Empirical electric quantity distribution method for hydropower station in electricity market environment

Abstract

The invention discloses a hydropower station experience electric quantity distribution method in an electric power market environment, particularly relates to the field of electric power optimization, and is used for solving the problems of accuracy and adaptability of electric quantity distribution of a hydropower station in a complex market environment. A multi-dimensional feature mapping matrix and a dynamic credibility evaluation system are constructed, and accurate quantitative analysis of historical prediction results is realized; based on the evaluation result, the short-term abnormal signal is accurately corrected by adopting a classification identification method, so that the prediction result can reflect the actual dynamic change in time; furthermore, by combining a nonlinear multi-objective optimization model, three core objectives of economic benefit maximization, risk minimization and ecological flow balance are organically fused, and an electric quantity distribution scheme with high economic benefit, low operation risk and ecological protection requirements is generated through a complex mathematical algorithm.

Description

Technical Field

[0001] The present invention relates to the field of power optimization, and more specifically, to an empirical power distribution method for a hydropower station in a power market environment. Background Art

[0002] In the power market environment, the power distribution method of hydropower stations usually relies on the combination of historical data and short-term real-time data. Historical data includes information such as accumulated power generation, hydrological conditions and market prices over many years, which can reveal the general trend of power generation laws and market fluctuations. Short-term real-time data reflects the current hydrological changes, equipment operating conditions and market transaction dynamics, and can provide instant feedback for decision-making. By combining these two types of data, hydropower stations hope to achieve a balance between long-term stability and short-term flexibility in power distribution. However, the differences in sources and characteristics between historical data and short-term data make the fusion processing of the two a complex challenge, especially in the credibility assessment of prediction results and decision optimization.

[0003] In the data processing process, the existing technology usually directly formulates the power allocation strategy based on the historical forecast results, and only makes simple adjustments when the short-term data deviates significantly from the forecast results. This method ignores the dynamic nature of the credibility of the forecast results, that is, there is a lack of scientific analysis of the deviation patterns and regularities that may exist between historical data and short-term data. Especially in scenarios where hydrological conditions change rapidly, key abnormal information in short-term data may not be incorporated into decision-making in a timely manner, resulting in a lack of accurate basis for the degree of trust in historical forecast results, thereby affecting the adaptability and accuracy of the power allocation plan.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an empirical power allocation method for a hydropower station in an electric power market environment. By constructing a multidimensional feature mapping system, the long-term laws of historical data and the dynamic changes of short-term real-time data are integrated in a unified feature space, eliminating the scale differences between data sources, and laying the foundation for subsequent credibility evaluation. Based on the dynamic calculation and interval division of the credibility index, the reliability level of the prediction results in different time periods is accurately identified, and the prediction results of the medium and low credibility intervals are dynamically adjusted through the partition correction strategy, ensuring the effective integration of short-term abnormal signals, and significantly improving the responsiveness of the prediction results to the real-time environment. In the optimization stage, a nonlinear multi-objective optimization model is constructed, and the benefits, risks and ecological balance goals are included in a unified framework. A set of power allocation schemes that comprehensively cover market demand, resource constraints and ecological requirements are generated through the high-order construction and optimization solution of the mathematical model, and finally the coordinated optimization of power generation efficiency, resource utilization and ecological protection is achieved. The present invention has a strict structure and clear logic, and has significant practical value and theoretical depth in dealing with complex market fluctuations and multiple constraints to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for allocating empirical electricity of a hydropower station in an electricity market environment comprises the following steps:

[0008] S1, decomposes the features of historical data and short-term real-time data, and constructs a unified multi-dimensional feature mapping matrix.

[0009] S2, calculates the credibility index of historical prediction results based on the feature mapping matrix and divides them into high credibility, medium credibility and low credibility intervals.

[0010] S3, performs short-term anomaly correction on the prediction results based on the credible interval and generates dynamically adjusted corrected results.

[0011] S4, input the correction results into the multi-objective optimization model to generate an electricity allocation plan that meets the constraints of benefits, risks and ecological balance.

[0012] In a preferred embodiment, step S1 includes the following contents:

[0013] S1.1, firstly, the historical data and short-term real-time data are decomposed separately, and the historical data D is decomposed into h Decomposed into the following three levels: D h =T h +S h +A h ; Among them, T h represents the long-term trend term; S hA is a seasonal regular term; h It represents the abnormal deviation term.

[0014] For short-term real-time data s , also using time series decomposition, decompose it into trend terms T s , Fluctuation term W s and abnormal signal A s :D s =T s +W s +A s .

[0015] S1.2, the decomposed data at each level are mapped in a unified feature space to construct a multi-dimensional feature mapping matrix, and the construction process includes the following steps:

[0016] a1. Feature standardization and dimensionality reduction: Adaptive principal component analysis is used to analyze T h ,S h ,A h and T s ,W s ,A s Standardization is performed to eliminate the influence of different feature dimensions and extract the main feature components.

[0017] a2. Feature fusion and mapping: Using the multi-scale mutual information fusion algorithm, the correlation weights of each feature level are determined by calculating the mutual information between historical features and short-term features; specifically, for each pair of corresponding level features and Calculating mutual information And build the weight matrix W based on this: Where i represents the i-th level of historical data, and j represents the j-th level of short-term data.

[0018] a3. Feature mapping matrix formation: Use the weight matrix W to perform weighted combination of historical features and short-term features to generate a unified multi-dimensional feature mapping matrix M: in, Represents the concatenation operation of features.

[0019] S1.3, use the dynamic scale normalization algorithm to unify the scales of historical data and short-term data in the mapping matrix.

[0020] S1.4, after the feature mapping matrix is ​​constructed, the historical prediction results are preliminarily matched with the short-term real-time data through a multi-dimensional feature matching algorithm to identify potential deviation patterns. The specific methods include:

[0021] Feature correlation analysis: Calculate the correlation coefficient matrix R between the features in the mapping matrix to identify strong correlations between historical and short-term features: in and Represents the first and Features.

[0022] Deviation pattern recognition: Based on the correlation analysis results, the self-organizing map network is used to cluster the feature mapping matrix to identify the main deviation patterns and their corresponding feature combinations.

[0023] Preliminary deviation quantification: For each deviation pattern, the deviation degree is quantified by a nonlinear deviation function φ: in, Indicates The deviation of the feature, is the eigenvalue in the mapping matrix.

[0024] In a preferred embodiment, step S2 includes the following contents:

[0025] S2.1, evaluate the credibility of historical prediction results, use the dynamic correlation adjustment matrix, which is calculated based on the mutual information between features and the dynamic weight distribution. The specific steps are as follows:

[0026] Mutual information calculation: For each pair of features in the mapping matrix Calculate their mutual information Reflect the dependencies between features: in, It is a feature and The joint probability density function of and are their marginal probability density functions respectively.

[0027] Dynamic weight distribution: Based on the changing trend of time series and the impact of emergencies, a time-related weight function ω(t) is constructed to reflect the dynamic adjustment of feature correlation in different time periods: ω(t) = e -λt sin(θt)+1; where λ and θ are adjustment parameters used to control the decay rate and oscillation frequency of the weight function.

[0028] Correlation adjustment: Combine mutual information with dynamic weights to obtain the adjusted correlation matrix C:

[0029] S2.2, using the adjusted correlation matrix and preliminary deviation quantification value, construct a credibility index calculation model:

[0030] Nonlinear transformation: Apply a hyperbolic tangent nonlinear function to the deviation quantization value. Among them, k is a tuning parameter.

[0031] Multi-dimensional integration: Perform matrix multiplication on the deviation quantization value after nonlinear transformation and the correlation matrix to generate a comprehensive deviation vector Θ: Θ = C·φ(Δ).

[0032] Exponential fusion: A multi-layer exponential fusion method is used to perform multiple exponential transformations on the comprehensive deviation vector. in, is the fusion strength parameter, represents the element-wise product, and ⊙ represents the element-wise product.

[0033] Credibility index generation: Finally, the credibility index κ of each prediction period is calculated through the normalized multi-layer index fusion results: in, For the The fusion results of each time period ensure that the sum of the credibility index is 1.

[0034] S2.3, based on the calculated credibility index, the prediction results are divided into three intervals: high credibility, medium credibility and low credibility. The division process is as follows:

[0035] Cluster analysis: The fuzzy C-means clustering algorithm is used to cluster the credibility index and determine three cluster centers, corresponding to the high credibility, medium credibility and low credibility intervals respectively:

[0036] Interval definition: Based on the clustering results, the following credible intervals are defined:

[0037] A high confidence interval is defined as a confidence index equal to or higher than the highest cluster center V 高 The range indicates that the prediction results are highly reliable and suitable for direct reference; the medium credible interval is defined as the confidence index in the middle cluster center V 中 and the highest cluster center V 高 The range between , indicating that the prediction results have a certain reference value, but need to be evaluated in combination with short-term corrections; the low credible interval corresponds to a credibility index lower than the medium cluster center V 低 This indicates that the forecast results are less reliable and need to be significantly adjusted to adapt to short-term dynamic changes.

[0038] In a preferred embodiment, step S3 includes the following contents:

[0039] S3.1, abnormal signal extraction: using the abnormal signal and preliminary deviation quantization value output from step S1, combined with the time series pattern in the short-term data, significant abnormal signals are identified through the time series segmentation model.

[0040] Signal classification: The anomaly classification method based on the mixed Gaussian model is applied to classify abnormal signals into strong fluctuation anomalies and gradual change anomalies.

[0041] S3.2, according to the credible interval division result of step S2, the high credible, medium credible and low credible intervals are respectively modified.

[0042] S3.3, to ensure that the correction results can dynamically adapt to real-time data changes, use the time series backtracking mechanism:

[0043] Backtracking window: Set a time window of fixed length τ, perform a backtracking check on the correction results of the previous τ time steps, and calculate the error change rate after correction; if the change rate exceeds the set threshold, adjust the correction function parameters and re-perform the correction calculation.

[0044] S3.4, Correction results: After the correction is completed, a comprehensive correction result P containing high-confidence, medium-confidence and low-confidence interval correction strategies is generated. 修正 , covering the final predicted value for all time steps.

[0045] In a preferred embodiment, step S4 includes the following contents:

[0046] S4.1, the multi-objective optimization model should include the following three core objectives:

[0047] Economic benefit target J1: The main target of the hydropower station is to obtain the maximum economic benefit in the market through electricity distribution. The formula is: Among them, P 分配 (t) is the power generation allocated in the tth period, p 市场 (t) is the market electricity price for the corresponding period.

[0048] Risk control objective J2: Risk comes from resource waste or market price fluctuations caused by forecast errors and short-term abnormal signals. To control risks, the following objective function is defined: Here, ∈ is a small value that prevents the denominator from being zero, and the risk function quantifies the deviation of the allocation from the corrected prediction value by means of squared error.

[0049] Ecological balance objective J3: The operation of the hydropower station needs to meet the downstream ecological needs. The impact of power generation on ecological flow is evaluated through the following objective function: Among them, Q 生态 is the minimum flow required for ecological protection, Q 实际 (t) is the actual flow at time period t, and the goal is to minimize the ecological deviation.

[0050] S4.2, the following constraints need to be met during the optimization process:

[0051] Resource constraint: The allocated power generation cannot exceed the total amount of water resources currently available: Among them, W 可用 The total amount of water resources.

[0052] Equipment operation constraints: The operating power of the generator set should be within the technical limit: P 最小 ≤P 分配 (t)≤P 最大 , Among them, P 最小 and P 最大 The lower and upper limits of the unit power.

[0053] Market balance constraint: The amount of electricity allocated should meet the fluctuation of market demand: P 分配 (t)≈P 市场 (t),

[0054] S4.3, multi-objective optimization uses the Pareto optimization technique based on evolutionary algorithm to find the balance point between objective functions through iterative calculation.

[0055] S4.4, Pareto optimization: Use a multi-objective genetic algorithm to generate a set of Pareto optimal solutions that meet the constraints.

[0056] Fuzzy comprehensive decision-making: Perform fuzzy comprehensive evaluation on the Pareto optimal solution set and select the solution with the highest comprehensive score as the final allocation plan.

[0057] S4.5, output the optimal power allocation plan and clarify the power generation allocation in each time period.

[0058] The technical effects and advantages of the method for allocating power volume of a hydropower station under the power market environment of the present invention are as follows:

[0059] The present invention systematically integrates historical data and short-term real-time data, constructs a multi-dimensional feature mapping matrix and a dynamic credibility evaluation system, and realizes accurate quantitative analysis of historical prediction results. Based on this evaluation result, a classification and recognition method is used to accurately correct short-term abnormal signals to ensure that the prediction results can timely reflect actual dynamic changes. Furthermore, combined with a nonlinear multi-objective optimization model, the three core goals of maximizing economic benefits, minimizing risks and balancing ecological flows are organically integrated, and a power distribution plan with high economic benefits, low operating risks and ecological protection requirements is generated through a complex mathematical algorithm. In the data processing process, the method adopts advanced feature decomposition, dynamic correlation adjustment and fuzzy clustering analysis technology to improve the adaptability and accuracy of the prediction; in the optimization process, the global optimality and practicality of the plan are ensured through multi-level index fusion and evolutionary algorithms. On the whole, the present invention achieves high accuracy and flexibility in power distribution, significantly improves the operating efficiency and resource utilization efficiency of hydropower stations in a complex power market environment, and ensures the coordination and consistency of economic and ecological goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of a method for allocating empirical electricity volume of a hydropower station in an electricity market environment according to the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Embodiment 1: Figure 1 The present invention provides an empirical power distribution method for a hydropower station in a power market environment, comprising:

[0063] S1, decomposes the features of historical data and short-term real-time data, and constructs a unified multi-dimensional feature mapping matrix.

[0064] S2, calculates the credibility index of historical prediction results based on the feature mapping matrix and divides them into high credibility, medium credibility and low credibility intervals.

[0065] S3, performs short-term anomaly correction on the prediction results based on the credible interval and generates dynamically adjusted corrected results.

[0066] S4, input the correction results into the multi-objective optimization model to generate an electricity allocation plan that meets the constraints of benefits, risks and ecological balance.

[0067] The power generation of hydropower stations is affected by many factors, including long-term hydrological trends, seasonal precipitation changes, market electricity price fluctuations, and sudden hydrological events. Historical data provides long-term power generation laws and market fluctuation characteristics, while short-term real-time data reflects the current hydrological status and market dynamics. In order to optimize power distribution, these two types of data must be effectively fused to retain the stability of historical laws and flexibly respond to the dynamics of short-term changes. However, there are significant differences between historical data and short-term data in time scale, data characteristics, and change rate. Traditional linear fusion methods are difficult to fully capture the complex relationship between them. Therefore, it is particularly important to design a method that can hierarchically decompose and map these two types of data in multiple dimensions.

[0068] Step S1 includes the following contents:

[0069] S1.1, first, the historical data and short-term real-time data are characterized and decomposed respectively. The historical data contains many years of power generation records, hydrological data and market price information. The historical data D is decomposed into h Decomposed into the following three levels: D h =T h +S h +A h ; Among them, T h It represents the long-term trend item, reflecting the hydrological changes and market development trends accumulated over many years; S h A is a seasonal regularity term, capturing the periodic fluctuations of seasonal precipitation and power generation; h represents the abnormal deviation term, which is used to identify extreme hydrological events and abnormal market fluctuations.

[0070] For short-term real-time data s , also using time series decomposition, decompose it into trend terms T s , Fluctuation term W s and abnormal signal A s :D s =T s +W s +A s ; Among them, T s Reflecting the current traffic trend, W s Capture the real-time fluctuations of market prices, s It is used to identify sudden hydrological events or market anomalies.

[0071] S1.2, the decomposed data at each level needs to be mapped in a unified feature space for subsequent credibility evaluation. To this end, a multi-dimensional feature mapping matrix is ​​constructed, and its construction process includes the following steps:

[0072] a1. Feature standardization and dimensionality reduction: Adaptive principal component analysis is used to analyze T h ,Sh ,A h and T s ,W s ,A s Standardization is performed to eliminate the influence of different feature dimensions, and the main feature components are extracted through adaptive dimensionality reduction technology. The standardized historical feature F h and short-term characteristics F s Respectively expressed as: F h =AdaptivePCA(T h ,S h ,A h );F s =AdaptivePCA(T s ,W s ,A s );

[0073] a2. Feature fusion and mapping: Use the multi-scale mutual information fusion algorithm to calculate the mutual information between historical features and short-term features to determine the correlation weight of each feature level. Specifically, for each pair of corresponding level features and Calculating mutual information And build the weight matrix W based on this: Where i represents the i-th level of historical data, and j represents the j-th level of short-term data.

[0074] a3. Feature mapping matrix formation: Use the weight matrix W to perform weighted combination of historical features and short-term features to generate a unified multi-dimensional feature mapping matrix M: in, Represents the concatenation operation of features, ensuring that historical and short-term features are arranged in order in the same matrix.

[0075] S1.3, using a dynamic scale normalization algorithm to unify the scales of historical data and short-term data in the mapping matrix. This algorithm adaptively adjusts the scale factors α and β of each feature according to the current hydrological and market conditions to achieve dynamic normalization of the feature values:

[0076] Specifically, the scaling factors α and β are dynamically adjusted through real-time calculation of the standard deviation ratio of historical and short-term data and trend consistency indicators to adapt to different hydrological and market environments.

[0077] For example, here is how the scaling factors α and β are calculated:

[0078] The dynamic calculation of the scale factors α and β is based on the statistical characteristics of the two types of data, including standard deviation ratio and trend consistency. The specific formula is as follows:

[0079] σ h and σ s : They are the standard deviations of historical data and short-term data, respectively, used to measure the fluctuation range of the two types of data.

[0080] corr(T s ,T h ): It is the correlation coefficient between the historical data trend item and the short-term data trend item, reflecting the overall trend consistency of the two.

[0081] Δ t : is the time span difference between historical data and short-term data, defined as: in, and are the end time of historical data and the start time of short-term data respectively. γ: is the time decay factor, which controls the influence of time span on scale factor. exp(-γ·Δ t ): Used to attenuate data with a large time span to avoid excessive influence of historical data on short-term data.

[0082] S1.4, after the feature mapping matrix is ​​constructed, the historical prediction results are preliminarily matched with the short-term real-time data through the multi-dimensional feature matching algorithm to identify potential deviation patterns. The specific methods include:

[0083] Feature correlation analysis: Calculate the correlation coefficient matrix R between the features in the mapping matrix to identify strong correlations between historical and short-term features: in and Represents the first and Features.

[0084] Deviation pattern recognition: Based on the correlation analysis results, the self-organizing map network is used to cluster the feature mapping matrix to identify the main deviation patterns and their corresponding feature combinations.

[0085] Preliminary deviation quantification: For each deviation pattern, the deviation degree is quantified by a nonlinear deviation function φ: in, Indicates The deviation of the feature, To map the eigenvalues ​​in the matrix, the function τ is designed to enhance the sensitivity to abnormal deviations while suppressing the impact of small fluctuations.

[0086] The output of step S1 includes a multi-dimensional feature mapping matrix, deviation pattern recognition results, and preliminary deviation quantification values. These outputs serve as input variables for the credibility quantitative evaluation in step S2, ensuring that the subsequent steps can perform dynamic credibility calculation and decision optimization based on detailed feature analysis results.

[0087] Step S1 constructs a unified and high-dimensional data structure through multi-level feature decomposition, dynamic scale normalization, multi-dimensional feature mapping and deviation pattern recognition, and realizes the deep integration of historical data and short-term real-time data. This process not only retains the long-term regularity of historical data and the instant dynamics of short-term data, but also improves the accuracy and flexibility of data processing through complex nonlinear calculation formulas and adaptive algorithms, providing a solid technical foundation for subsequent credibility assessment and optimization of power distribution decisions.

[0088] In the power market environment, the power distribution strategy of hydropower stations not only depends on the long-term laws of historical data, but also needs to be combined with the dynamic changes of short-term real-time data. Step S1 has completed the multi-dimensional feature mapping of historical data and short-term data, and generated a unified feature mapping matrix, deviation pattern recognition results, and preliminary deviation quantification values. Next, the goal of step S2 is to systematically evaluate the credibility of historical prediction results based on these outputs, and divide them into three intervals of high credibility, medium credibility, and low credibility, to provide a scientific basis for subsequent decision optimization.

[0089] Step S2 includes the following contents:

[0090] S2.1, in order to comprehensively evaluate the credibility of historical prediction results, it is necessary to comprehensively consider the correlation between the features in the multi-dimensional feature mapping matrix and their impact on the prediction results. Use a dynamic correlation adjustment matrix, which is calculated based on the mutual information between features and the dynamic weight distribution. The specific steps are as follows:

[0091] Mutual information calculation: For each pair of features in the mapping matrix Calculate their mutual information Reflect the dependencies between features: in, It is a feature and The joint probability density function of and are their marginal probability density functions respectively.

[0092] Dynamic weight distribution: Based on the changing trend of time series and the impact of emergencies, a time-related weight function ω(t) is constructed, which reflects the dynamic adjustment of feature correlation in different time periods: ω(t) = e -λt sin(θt)+1; where λ and θ are adjustment parameters used to control the decay rate and oscillation frequency of the weight function.

[0093] Correlation adjustment: Combine mutual information with dynamic weights to obtain the adjusted correlation matrix C:

[0094] S2.2, using the adjusted correlation matrix and the preliminary deviation quantification value, construct a credibility index calculation model. This model achieves a quantitative assessment of the credibility of the prediction results through nonlinear transformation and multi-dimensional integration:

[0095] Nonlinear transformation: A hyperbolic tangent nonlinear function is applied to the deviation quantization value to enhance the sensitivity to large deviations while suppressing the noise effect of small deviations: Among them, k is an adjustment parameter that controls the steepness of the nonlinear function.

[0096] Multi-dimensional integration: Perform matrix multiplication on the deviation quantization value after nonlinear transformation and the correlation matrix to generate a comprehensive deviation vector Θ: Θ = C·Φ(Δ);

[0097] Exponential fusion: A multi-layer exponential fusion method is used to perform multiple exponential transformations on the comprehensive deviation vector to capture complex nonlinear relationships and high-order interaction effects: in, is the fusion strength parameter, represents the element-wise product, and ⊙ represents the element-wise product.

[0098] Credibility index generation: Finally, the credibility index κ of each prediction period is calculated through the normalized multi-layer index fusion results: in, For the The fusion results of each time period ensure that the sum of the credibility index is 1.

[0099] S2.3, based on the calculated credibility index, the prediction results are divided into three intervals: high credibility, medium credibility and low credibility, so as to facilitate the subsequent decision optimization process. The division process is as follows:

[0100] Cluster analysis: The fuzzy C-means clustering algorithm is used to cluster the credibility index and determine three cluster centers, corresponding to the high credibility, medium credibility and low credibility intervals respectively:

[0101] Interval definition: Based on the clustering results, the following credible intervals are defined:

[0102] High confidence interval: The cluster center corresponds to the highest confidence index, reflecting that the prediction result is highly reliable.

[0103] Medium Credible Interval: The cluster center is in a medium position, and the prediction result has a certain degree of credibility, but it needs to be treated with caution.

[0104] Low credible interval: The cluster center corresponds to the lowest credibility index, the reliability of the prediction result is low, and significant adjustments are required.

[0105] Based on the location of the cluster center, set a specific threshold for the credible interval. For example:

[0106] A high confidence interval is defined as a confidence index equal to or higher than the highest cluster center V 高 The range indicates that the prediction results are highly reliable and suitable for direct reference; the medium credible interval is defined as the confidence index in the middle cluster center V 中 and the highest cluster center V 高 The range between , indicating that the prediction results have a certain reference value, but need to be evaluated in combination with short-term corrections; the low credible interval corresponds to a credibility index lower than the medium cluster center V 低 This indicates that the forecast results are less reliable and need to be significantly adjusted to adapt to short-term dynamic changes.

[0107] Step S2 outputs the credibility index vector and the credibility interval division result.

[0108] The processing of step S2 is based on the multidimensional feature mapping matrix. Through the construction of the dynamic correlation matrix and the application of the nonlinear deviation function, the reliability of the historical prediction results is quantified as a credibility index from the perspective of multidimensional features. This credibility index combines the deviation law and dynamic change characteristics of historical data and short-term data, and can fully reflect the difference and consistency between the two. Subsequently, combined with the fuzzy clustering method, the credibility index is partitioned, which not only realizes the scientific division of the prediction results into high credibility, medium credibility and low credibility intervals, but also ensures that the distribution within each interval has clear boundaries and distinction criteria. The final output of the credibility index and interval division results provides a clear numerical basis for the subsequent short-term anomaly correction, which not only implements the results of the feature mapping, but also lays a data foundation for the dynamic adjustment of the optimization strategy.

[0109] The forecast results of electricity distribution need to reflect both historical laws and short-term dynamic changes. However, in actual scenarios, short-term abnormal hydrological conditions or market fluctuations may cause the forecast results to deviate from the actual situation. Step S2 has divided the forecast results into high-credibility, medium-credibility and low-credibility intervals through the calculation and partitioning of the credibility index, and clarified the reliability level of the forecast results in different intervals. The goal of step S3 is to reasonably adjust the forecast results of the low-credibility and medium-credibility intervals through dynamic correction methods for these intervals, so that they can fully integrate short-term abnormal signals, thereby generating revised forecast results and improving their accuracy and applicability.

[0110] Step S3 includes the following contents:

[0111] S3.1, abnormal signal extraction: using the abnormal signal A output in step S1 s The preliminary deviation quantification value Δ is combined with the temporal patterns in short-term data to identify significant abnormal signals through the time series segmentation model.

[0112] Signal classification: An anomaly classification method based on a mixed Gaussian model is applied to classify abnormal signals into strong fluctuation anomalies (such as flood events) and gradual change anomalies (such as continuous rainfall or market trend changes). The classification function is expressed as: Among them, μ k and are the mean and variance of the kth class, A s Indicates the abnormal signal characteristic value.

[0113] S3.2, according to the credible interval division result of step S2, respectively revise the high credible interval, the medium credible interval and the low credible interval;

[0114] High confidence interval: The prediction results in this interval are highly reliable and do not need to be corrected.

[0115] Medium credible interval: For the medium credible interval, the correction formula is: Among them, P 修正 is the original predicted value, is the correction amplitude adjustment factor, f(A s ,Δ) is a dynamic correction function based on abnormal signals and deviation quantization values, defined as: This function is highly sensitive to abnormal deviation terms and suppresses noise through nonlinear regulation of the deviation value.

[0116] Low credible interval: For the low credible interval, the original forecast value is completely replaced and the forecast result is directly recalculated based on short-term data. It is a reconstruction function that integrates short-term trends, fluctuation characteristics and abnormal signals, expressed as:

[0117] S3.3, to ensure that the correction results can dynamically adapt to real-time data changes, use the time series backtracking mechanism:

[0118] Backtracking window: Set a fixed-length time window τ, perform a backtracking check on the correction results of the first τ time steps, and calculate the error change rate after correction:

[0119] If the rate of change exceeds the set threshold, the correction function parameters are adjusted and the correction calculation is performed again.

[0120] The model is updated and corrected in combination with real-time data to ensure that the impact of abnormal signals on the prediction results can be dynamically reflected in the correction process.

[0121] S3.4, Correction results: After the correction is completed, a comprehensive correction result P containing high-confidence, medium-confidence and low-confidence interval correction strategies is generated. 修正 , covering the final predicted value for all time steps.

[0122] Corrected prediction result P修正 and correction process data to provide accurate input for the optimization model in step S4.

[0123] Step S3 scientifically integrates the impact of short-term abnormal signals into the prediction results by partitioning the credible interval and designing a dynamic correction strategy. The high credible interval retains the stability of the prediction results, the medium credible interval makes partial adjustments to the abnormal signals, and the low credible interval directly reconstructs the prediction value based on short-term data. The entire correction process makes full use of the classification characteristics of abnormal signals and the nonlinear regulation of deviation quantization values, combines the time series backtracking mechanism to achieve real-time optimization, and finally outputs a fully corrected prediction result, providing high-quality input data for subsequent multi-objective optimization, ensuring that the correction process is rigorous and highly applicable.

[0124] The optimization of the power distribution plan of the hydropower station needs to find a dynamic balance between maximizing economic benefits, minimizing risks and ecological balance constraints. Step S3 has generated the corrected prediction result correction P 修正 ,These results reflect the combined impact of historical laws and short-term dynamic changes. However, the power allocation plan cannot be directly formulated based on the correction results alone. Further multi-objective optimization must be carried out to make the allocation strategy achieve the global optimality between market demand fluctuations, resource utilization efficiency and ecological protection. Step S4 aims to build a multi-objective optimization model to generate the final power allocation plan based on the correction results.

[0125] Step S4 includes the following contents:

[0126] S4.1, the multi-objective optimization model should include the following three core objectives:

[0127] Economic benefit target J1: The main target of the hydropower station is to obtain the maximum economic benefit in the market through electricity distribution. The formula is: Among them, P 分配 (t) is the power generation allocated in the tth period, p 市场 (t) is the market electricity price for the corresponding period.

[0128] Risk control objective J2: Risks come from forecast errors and short-term abnormal signals that may lead to resource waste or market price fluctuations. To control risks, the following objective function is defined: Here, ∈ is a small value that prevents the denominator from being zero, and the risk function quantifies the deviation of the allocation from the corrected prediction value by means of squared error.

[0129] Ecological balance objective J3: The operation of the hydropower station needs to meet the downstream ecological needs. The impact of power generation on ecological flow is evaluated through the following objective function: Among them, Q 生态 is the minimum flow required for ecological protection, Q 实际(t) is the actual flow at time period t, and the goal is to minimize the ecological deviation.

[0130] S4.2, the following constraints need to be met during the optimization process:

[0131] Resource constraint: The allocated power generation cannot exceed the total amount of water resources currently available: Among them, W 可用 The total amount of water resources.

[0132] Equipment operation constraints: The operating power of the generator set should be within the technical limit: P 最小 ≤P 分配 (t)≤P 最大 , Among them, P 最小 and P 最大 The lower and upper limits of the unit power.

[0133] Market balance constraint: The amount of electricity allocated should meet the fluctuation of market demand as much as possible: P 分配 (t)≈P 市场 (t),

[0134] S4.3, the multi-objective optimization problem can finally be expressed as the following nonlinear model: min{-J1,J2,J3};

[0135] In order to solve the nonlinear model, the Pareto optimization technology based on evolutionary algorithm is used to find the balance point between objective functions through iterative calculation.

[0136] S4.4, Pareto optimization: A multi-objective genetic algorithm (NSGA-II) is used to generate a set of Pareto optimal solutions that meet the constraints.

[0137] Fuzzy comprehensive decision-making: Perform fuzzy comprehensive evaluation on the Pareto optimal solution set and select the solution with the highest comprehensive score as the final allocation plan. The calculation formula for the comprehensive score S is: Among them, ξ is a smoothing factor, which is used to control the balance between objectives.

[0138] S4.5, output the optimal power allocation plan and clarify the power generation allocation in each time period.

[0139] Step S4 takes the multi-objective optimization model as the core, converts the corrected prediction results into a specific electricity allocation plan, and realizes the systematic planning of electricity allocation by constructing three objective functions of benefit, risk and ecological balance. In the optimization process, a nonlinear construction method is adopted, and a Pareto optimal solution set is generated by a genetic algorithm to ensure the balance between the various objectives. The risk control objective improves the accuracy of power generation resource utilization by constraining the dynamic deviation between the allocation amount and the corrected prediction value; the ecological balance objective makes the allocation plan meet the needs of ecological protection by accurately quantifying the flow deviation; the economic benefit objective ensures the economy of the plan by coupling optimization of market electricity prices and power generation. The final electricity allocation plan not only meets multiple constraints such as water resources, equipment and market demand, but also provides a set of highly adaptable allocation strategies that adapt to the dynamic market environment, which fully reflects the rigor of the theoretical model and the feasibility of practical application.

[0140] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0141] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0142] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0143] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for allocating empirical electricity of a hydropower station in a power market environment, characterized in that: Includes steps: S1, decomposes the features of historical data and short-term real-time data, and constructs a unified multi-dimensional feature mapping matrix; S2, calculates the credibility index of historical prediction results based on the feature mapping matrix and divides them into high credibility, medium credibility and low credibility intervals; S3, performs short-term anomaly correction on the forecast results according to the credible interval and generates dynamically adjusted revised results; S4, input the correction results into the multi-objective optimization model to generate an electricity allocation plan that meets the constraints of benefits, risks and ecological balance.

2. The method for allocating power volume of a hydropower station under a power market environment according to claim 1 is characterized by: Step S1 includes the following contents: S1.1, firstly, the historical data and short-term real-time data are decomposed separately, and the historical data D is decomposed into h Decomposed into the following three levels: D h =T h +S h +A h ; Among them, T h represents the long-term trend term; S h A is a seasonal regular term; h It represents the abnormal deviation term; For short-term real-time data s , also using time series decomposition, decompose it into trend terms T s , Fluctuation term W s and abnormal signal A s :D s =T s +W s +A s ; S1.2, the decomposed data at each level are mapped in a unified feature space to construct a multi-dimensional feature mapping matrix, and the construction process includes the following steps: a1. Feature standardization and dimensionality reduction: Adaptive principal component analysis is used to analyze T h ,S h ,A h and T s ,W s ,A s Perform standardization to eliminate the influence of different feature dimensions and extract the main feature components; a2. Feature fusion and mapping: Using the multi-scale mutual information fusion algorithm, the correlation weights of each feature level are determined by calculating the mutual information between historical features and short-term features; specifically, for each pair of corresponding level features and Calculating mutual information And build the weight matrix W based on this: Where i represents the i-th level of historical data, and j represents the j-th level of short-term data; a3. Feature mapping matrix formation: Use the weight matrix W to perform weighted combination of historical features and short-term features to generate a unified multi-dimensional feature mapping matrix M: M = W · [F h ⊕F s ]; where ⊕ represents the concatenation operation of features.

3. The method for allocating power volume of a hydropower station under a power market environment according to claim 2 is characterized by: S1.3, using a dynamic scale normalization algorithm to unify the scales of historical data and short-term data in the mapping matrix; S1.4, after the feature mapping matrix is ​​constructed, the historical prediction results are preliminarily matched with the short-term real-time data through a multi-dimensional feature matching algorithm to identify potential deviation patterns. The specific methods include: Feature correlation analysis: Calculate the correlation coefficient matrix R between each feature in the mapping matrix to identify strong correlations between historical and short-term features: in and Represents the first and Features Deviation pattern recognition: Based on the correlation analysis results, the self-organizing map network is used to cluster the feature map matrix to identify the main deviation patterns and their corresponding feature combinations; Preliminary deviation quantification: For each deviation pattern, the deviation degree is quantified by a nonlinear deviation function φ: in, Indicates The deviation of the feature, is the eigenvalue in the mapping matrix.

4. The method for allocating power volume of a hydropower station under a power market environment according to claim 3 is characterized by: Step S2 includes the following contents: S2.1, evaluate the credibility of historical prediction results, use the dynamic correlation adjustment matrix, which is calculated based on the mutual information between features and the dynamic weight distribution. The specific steps are as follows: Mutual information calculation: For each pair of features in the mapping matrix Calculate their mutual information Reflect the dependencies between features: in, It is a feature and The joint probability density function of and are their marginal probability density functions respectively; Dynamic weight distribution: Based on the changing trend of time series and the impact of emergencies, a time-related weight function ω(t) is constructed to reflect the dynamic adjustment of feature correlation in different time periods: ω(t) = e -λt sin(θt)+1; where λ and θ are adjustment parameters used to control the decay rate and oscillation frequency of the weight function; Correlation adjustment: Combine mutual information with dynamic weights to obtain the adjusted correlation matrix C: S2.2, using the adjusted correlation matrix and preliminary deviation quantification value, construct a credibility index calculation model: Nonlinear transformation: Apply a hyperbolic tangent nonlinear function to the deviation quantization value. Among them, k is the adjustment parameter; Multi-dimensional integration: Perform matrix multiplication on the deviation quantization value after nonlinear transformation and the correlation matrix to generate a comprehensive deviation vector Θ: Θ = C·Φ(Δ); Exponential fusion: A multi-layer exponential fusion method is used to perform multiple exponential transformations on the comprehensive deviation vector. in, is the fusion strength parameter, represents the element-level product, ⊙ represents the element-wise product; Credibility index generation: Finally, the credibility index κ of each prediction period is calculated through the normalized multi-layer index fusion results: in, For the The fusion results of each time period ensure that the sum of the credibility index is 1.

5. The method for allocating power volume of a hydropower station under a power market environment according to claim 4 is characterized by: S2.3, based on the calculated credibility index, the prediction results are divided into three intervals: high credibility, medium credibility and low credibility. The division process is as follows: Cluster analysis: The fuzzy C-means clustering algorithm is used to cluster the credibility index and determine three cluster centers, corresponding to the high credibility, medium credibility and low credibility intervals respectively: Interval definition: Based on the clustering results, the following credible intervals are defined: A high confidence interval is defined as a confidence index equal to or higher than the highest cluster center V 高 The range indicates that the prediction results are highly reliable and suitable for direct reference; the medium credible interval is defined as the confidence index in the middle cluster center V 中 and the highest cluster center V 高 The range between , indicating that the prediction results have a certain reference value, but need to be evaluated in combination with short-term corrections; the low credible interval corresponds to a credibility index lower than the medium cluster center V 低 This indicates that the forecast results are less reliable and need to be significantly adjusted to adapt to short-term dynamic changes.

6. The method for allocating power volume of a hydropower station under a power market environment according to claim 5 is characterized by: Step S3 includes the following contents: S3.1, abnormal signal extraction: using the abnormal signal output from step S1 and the preliminary deviation quantization value, combined with the time series pattern in the short-term data, the significant abnormal signal is identified through the time series segmentation model; Signal classification: An anomaly classification method based on a mixed Gaussian model is used to classify abnormal signals into strong fluctuation anomalies and gradual change anomalies; S3.2, according to the credible interval division result of step S2, respectively revise the high credible interval, the medium credible interval and the low credible interval; S3.3, to ensure that the correction results can dynamically adapt to real-time data changes, use the time series backtracking mechanism: Backtracking window: Set a fixed-length time window τ, perform a backtracking check on the correction results of the previous τ time steps, and calculate the error change rate after correction; If the rate of change exceeds the set threshold, the correction function parameters are adjusted and the correction calculation is performed again; S3.4, Correction results: After the correction is completed, a comprehensive correction result P containing high-confidence, medium-confidence and low-confidence interval correction strategies is generated. 修正 , covering the final predicted value for all time steps.

7. The method for allocating power volume of a hydropower station under a power market environment according to claim 6 is characterized by: Step S4 includes the following contents: S4.1, the multi-objective optimization model should include the following three core objectives: Economic benefit target J1: The main target of the hydropower station is to obtain the maximum economic benefit in the market through electricity distribution. The formula is: Among them, P 分配 (t) is the power generation allocated in the tth period, p 市场 (t) is the market electricity price for the corresponding period; Risk control objective J2: Risk comes from resource waste or market price fluctuations caused by forecast errors and short-term abnormal signals. To control risks, the following objective function is defined: Among them, ∈ is a small value that prevents the denominator from being zero, and the risk function quantifies the deviation of the allocation amount from the corrected predicted value by means of squared error; Ecological balance objective J3: The operation of the hydropower station needs to meet the downstream ecological needs. The impact of power generation on ecological flow is evaluated through the following objective function: Among them, Q 生态 is the minimum flow required for ecological protection, Q 实际 (t) is the actual flow at time period t, and the goal is to minimize the ecological deviation.

8. The method for allocating power volume of a hydropower station under a power market environment according to claim 7 is characterized by: S4.2, the following constraints need to be met during the optimization process: Resource constraint: The allocated power generation cannot exceed the total amount of water resources currently available: Among them, W 可用 is the total amount of water resources; Equipment operation constraints: The operating power of the generator set should be within the technical limit: P 最小 ≤P 分配 (t)≤P 最大 , Among them, P 最小 and P 最大 is the lower and upper limits of the unit power; Market balance constraint: The amount of electricity allocated should meet the fluctuation of market demand: P 分配 (t)≈P 市场 (t), S4.3, multi-objective optimization uses Pareto optimization technology based on evolutionary algorithm to find the balance point between objective functions through iterative calculation; S4.4, Pareto optimization: Use a multi-objective genetic algorithm to generate a set of Pareto optimal solutions that meet the constraints; Fuzzy comprehensive decision-making: Perform fuzzy comprehensive evaluation on the Pareto optimal solution set and select the solution with the highest comprehensive score as the final allocation plan; S4.5, output the optimal power allocation plan and clarify the power generation allocation in each time period.

Images