constraint unit is further analyzed and determined to form a complete market constraint structure topology.

[0038] Specifically, determining the dependencies between the constraint-acting units includes: According to market trading rules, market trading constraints are broken down into multiple constraint units; It should be understood that this embodiment is based on specific market rules, and decomposes complex market transaction constraints into multiple constraint units with independent operating logic, including but not limited to: The day-ahead market constraint unit includes the day-ahead market bid price upper and lower limit unit, the day-ahead clearing price formation unit, and the day-ahead market electricity matching unit; The real-time market constraint unit includes a real-time market electricity price adjustment unit and a real-time deviation electricity settlement unit; Capacity constraint units include electricity sales company power purchase capacity restriction units and time-of-use power purchase quota units; System operation safety constraint units include system backup capacity requirement units, load real-time balance requirement units, etc. Through the above steps, this embodiment constructs and obtains a set of constraint action units. Specifically, it is expressed as: ; In the formula, Indicates the first A specific constraint unit, This indicates the total number of all constraint units; the specific number is determined by the actual market rules.

[0039] The dependencies between constraint-acting units are determined based on the logical order of each constraint-acting unit.

[0040] In specific implementation, after completing the above-mentioned decomposition of the constraint action unit, this embodiment further performs logical structure analysis and dependency confirmation on all constraint action units to determine the action order and data transmission relationship between each constraint action unit.

[0041] Specifically, this embodiment is based on the market transaction process and analyzes the input-output logical relationships between the constraint units. For example, the output of the day-ahead market clearing price formation unit directly affects the input data of the real-time deviation electricity settlement unit, and the results of the day-ahead quotation upper and lower limit constraint unit will constrain the day-ahead market electricity matching unit. Thus, a dependency matrix between the constraint units is constructed. : ; Based on the above analysis, a complete topological diagram of the dependency relationships of constraint action units is obtained, which forms the structural basis for determining the load characteristic transmission path.

[0042] By incorporating load fractal characteristics into the dependency relationship, the transmission path of load fractal characteristics in the constraint unit is analyzed, and the response characteristic values ​​of load fractal characteristics at each constraint unit node are obtained. In specific implementation, this embodiment uses the aforementioned load fractal characteristics as input parameters and embeds them into the constraint-dependent structure to specifically analyze the step-by-step mapping and evolution process of load characteristics between each unit.

[0043] First, define the constraint action operator. Used to characterize constraint action unit It can be used for boundary clipping or nonlinear transformation of input features.

[0044] It should be noted that, in this embodiment, the constraint action operator It depends on the specific market rules, such as constraints on the range of market quotations in the current day. It is constructed as a saturated cutoff function to simulate the suppressive effect of the price ceiling on the load-price correlation vector; for clearing price formation constraints, It is constructed as a nonlinear transfer function based on the logic of supply and demand balance, in order to characterize the distortion process of fractal features under market competition.

[0045] The load fractal features at different scales are along the dependency matrix. The defined path is used for iterative calculation to obtain the response characteristic values ​​of the load fractal characteristics at each constraint element node. The response eigenvalue characterizes the fractal features after the first... The state transition after each market constraint unit; specifically represented as: ; In the formula, This represents the vector of physical constraint parameters for the corresponding constraint unit, such as price cap or peak shaving capacity.

[0046] Through the above unit-by-unit analysis, the dynamic response results of the feature among different constraint units are determined.

[0047] Based on the structural turning points of the response characteristic values ​​in the transmission path, risk-sensitive nodes are determined.

[0048] In specific implementation, this embodiment accurately identifies risk-sensitive node locations that can cause significant fluctuations in electricity purchase costs based on the structural changes in response characteristic values ​​that evolve from load fractal characteristics in the transmission path.

[0049] Specifically, by calculating the structural change amplitude of the load fractal characteristics in the response characteristic values ​​between adjacent constraint elements, locations of drastic structural changes are identified as risk-sensitive nodes; in this embodiment, the structural change amplitude (i.e., the change in response intensity) is defined as: ; In the formula, Indicates the fractal characteristics of the load from the first Unit 1 to the 1st When there are 1 unit, the constrained operator The relative change amplitude produced by the action , These represent the response feature values ​​at adjacent constraint nodes in the path.

[0050] Understandably, the magnitude of this value directly reflects the degree to which a specific market constraint distorts the inherent fractal characteristics of the load. The larger the magnitude, the more significant the impact of the constraint unit on the load evolution trend.

[0051] Furthermore, a structural change threshold is set. (Value range is 0.1 to 0.3), when Greater than At that time, the first Each constraint-acting unit is located at a risk-sensitive node; finally, by traversing all constraint-acting units along the entire propagation path, the complete set of risk-sensitive nodes is determined. ;in, The total number of risk-sensitive nodes identified. The constraint unit that satisfies the threshold condition in the corresponding transmission path.

[0052] By traversing all constraint action units throughout the entire transmission path Complete the set of risk-sensitive nodes Extraction.

[0053] Based on the risk-sensitive nodes, the load fractal characteristics are perturbation propagation correction to generate perturbation-constrained load prediction results; It should be understood that, based on the aforementioned risk-sensitive nodes, this embodiment analyzes the influence area and degree of each node on the load fractal characteristics under disturbance, and then performs structured correction on the original load fractal characteristics. On this basis, a load prediction model is established, and the load prediction results after disturbance constraint correction are output.

[0054] In a specific implementation, generating the perturbation-constrained load forecast result includes: Based on the location of the risk-sensitive node in the constraint unit, identify the affected fractal structure in the load fractal characteristics; It is understood that this embodiment determines the specific impact range of the disturbance within the load fractal characteristics based on the risk-sensitive node and the specific structural location of the market transaction constraint unit.

[0055] The identification of affected fractal structures in the load fractal features includes: The scope of the disturbance impact is determined based on the location of the constraint unit corresponding to the risk-sensitive node; Specifically, for each risk-sensitive node, its position in the market constraint unit structure is determined, and the disturbance diffusion area at that position is defined as the disturbance influence range of the risk-sensitive node. In this embodiment, the determination of the disturbance's impact range is specifically achieved by establishing a node impact factor: The node impact factor is defined as the disturbance gain index of risk-sensitive nodes on the fractal structure evolution of the load, used to quantify the degree of correction of load morphological characteristics by specific constraints; in this embodiment, it is determined by fusing historical sensitivity prior values ​​with the current structural change intensity, and the specific calculation formula is as follows: ; In the formula, Indicates the first The node impact factor of each risk-sensitive node. This represents the partial derivative of the electricity purchase cost with respect to the load fractal characteristics under similar constraints, extracted from the historical electricity trading database; it is also known as the sensitivity coefficient. Indicates the first The maximum value of structural change amplitude generated by a risk-sensitive node across all modal scales.

[0056] Furthermore, by using the magnitude of the node influence factor, the scope of the disturbance's impact can be quantified into a set of related nodes centered on the risk-sensitive node, which can be represented as: ; In the formula, Indicates the first The scope of the impact of disturbances on each risk-sensitive node. , These respectively represent risk-sensitive nodes The constraint unit is the one that affects the center before and after. , The value of is determined based on the node influence factor, and the value ranges from 1 to 3.

[0057] Based on the range of the disturbance, the corresponding fractal structure in the load fractal characteristics is identified.

[0058] In specific implementation, this embodiment is based on the aforementioned disturbance influence range. This involves identifying the specific location of the corresponding fractal structure within the load fractal characteristics; specifically, this embodiment uses the correlation mapping relationship between load fractal characteristics and constraint action units to define the disturbance influence range. The mapping relationship to specific modal structures in load fractal features is defined as follows: ; in, Indicates the first The perturbed fractal structure of the load fractal characteristics corresponding to each risk-sensitive node. This represents the mapping function from the range of disturbance influence to the fractal characteristics of the load, specifically implemented using the correlation matching mapping method. This represents the set of load fractal characteristics.

[0059] By following the steps above, the specific structural locations affected by disturbances in the load fractal characteristics are determined, and the set of fractal structures corresponding to each risk-sensitive node is obtained.

[0060] The affected fractal structure is corrected to form a load fractal feature after disturbance conduction correction; In practice, for each affected load fractal structure, the node impact factor of the corresponding risk-sensitive node is used. Structural adjustments are made to reflect the corrective effect of market constraints on load evolution trends; specifically: ; In the formula, Indicates the corrected number The load mode and the first The load fractal characteristics corresponding to each pricing mode are in The value of the moment. This represents the original load fractal characteristic value before correction. This represents the correction intensity factor, with a value ranging from 0.1 to 0.5. The specific value is determined based on historical data statistics. This represents a numerical truncation function used to ensure that the corrected fractal eigenvalues ​​(such as correlation coefficients) remain within a reasonable physical definition range (e.g., [-1, 1]).

[0061] It should be noted that if a fractal structure is simultaneously within the perturbation range of multiple risk-sensitive nodes... Within the range, the node with the largest influence factor is selected. As a basis for revision, or for multiple A weighted average is then applied and corrected to reflect the characteristic distortions under composite constraints.

[0062] After correcting the load fractal features affected by the disturbance one by one through the above steps, a complete set of load fractal features after disturbance conduction correction is formed: .

[0063] A load prediction model is constructed using the load fractal characteristics after disturbance conduction correction, and load prediction results are generated.

[0064] In this specific implementation, the load prediction model is constructed using a deep learning method based on Long Short-Term Memory (LSTM) networks. The specific model construction process is as follows: First, the load fractal feature set after perturbation conduction correction is used. Using historical actual load data as input, a training dataset is constructed as supervised output samples. The dataset is in the following format: ; in, This represents the corrected load fractal characteristic input. This indicates the prediction time lag window, with values ​​ranging from 24 to 96 time periods (e.g., 6 hours to 1 day). Indicates the corresponding time period Actual historical load value, This represents the total number of training samples.

[0065] Secondly, the model is trained using an LSTM network. Specifically, the corrected load fractal feature sequence is modeled using multidimensional long short-term memory to capture the nonlinear evolution of fractal features under constraint perturbation. During the training process, mean square error is used as the loss function, and the weight matrix and bias terms in the network are continuously updated through the backpropagation algorithm until the predicted vector output by the model can accurately fit the actual historical load fluctuation trend after being affected by constraints.

[0066] After the model has been fully trained, the trained LSTM model is used as input, along with the real-time obtained corrected load fractal features, to generate load forecast results for future periods: ; This completes the load forecasting process based on disturbance propagation correction at risk-sensitive nodes, ensuring that the output load forecast results can effectively reflect the actual evolution trend of the load under market transaction constraint disturbances.

[0067] Based on the load forecast results and the risk-sensitive nodes, a correlation topology between load fractal characteristics and electricity purchase costs is established to output the electricity purchase plan on the sales side.

[0068] It is understandable that the structured and quantitative mapping from load characteristics to cost characteristics is achieved through the aforementioned associated topology, thereby effectively supporting the intelligent decision-making and optimization generation of electricity purchase plans on the electricity sales side.

[0069] In a specific implementation, the process of establishing the associated topology includes: The load forecasting results are structurally mapped to the constraint action units corresponding to the risk-sensitive nodes to determine the forecast-constraint action structure. The determination of the prediction-constraint action structure includes: Determine the structural location of risk-sensitive nodes in the load forecast results; In practical implementation, this embodiment addresses each risk-sensitive node. Perform structured analysis to identify its role in load forecasting results. The corresponding position in the text; the method for determining the corresponding position includes: First, by determining the location of the risk-sensitive node within the constraint unit, the corresponding constraint unit's effective period (e.g., day-ahead market clearing period, real-time market adjustment period, etc.) is identified. Second, using this effective period as a reference window, the load forecast value corresponding to this period is extracted from the load forecast results to form the node load forecast vector. .

[0070] Based on the corresponding positions of the structure, the load prediction results are mapped to the constraint action units to form a prediction-constraint action structure.

[0071] In practical implementation, a mapping relationship between the node load prediction vector and the structural parameters of the constraint unit is further established based on the structural characteristics of the constraint unit corresponding to the risk-sensitive node. This embodiment implements the above-mentioned correlation mapping process through a linear or nonlinear mapping function, and the specific mapping formula is expressed as follows: ; In the formula, Represents a node The output value of the prediction-constraint structure formed at the location, The mapping function representing the relationship between nodal load forecasting and constraint units is used in this embodiment, employing a radial basis function as the mapping kernel to characterize the forecasted load under complex electricity market parameters. Nonlinear response characteristics under these conditions; This represents the set of structural parameters of the constraint unit, including parameters such as market price factor and market electricity limit factor.

[0072] Through the above steps, a complete set of prediction-constraint action structures is obtained. ; in, This indicates the number of risk-sensitive nodes.

[0073] Based on the prediction-constraint structure, a correlation topology between load fractal characteristics and electricity purchase cost is constructed.

[0074] In specific implementation, this embodiment is based on a set of prediction-constraint action structures. Furthermore, a correlation topology is constructed between load fractal characteristics and electricity purchase cost. The correlation topology is used to describe the quantitative impact of load fractal characteristics on electricity purchase cost under different market constraints, specifically including two types of elements: topology nodes and topology edges.

[0075] Specifically, in this embodiment, the topology node is defined as the prediction-constraint structure at the risk-sensitive node. Topological edges are defined as the cost impact weights of the propagation effects between nodes. The calculation is as follows: ; In the formula, Indicates the first The node and the first The weights of the topological edges between nodes. The function representing the Pearson correlation coefficient between two nodes is... Represents a node With nodes The logical dependency coefficient within the market constraint unit, when the node... Directly affected by nodes When the effect of the action is affected, ,otherwise .

[0076] The correlation topology between load fractal characteristics and electricity purchase cost is obtained using the above definition method. : .

[0077] Specifically, the power purchase plan on the power sales side includes: Based on the prediction-constraint structure, the cost sensitivity distribution of load fractal characteristics under different constraint units is determined; In practical implementation, this embodiment is based on a prediction-constraint structure. To further determine the distribution of the sensitivity of load fractal characteristics to electricity purchase costs under different constraint units, specific methods include: First, take the node output value Based on this, the partial derivative of the output value of the prediction-constraint structure at each node with respect to the overall electricity purchase cost is calculated to quantify the cost sensitivity of different risk-sensitive nodes; and this is defined as a cost sensitivity index. .

[0078] It should be noted that the aforementioned cost sensitivity index It reflects the weight of the impact of changes in the constraint state at risk-sensitive nodes on the final electricity purchase cost; in specific implementation, the partial derivative is obtained by performing multiple linear regression fitting on historical electricity price and load data, or by numerically calculating by changing the constraint boundary parameters of the node in the power market simulation model and observing the change in electricity purchase cost.

[0079] Secondly, by summarizing the sensitivity indicators of each node, the cost sensitivity distribution vector is obtained: ; Among them, nodes with higher cost sensitivity have a more significant impact on the optimization of electricity purchase plans on the sales side, and will be given priority consideration in the subsequent establishment of cost mapping relationships.

[0080] A hierarchical quantitative mapping relationship between load fractal characteristics and electricity purchase cost is established based on cost sensitivity distribution; In specific implementation, this embodiment is based on cost sensitivity distribution and establishes a quantitative mapping relationship between load fractal characteristics and electricity purchase cost through a hierarchical quantification method, specifically including the following steps: First, cost sensitivity indicators Sensitivity levels are categorized based on numerical values; for example, high sensitivity level (greater than 0.5), medium sensitivity level (greater than or equal to 0.2 and less than or equal to 0.5), and low sensitivity level (less than 0.2). Secondly, a quantitative mapping function between load characteristics and electricity purchase cost is established for different sensitivity levels. The specific mapping function can be a piecewise linear function or a logistic function, etc. In this embodiment, it can be expressed as: ; In the formula, This represents the estimated cost of electricity purchase under the corresponding sensitivity level. Indicates the fractal characteristics of the load. and These represent the cost mapping function parameters corresponding to the sensitivity levels, which are determined through regression analysis of historical data.

[0081] The structured optimization calculation is performed using the hierarchical quantization mapping relationship to form a power purchase plan for the electricity sales side.

[0082] In specific implementation, this embodiment performs differentiated optimization calculations on the decision space corresponding to different sensitivity levels in the hierarchical quantization mapping structure, as follows: For market trading periods and key constraint nodes corresponding to high sensitivity levels, the corrected load forecast results are used as rigid demand constraints. With minimizing the total electricity purchase cost as the objective function, particle swarm optimization algorithm is used to refine the electricity declaration ratio and bidding strategy parameters of the electricity sales side in various trading varieties (day-ahead and real-time). For decision variables corresponding to medium and low sensitivity levels, genetic algorithm or differential evolution algorithm is used to perform structured optimization calculations on the adjustment ratio or deviation assessment margin of medium and long-term contracts, thereby realizing multi-level optimization of the electricity purchase plan under different sensitivity dimensions, and finally outputting the electricity purchase plan scheme of the electricity sales side.

[0083] Example 2 like Figure 2 As shown, this embodiment discloses a power purchase plan optimization system based on load fractal characteristics. For details not covered in this embodiment, please refer to the relevant parts of Embodiment 1 above. The system includes: Fractal extraction module: used to perform modal decomposition processing on historical load data and historical market price data, and extract load fractal features that characterize the mapping relationship between load and market price; Constraint Analysis Module: Used to perform disturbance transmission analysis on the load fractal characteristics based on market transaction constraints, and to identify risk-sensitive nodes that cause fluctuations in electricity purchase costs; Correction and prediction module: used to perform disturbance transmission correction on the load fractal characteristics based on the risk-sensitive nodes, and generate a load prediction result constrained by disturbance; Topology optimization module: used to establish a correlation topology between load fractal characteristics and electricity purchase cost based on the load forecast results and the risk-sensitive nodes, so as to output the electricity purchase plan on the sales side.

[0084] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0085] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for optimizing electricity purchase plans on the retail side based on load fractal characteristics, characterized in that, The method includes: Modal decomposition is performed on historical load data and historical market price data to extract load fractal features that characterize the mapping relationship between load and market price. Based on market transaction constraints, a disturbance transmission analysis is performed on the load fractal characteristics to identify risk-sensitive nodes that cause fluctuations in electricity purchase costs. Based on the risk-sensitive nodes, the load fractal characteristics are perturbation propagation correction to generate perturbation-constrained load prediction results; Based on the load forecast results and the risk-sensitive nodes, a correlation topology between load fractal characteristics and electricity purchase costs is established to output the electricity purchase plan on the sales side.

2. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 1, characterized in that, The extraction of load fractal features representing the mapping relationship between load and market price includes: Based on the time-scale unfolding structure of historical load data, modal decomposition is performed on the historical load data to obtain a set of load modes that characterize the multi-scale fluctuation pattern of load. Based on the temporal evolution structure of historical market price data, modal decomposition is performed on the historical market price data to obtain a set of price modes used to characterize the multi-scale change pattern of prices. Based on the structural correspondence between the load mode set and the price mode set at different time scales, the load fractal characteristics are determined.

3. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 1, characterized in that, The identification of risk-sensitive nodes that trigger fluctuations in electricity purchase costs includes: The market transaction constraints are parsed into multiple constraint action units, and the dependencies between constraint action units are determined. By incorporating load fractal characteristics into the dependency relationship, the transmission path of load fractal characteristics in the constraint unit is analyzed, and the response characteristic values ​​of load fractal characteristics at each constraint unit node are obtained. Based on the structural turning points of the response characteristic values ​​in the transmission path, risk-sensitive nodes are determined.

4. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 3, characterized in that, The determination of the dependencies between constraint-acting units includes: According to market trading rules, market trading constraints are broken down into multiple constraint units; The dependencies between constraint-acting units are determined based on the logical order of each constraint-acting unit.

5. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 3, characterized in that, The generation of perturbation-constrained load forecast results includes: Based on the location of the risk-sensitive node in the constraint unit, identify the affected fractal structure in the load fractal characteristics; The affected fractal structure is corrected to form a load fractal feature after disturbance conduction correction; A load prediction model is constructed using the load fractal characteristics after disturbance conduction correction, and load prediction results are generated.

6. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 5, characterized in that, The identification of affected fractal structures in the load fractal features includes: The scope of the disturbance impact is determined based on the location of the constraint unit corresponding to the risk-sensitive node; Based on the range of the disturbance, the corresponding fractal structure in the load fractal characteristics is identified.

7. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 3, characterized in that, The process of establishing the associated topology includes: The load forecasting results are structurally mapped to the constraint action units corresponding to the risk-sensitive nodes to determine the forecast-constraint action structure. Based on the prediction-constraint structure, a correlation topology between load fractal characteristics and electricity purchase cost is constructed.

8. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 7, characterized in that, The determination of the prediction-constraint action structure includes: Determine the structural location of risk-sensitive nodes in the load forecast results; Based on the corresponding positions of the structure, the load prediction results are mapped to the constraint action units to form a prediction-constraint action structure.

9. The method for optimizing electricity purchase plans on the retail side based on load fractal characteristics according to claim 7, characterized in that, The power purchase plan on the power sales side includes: Based on the prediction-constraint structure, the cost sensitivity distribution of load fractal characteristics under different constraint units is determined; A hierarchical quantitative mapping relationship between load fractal characteristics and electricity purchase cost is established based on cost sensitivity distribution; The structured optimization calculation is performed using the hierarchical quantization mapping relationship to form a power purchase plan for the electricity sales side.

10. A power purchase plan optimization system based on load fractal characteristics, implemented based on the power purchase plan optimization method based on load fractal characteristics according to any one of claims 1-9, characterized in that, The system includes: Fractal extraction module: used to perform modal decomposition processing on historical load data and historical market price data, and extract load fractal features that characterize the mapping relationship between load and market price; Constraint Analysis Module: Used to perform disturbance transmission analysis on the load fractal characteristics based on market transaction constraints, and to identify risk-sensitive nodes that cause fluctuations in electricity purchase costs; Correction and prediction module: used to perform disturbance transmission correction on the load fractal characteristics based on the risk-sensitive nodes, and generate a load prediction result constrained by disturbance; Topology optimization module: used to establish a correlation topology between load fractal characteristics and electricity purchase cost based on the load forecast results and the risk-sensitive nodes, so as to output the electricity purchase plan on the sales side.

Images