Thermal index correction method and system based on accumulated temperature effect

Abstract

The invention discloses a thermal index correction method and system based on accumulated temperature effect, and the method comprises the steps: obtaining the historical data of a target region, and the data comprise power load, temperature and humidity, solar radiation intensity and non-meteorological characteristics; coding the non-meteorological characteristics and clustering in combination with the load data, and dividing a plurality of user behavior portraits; calculating a basic intense heat index, and constructing a dual correction model containing an air heat accumulated temperature effect and a building radiation accumulated temperature effect; for each behavior portrait, cooperatively solving a differentiated accumulated temperature effect parameter matrix by using a genetic algorithm; and finally, calling corresponding parameters according to the portrait to which the target day belongs, calculating the corrected exactly hot index and reversely deducing the equivalent highest temperature. According to the method, by introducing the solar radiation correction term and the behavior portrait classification mechanism, double accurate correction of air heat lag and building radiation lag is realized, the problems of single parameter and dimension deficiency of a traditional model are effectively solved, and the accuracy of summer power load prediction is improved.

Description

Technical Field

[0001] This invention relates to a meteorological technology, and more specifically to a method and system for correcting the heat index based on the accumulated temperature effect. Background Technology

[0002] In recent years, with the development of the social economy and the improvement of residents' living standards, residential electricity demand has continued to grow. Coupled with the high temperatures in summer, cooling loads such as air conditioning have increased significantly, leading to a rapid rise in grid load. Short-term load forecasting, as a crucial aspect of power system dispatching and operation, directly impacts the safe and stable operation of the power grid and the rational allocation of power resources.

[0003] Existing short-term load forecasting methods are generally based on meteorological factors, historical load data, and time series characteristics, with temperature being the core factor affecting residential load in summer. However, during periods of sustained high temperatures, residential electricity load is not only affected by the highest temperature of the day but also by the "cumulative temperature effect." This means that during periods of continuous high temperatures, environmental heat gradually accumulates, causing cooling demand to lag and accumulate, resulting in actual load exceeding the theoretical load determined solely by the daily temperature.

[0004] To address the accumulated temperature effect, existing technologies mostly use the heat index to comprehensively reflect the thermal impact of temperature and humidity on human perception. However, existing correction methods still have the following significant limitations when applied to refined load forecasting: First, traditional heat indices are primarily calculated based on the temperature and humidity of air in shaded areas, often neglecting the continuous heating effect of solar radiation on building envelopes (walls and roofs). In fact, the release of heat stored in buildings due to direct sunlight has a stronger lag and is a significant factor contributing to persistently high air conditioning loads; ignoring radiation factors leads to insufficient correction.

[0005] Secondly, existing correction models typically use a uniform accumulated temperature parameter for all dates. In reality, residents' work and rest patterns and air conditioning usage habits vary significantly between weekdays, weekends, holidays, and special days affected by electricity pricing policies. This results in different load sensitivities and lag characteristics to the accumulated temperature effect under different behavioral patterns. Using a single model cannot accurately match the complex and ever-changing electricity consumption patterns, leading to a significant drop in prediction accuracy in specific scenarios.

[0006] Therefore, there is an urgent need to propose a heat index correction method that integrates the dual accumulated temperature effects of air heat and building radiant heat, and can make parameter differential corrections based on user behavior profiles, so as to improve the accuracy of short-term residential load forecasting in summer. Summary of the Invention

[0007] The main technical problem solved by this invention is: 1. Existing heat index models typically only consider air temperature and humidity, neglecting the cumulative temperature effect of solar radiation on building envelopes. This results in insufficient correction for lagging loads during periods of continuous high temperatures, leading to significant prediction bias.

[0008] 2. Existing accumulated temperature correction methods typically use uniform correction parameters, which cannot adapt to the differentiated load response characteristics under different user behavior profiles such as weekdays and holidays, resulting in poor applicability of the model in specific scenarios.

[0009] 3. Key parameters in existing models, such as accumulated temperature threshold, backtracking days, and decay coefficient, are mostly set based on experience and lack a data-driven global optimization mechanism, resulting in insufficient generalization ability of the model and difficulty in matching real load characteristics.

[0010] Therefore, this invention discloses a heat index correction method and system based on accumulated temperature effect. The method involves acquiring historical data of the target area, including electricity load, temperature and humidity, solar radiation intensity, and non-meteorological characteristics. The non-meteorological characteristics are encoded and clustered with load data to create several user behavior profiles. A basic heat index is calculated, and a dual correction model incorporating both air thermal accumulated temperature effect and building radiation accumulated temperature effect is constructed. For each behavior profile, a genetic algorithm is used to collaboratively solve for the differentiated accumulated temperature effect parameter matrix. Finally, based on the profile of the target day, the corresponding parameters are called to calculate the corrected heat index and inversely deduce the equivalent maximum temperature. This invention, by introducing a solar radiation correction term and a behavior profile classification mechanism, achieves accurate dual correction for air thermal lag and building radiation lag, effectively solving the problems of single parameters and missing dimensions in traditional models, and improving the accuracy of summer electricity load forecasting.

[0011] The present invention adopts the following technical solution.

[0012] A method for correcting the heat index based on the accumulated temperature effect, comprising: S1: Obtain historical data of the target area within a preset time period, including residential electricity load data, meteorological characteristic data, and non-meteorological characteristic data; S2: Construct a behavioral feature vector based on the residential electricity load data; use a clustering algorithm based on the behavioral feature vector to divide each day within the preset time period into several behavioral profiles, and establish a correspondence between the non-meteorological feature data and the behavioral profiles; S3: Based on the meteorological characteristic data, establish a heat index calculation model to calculate the maximum daily heat index within the preset time period; S4: Construct a dual correction model that includes air heat accumulation effect and building radiation accumulation effect; for each behavior profile obtained in S2, use a subset of historical data belonging to that behavior profile to maximize the correlation between the corrected daily maximum heat index sequence and the residential electricity load data sequence in the historical data subset as the optimization objective, solve the parameters of the dual correction model, and generate a behavior profile-accumulation effect parameter matrix containing differentiated parameter groups under different profiles; S5: For the target day to be corrected, identify its behavioral profile based on its non-meteorological feature data, call the corresponding differentiated parameter group from the behavioral profile-accumulated temperature effect parameter matrix, and substitute it into the dual correction model to calculate the corrected daily maximum heat index.

[0013] Preferably, in S1, the historical data includes: The residential electricity load data includes daily maximum electricity load data; The meteorological data includes daily maximum temperature, daily maximum humidity, and daily solar radiation intensity. The non-meteorological feature data includes date type labels, electricity price type labels, and electricity demand-side management event information.

[0014] Preferably, in S2, constructing a behavioral feature vector based on the residential electricity load data includes: The date type label and the electricity price type label are converted using unique thermal encoding, and the electricity demand-side management event information is converted into binary values ​​to obtain non-meteorological numerical features; The non-meteorological numerical features and the residential electricity load data are normalized, and the normalized data are combined to construct the behavioral feature vector.

[0015] Preferably, in S2, based on the behavioral feature vector, a clustering algorithm is used to divide each day within the preset time period into several behavioral profiles, including: The behavior feature vectors are clustered using a clustering algorithm. The number of clusters is determined according to a preset clustering evaluation index, forming several behavior pattern clusters. The behavior pattern cluster to which each day belongs within the preset time period is also determined. The behavior pattern clusters are defined as corresponding behavior profiles, and the distribution patterns of non-meteorological feature data corresponding to each behavior profile are statistically analyzed to construct the correspondence between non-meteorological feature data and behavior profiles.

[0016] Preferably, in S4, the method for constructing the dual correction model includes: Calculate the air heat accumulated temperature correction and the building radiation accumulated temperature correction separately, and then add the air heat accumulated temperature correction and the building radiation accumulated temperature correction to the daily maximum heat index to obtain the corrected daily maximum heat index. The calculation of the air heat accumulated temperature correction includes: within the effective air heat backtracking days, determining the corresponding discrete air heat accumulated temperature effect coefficient based on the numerical range of the daily maximum heat index and the number of lag days; calculating the difference between the daily maximum heat index and the boundary air heat index; multiplying the difference by the discrete air heat accumulated temperature effect coefficient and summing them to obtain the air heat accumulated temperature correction. The calculation of the building radiation accumulated temperature correction includes: within the effective building radiation backtracking days, determining the corresponding discretized building radiation accumulated temperature effect coefficient based on the numerical range of the daily solar radiation intensity and the number of lag days; calculating the difference between the daily solar radiation intensity and the limit building radiation index; multiplying the difference by the discretized building radiation accumulated temperature effect coefficient and summing them to obtain the building radiation accumulated temperature correction. The determination of the effective air heat backtracking days includes: taking the smaller value between the theoretical maximum air heat backtracking days and the first consecutive days, where the first consecutive days refers to the number of days prior to day i where the daily maximum heat index is greater than the threshold air heat index; the determination of the effective building radiation backtracking days includes: taking the smaller value between the theoretical maximum building radiation backtracking days and the second consecutive days, where the second consecutive days refers to the number of days prior to day i where the daily solar radiation intensity is greater than the threshold building radiation index.

[0017] In S4, the parameters of the dual-correction model are solved using the corresponding subset of historical data, specifically including: A constrained optimization model is constructed, the objective function of which is to maximize the Pearson correlation coefficient between the corrected daily maximum heat index sequence calculated by the dual correction model and the residential electricity load data sequence in the historical data subset. The constraints of the constrained optimization model include: the discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient are both within a preset value range, and within the same value range, the discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient decrease or remain unchanged as the number of lag days increases. A heuristic global optimization algorithm is used to perform global collaborative optimization on the constrained optimization model. The discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient are encoded as a unified individual and iteratively solved to obtain the behavior profile-accumulated temperature effect parameter matrix.

[0018] Preferably, the method further includes: Based on the revised daily maximum heat index, the equivalent daily maximum temperature is calculated by reverse calculation, specifically including: A numerical solution equation for the temperature to be determined is constructed. The numerical solution equation takes the temperature to be determined and the daily maximum humidity of the target day as input to the heat index calculation model, and takes the corrected daily maximum heat index as the target output value. A numerical approximation algorithm is used to solve for the temperature to be determined. Within a preset temperature search interval, a unique solution is found that makes the output value of the heat index calculation model equal to the corrected daily maximum heat index. The unique solution is then determined as the equivalent daily maximum temperature.

[0019] The second part describes a heat index correction system based on accumulated temperature effect, which operates a heat index correction method based on accumulated temperature effect as described in the first part, including: The data acquisition module is used to acquire historical data of the target area within a preset time period, including residential electricity load data, meteorological characteristic data, and non-meteorological characteristic data. The behavior profile construction module is used to construct a behavior feature vector based on the residential electricity load data; based on the behavior feature vector, a clustering algorithm is used to divide each day within the preset time period into several behavior profiles, and a correspondence between the non-meteorological feature data and the behavior profiles is established. The basic index calculation module is used to establish a heat index calculation model based on the meteorological characteristic data and calculate the maximum daily heat index within the preset time period. The parameter matrix generation module is used to construct a dual correction model that includes the air heat accumulation effect and the building radiation accumulation effect. For each behavior profile obtained by the behavior profile construction module, the historical data subset belonging to that behavior profile is used to optimize the model by maximizing the correlation between the corrected daily maximum heat index sequence and the residential electricity load data sequence in the historical data subset. The module solves the parameters of the dual correction model to generate a behavior profile-accumulation effect parameter matrix containing differentiated parameter groups under different profiles. The index correction calculation module is used to identify the behavioral profile of the target day to be corrected based on its non-meteorological characteristic data, call the corresponding differentiated parameter group from the behavioral profile-accumulated temperature effect parameter matrix, and substitute it into the dual correction model to calculate the corrected daily maximum heat index.

[0020] The third part is a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method described in the first part.

[0021] Part four is a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in Part one.

[0022] The beneficial effects of this invention are compared with those of the prior art: 1. A dual-correction model for air thermal accumulated temperature and building radiation accumulated temperature is constructed to make the heat index more accurately reflect the load response under continuous high temperature environment. This invention simultaneously considers the air thermal lag effect and the building heat storage lag effect, making up for the deficiency of the traditional heat index in not being able to characterize the building radiation accumulated temperature, and realizing accurate correction of the dynamic changes of load in continuous hot weather.

[0023] 2. A differentiated accumulated temperature parameter matrix based on user behavior profiles is proposed to achieve refinement and adaptability in the heatwave correction process. By clustering non-meteorological features and load behavior, multiple behavior profiles are formed, and accumulated temperature parameters are solved separately for each profile. This allows the model to automatically call differentiated parameters according to different date types, significantly improving its ability to match load changes under multiple scenarios and behavior patterns.

[0024] 3. A data-driven approach is adopted to solve the threshold, backtracking days, and decay coefficient of the accumulated temperature model, thereby improving the model's generalization ability and robustness. This invention utilizes cubic spline fitting, correlation analysis, and genetic algorithms to construct a fully data-driven parameter optimization mechanism, enabling the modified model to adapt to meteorological environments in different regions and years, thus improving the model's transferability. Attached Figure Description

[0025] Figure 1 This is a flowchart of a heat index correction method based on accumulated temperature effect provided by the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0027] Example 1: like Figure 1 As shown, a heat index correction method based on accumulated temperature effect includes the following steps: S1: Obtain historical data of the target area within a preset time period, including residential electricity load data, meteorological characteristic data, and non-meteorological characteristic data; In S1, the historical data includes: The residential electricity load data includes daily maximum electricity load data; The meteorological data includes daily maximum temperature, daily maximum humidity, and daily solar radiation intensity. The non-meteorological feature data includes date type labels, electricity price type labels, and electricity demand-side management event information.

[0028] In practical implementation, considering the delayed cumulative effect of accumulated temperature, the calculation of the accumulated temperature model often requires the state values ​​of the previous few days as input. Therefore, the preset time period is preferably configured to include the formal modeling period (e.g., the core summer months of July and August) and a preceding buffer period (e.g., 15-30 days before the start date of the formal modeling period). By obtaining the data from the preceding buffer period, the initial iteration of the double-correction model can be completed, avoiding cold start errors caused by data truncation.

[0029] The specific definitions and preprocessing methods for the acquired historical data are as follows: The residential electricity load data includes daily maximum electricity load data. If the original data is time series data measured in minutes or hours (such as 96-point or 24-point load curves), the daily maximum load is filtered out using the maximum value function. The meteorological characteristic data includes daily maximum temperature and daily maximum humidity; the meteorological characteristic data also includes daily solar radiation intensity, denoted as S. For daily solar radiation intensity, to accurately reflect the heat accumulation effect of solar radiation on the building envelope, the daily cumulative radiation is preferred. However, if data is limited, the daily maximum radiation intensity can be used as an alternative characteristic. This data is used for subsequent correction of building radiation accumulated temperature.

[0030] The non-meteorological feature data is used to assist in constructing user behavior profiles, including: Date type labels are used to distinguish the social attributes of different dates, such as marking them as weekdays, weekends, public holidays, or adjusted workdays; Electricity price type labels are used to reflect the guiding effect of electricity pricing policies on users' electricity consumption behavior. Specifically, this label can be defined as the electricity price rate model implemented on that day (e.g., the date of implementation of peak electricity pricing policy, the date of implementation of off-peak electricity pricing policy, etc.), or as the electricity price period attribute to which the time of peak load occurs each day (e.g., if the highest load on a certain day occurs during the peak electricity price period, that day is marked as "peak electricity price preference"; if it occurs during the off-peak electricity price period, it is marked as "off-peak electricity price preference"). Electricity demand-side management event information is used to record whether demand response or orderly electricity consumption measures were implemented on the same day, as well as the specific response level.

[0031] By combining meteorological and non-meteorological data, the model's ability to characterize behavioral factors can be enhanced, providing a more comprehensive data foundation for subsequent revisions to the heat index.

[0032] S2: Construct a behavioral feature vector based on the residential electricity load data; use a clustering algorithm based on the behavioral feature vector to divide each day within the preset time period into several behavioral profiles.

[0033] In this embodiment, the specific implementation process of S2 includes: S2.1: Perform numerical encoding processing on the non-meteorological feature data, and construct a behavioral feature vector based on the residential electricity load data, specifically including: One-Hot Encoding is used to vectorize the date type labels (such as weekdays and weekends) and the electricity price type labels (such as peak / flat / valley preferences), and the electricity demand-side management event information is converted into binary values ​​(e.g., 0 indicates no event and 1 indicates an event) to obtain non-meteorological numerical features; The non-meteorological numerical features and the residential electricity load data are normalized (for example, the data is mapped to the [0,1] interval using the Min-Max normalization method), and the normalized data are combined to construct the behavioral feature vector.

[0034] S2.2: Based on the behavioral feature vector, a clustering algorithm is used to divide each day within the preset time period into several behavioral profiles, specifically including: The behavioral feature vectors are clustered using a clustering algorithm (K-means clustering algorithm is preferred in this embodiment).

[0035] The number of clusters is determined based on preset clustering evaluation indicators, forming several behavioral pattern clusters, and the behavioral pattern cluster to which each day belongs within the preset time period is determined. In this embodiment, the preset clustering evaluation indicators preferably employ the elbow rule or the silhouette coefficient method.

[0036] The clusters of behavioral patterns are defined as corresponding behavioral profiles. For example, the clustering results may include the following typical profiles: Profile A (High-load rigid demand type): Corresponds to typical weekday behavior patterns, characterized by strong regularity in electricity consumption, high peak load, and significant impact from electricity price fluctuations; Profile B (Low Load Dispersive Type): Corresponds to weekend or rest day behavior patterns, characterized by low load levels and high dispersion in time periods; Image C (Specific Event Impact Type): Corresponds to holidays or major event days, where there may be peak shifts due to concentrated activities or family gatherings; Profile D (Controlled Adjustment Type): Corresponds to the daily behavior pattern of demand response, and the load curve shows that it is significantly affected by demand-side regulation intervention (such as peak shaving and valley filling).

[0037] Further statistical analysis of the distribution patterns of non-meteorological features in each behavioral profile is conducted to construct a correspondence between "non-meteorological feature combinations" and "behavioral profiles." For example, if the statistics show that the tag combination of "working day" and "non-demand response day" has a 95% probability of falling into "profile A," then a corresponding matching rule is established. This matching rule library is used in subsequent steps to index the corresponding behavioral profile based solely on the non-meteorological features of the target day, thereby achieving automatic identification of the behavioral background for future dates.

[0038] S3: Based on the meteorological characteristic data, establish a heat index calculation model to calculate the maximum daily heat index within the preset time period.

[0039] In this embodiment, a computational model is constructed based on the Rothfusz Regression Equation, which is commonly used in meteorology. Considering the sensitivity of power load to extreme high temperature and humidity environments, this step aims to construct a comprehensive index reflecting the most severe thermal stress within a single day.

[0040] If the daily maximum temperature obtained by S1 is in degrees Celsius, it is first converted to degrees Fahrenheit. The heat index calculation model is defined as follows:

[0041] in, To calculate the daily maximum heat index, The highest temperature of the day (unit: Fahrenheit °F). The maximum daily relative humidity (unit: %, values ​​are integers). For fixed coefficients, , , , , , , , , , The scaling factor is used to adjust the calculated heat index value to a range of magnitude that matches the characteristics of the electricity load, facilitating subsequent data processing and model convergence. In this embodiment, The value can be 1 (where H is the standard perceived temperature value). In other embodiments, it can be set to other positive numbers (e.g., 10 or 100) depending on the situation, so that it is on the same order of magnitude as the load value.

[0042] It is worth noting that although the highest temperature and the highest humidity do not usually occur at the same time in meteorology, this embodiment selects the extreme values ​​of the two for combined calculation to construct a theoretical upper limit heat index. This index can characterize the potential driving force of extreme weather conditions on the grid's cooling load to the greatest extent, preventing the omission of extreme load risks due to smoothing processing.

[0043] S4: Construct a dual correction model that includes air heat accumulation effect and building radiation accumulation effect; for each behavior profile obtained in S2, use a subset of historical data belonging to that behavior profile to maximize the correlation between the corrected daily maximum heat index sequence and the residential electricity load data sequence in the historical data subset as the optimization objective, solve the parameters of the dual correction model, and generate a behavior profile-accumulation effect parameter matrix containing differentiated parameter groups under different profiles.

[0044] In this embodiment, the specific execution process of step S4 is as follows: S4.1 Constructing the dual correction model includes: Calculate the air heat accumulated temperature correction and the building radiation accumulated temperature correction separately, and then add the air heat accumulated temperature correction and the building radiation accumulated temperature correction to the daily maximum heat index to obtain the corrected daily maximum heat index. The calculation of the air heat accumulated temperature correction includes: within the effective air heat backtracking days, determining the corresponding discrete air heat accumulated temperature effect coefficient based on the numerical range of the daily maximum heat index and the number of lag days; calculating the difference between the daily maximum heat index and the boundary air heat index; multiplying the difference by the discrete air heat accumulated temperature effect coefficient and summing them to obtain the air heat accumulated temperature correction. The calculation of the building radiation accumulated temperature correction includes: within the effective building radiation backtracking days, determining the corresponding discretized building radiation accumulated temperature effect coefficient based on the numerical range of the daily solar radiation intensity and the number of lag days; calculating the difference between the daily solar radiation intensity and the limit building radiation index; multiplying the difference by the discretized building radiation accumulated temperature effect coefficient and summing them to obtain the building radiation accumulated temperature correction. In this embodiment, the specific mathematical expression of the dual correction model is as follows: a heat index correction model under the accumulated temperature effect is established, and the "air heat effect" and "building radiant heat effect" are cumulatively modeled. The daily maximum heat index H calculated in step 3 is dynamically corrected to obtain the corrected daily maximum heat index. The formula is:

[0045] in, Indicates the first The daily maximum heat index for the day, Indicates the corrected number The daily maximum heat index for the day.

[0046] The first summation term represents the hysteretic accumulation of the air's thermal effect: Let be the daily maximum heat index on day ij (i.e., lagging by day j). The threshold temperature index (i.e., the starting threshold that causes a significant increase in load). This represents the discretized air thermal accumulated temperature effect coefficient for the i-th sample after a lag of j days. The specific value of this coefficient is determined by... The numerical range in which it is located determines the value (as shown in Table 1 below).

[0047] The second summation term represents the lagged cumulative effect of building radiative heat: Let be the solar radiation intensity on day ij. This indicates the radiation heat index of the boundary building. This represents the solar radiation accumulated temperature effect coefficient, the specific value of which is determined by... The range of values ​​it falls within determines this.

[0048] The effective number of days of air thermal backtracking The determination method includes: taking the smaller value between the theoretical maximum air heat backtracking days and the first consecutive days, where the first consecutive days refers to the number of days prior to day i where the daily maximum heat index is greater than the threshold air heat index, expressed as: .in, Indicates the first The continuous satisfaction of the previous day The first consecutive days, This indicates the theoretical maximum number of days for atmospheric thermal backtracking.

[0049] The cutoff logic is set based on the thermal inertia excitation-blocking mechanism, that is, when the temperature on a certain day is below a certain threshold. When the environmental thermal pressure is interrupted, the building envelope changes from the temperature accumulation excitation state to the heat dissipation recovery state, the historical temperature accumulation chain fails, and the temperature accumulation effect is cut off and recalculated.

[0050] Similarly, the effective building radiation backtracking days The determination method includes: taking the smaller value between the theoretical maximum building radiation backtracking days and the second consecutive days, where the second consecutive days refers to the number of days prior to day i where the daily solar radiation intensity is greater than the threshold building radiation index. This is expressed as: .in, Indicates the first The continuous satisfaction of the previous day The second consecutive day, This represents the theoretical maximum number of days for building radiation to be traced back. If the solar radiation intensity is lower than the threshold value on any given day, the radiation accumulation process is considered to be blocked, and the tracing window ends.

[0051] S4.2 Solving the parameter matrix based on behavioral profiling The parameters of the dual-correction model are solved using the corresponding subset of historical data. This involves two types of parameters in the model: structural parameters (…). , , , The structural parameters (vector k, vector w) and the coefficient parameters are solved in this embodiment using a step-by-step optimization strategy: first, the structural parameters are determined through numerical analysis (corresponding to S4.2.2-S4.2.3), and then a constrained optimization model is constructed on this basis to solve the coefficient parameters (corresponding to S4.2.4).

[0052] Specifically, the execution process of step S4.2 is as follows: S4.2.1 Solving the parameter matrix based on behavioral profiling The steps required to solve for the parameters of the dual-correction model using the corresponding subset of historical data are as follows: First, obtain the "behavioral profiles" for N days obtained from clustering in S2. Then, start a loop to iterate through each profile (taking profile A as an example): 1. Select all date samples belonging to "Portrait A" from historical data to form a training subset; 2. Based on this subset, execute steps S4.2.2 to S4.2.4 sequentially to solve for the air thermal effect parameter set. and building radiation effect parameter group .

[0053] 3. When the loop ends, output the "Behavioral Profile - Accumulated Temperature Effect" parameter matrix containing all profiles.

[0054] S4.2.2, Solving for the boundary threshold and .

[0055] Based on the selected training subset, the following operations are performed: Since the curve of residential electricity load changing with the daily maximum heat index (or daily solar radiation intensity) has different magnitudes and trends at each point, a functional relationship between residential electricity load L and the daily maximum heat index H is constructed using cubic spline fitting. And calculate the maximum value of its first derivative. :

[0056] The heat index value to which residential electricity load is most sensitive to temperature is called the threshold air heat index. Similarly, a functional relationship can be constructed between residential electricity load L and daily solar radiation intensity S. The S-value corresponding to the maximum value of the derivative is used as the boundary building radiation index. .

[0057] S4.2.3, Determine the maximum theoretical backtracking days and .

[0058] A combination of trial-and-error and correlation analysis was employed to determine the maximum number of backtracking days for the accumulated air temperature. For example: 1. Set a backtracking period of 1 year to be tested. The range of values ​​(e.g.) sky).

[0059] 2. For each sample i in the training subset (total length N), a truncation length of [length missing] is extracted. The heat index sequence and load sequence of a day constitute a valid data set.

[0060] 3. For each Calculate the Pearson correlation coefficient between the heat index and the load for each data set, and obtain the average correlation coefficient. The calculation formula is as follows:

[0061]

[0062] in, The total length of the samples in the training subset. Indicates the first The highest heat index of the day, Indicates the first The maximum load of the day Indicates the first The local average of the daily maximum heat index within each data set. Indicates the first Local average of daily maximum load within the group:

[0063]

[0064] Compare the average correlation coefficients corresponding to different d values The d value that maximizes this average value is selected as the final determined maximum theoretical backtracking number of days. .

[0065] Similarly, by replacing the heat index H in the above process with the solar radiation intensity S, the maximum number of days of building radiation backtracking can be obtained. .

[0066] S4.2.4, Solving the cumulative effect coefficient vector piecewise and .

[0067] 1. A discretization decomposition method is adopted, and the discretized air heat accumulated temperature effect coefficient is calculated and processed simultaneously. Discretized building radiation accumulated temperature effect coefficient Considering the varying cumulative intensity across different temperature ranges, the heat index H and solar radiation intensity S are divided into several arithmetic intervals. The step size for the heat index is set to m, and the total number of discretized intervals is set to M. For example, m=1.

[0068] Calculate the discrete air thermal accumulated temperature effect coefficient for each interval. The details are shown in Table 1.

[0069] Table 1 Discretized air thermal accumulated temperature effect coefficients

[0070] Similarly, the solar radiation intensity S is segmented, and corresponding discretized building radiation accumulation coefficients are established. .

[0071] 2. A constrained optimization model is constructed with the objective function of maximizing the Pearson correlation coefficient between the corrected heat index H' and the actual load L. The objective function of the constrained optimization model is to maximize the Pearson correlation coefficient between the corrected daily maximum heat index sequence calculated by the double-correction model and the residential electricity load data sequence in the historical data subset. The constraints of the constrained optimization model include: both the discretized air heat accumulation temperature effect coefficient and the discretized building radiation accumulation temperature effect coefficient are within a preset numerical range, and within the same numerical range, both the discretized air heat accumulation temperature effect coefficient and the discretized building radiation accumulation temperature effect coefficient decrease sequentially or remain unchanged with increasing lag days. This is expressed as:

[0072] Constraints:

[0073] in, The Pearson correlation coefficient is an evaluation index representing the effect of the correction. This represents the covariance between the corrected heat index sequence H' and the actual load sequence L; and These represent the variances of the corrected heat index series and the actual load series, respectively. This indicates that the correlation coefficient is a function of the air heat effect coefficient vector k and the building radiation effect coefficient vector w. The cumulative effect decreases as the lag days j increase.

[0074] 3. For this complex optimization problem, a heuristic global optimization algorithm is used to perform global collaborative optimization on the constrained optimization model. The discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient are encoded as a unified individual and iteratively solved to obtain the behavior profile-accumulated temperature effect parameter matrix.

[0075] Specifically, in this embodiment, the coefficients k and w are encoded into the same chromosome, and a genetic algorithm is used for global collaborative optimization. Through selection, crossover, and mutation operations, iterative calculations are performed until convergence to obtain the optimal coefficient vector set.

[0076] S5: For the target day to be corrected, identify its behavioral profile based on its non-meteorological feature data, call the corresponding differentiated parameter group from the behavioral profile-accumulated temperature effect parameter matrix, and substitute it into the dual correction model to calculate the corrected daily maximum heat index.

[0077] Specifically, using the target date to be corrected as the anchor point, a complete data time window is determined for calculating the dual accumulated temperature. This window covers the target date and the period preceding it. The day (i.e., the backtracking window).

[0078] For each day within this time window, meteorological characteristic data (including daily maximum temperature, daily maximum humidity, and daily solar radiation intensity) are dynamically assembled based on their relative relationship with the current system date: For the target day: meteorological forecast data is directly called; For the preceding days in the retrospective window: if a day is earlier than or equal to the current system date, historical measured data is called; if a day is later than the current system date (i.e., in the transition period between the current date and the target day), meteorological forecast data is called.

[0079] Using non-meteorological data of the target day as input, the matching rules (or mapping relationships) established in step S2 are used for retrieval to identify the behavioral profile to which the target day belongs. For example: obtaining... If the non-meteorological characteristic is "Portrait C is a holiday behavior pattern", then the target day is determined to belong to Portrait C.

[0080] Based on the recognition results, the exclusive differentiated parameter group corresponding to "Portrait C" is dynamically called from the behavioral profile-accumulated temperature effect parameter matrix generated by S4.2.

[0081] Using this set of selected parameters, substitute them into the double correction model constructed in step S4.1 to solve for the corrected daily maximum heat index for the target day. .

[0082] S6 calculates the equivalent maximum temperature based on the corrected exponent. The method further includes: calculating the equivalent daily maximum temperature based on the corrected daily maximum heat index, specifically including: A numerical equation for solving the temperature to be determined is constructed. This equation uses the temperature to be determined and the daily maximum humidity of the target day as inputs to the heat index calculation model, and the corrected daily maximum heat index as the target output value. The equation is expressed as follows:

[0083] in, Substitute the desired temperature T and the constant relative humidity for the day. The calculated value of the Rothfusz heat index is as follows. This is the target value for the corrected heat index calculated in this step.

[0084] By solving To deduce the temperature value, a numerical approximation algorithm (such as the bisection method or Newton's iteration method) is used to solve for the desired temperature. First, the temperature is set... Reasonable preset temperature search range, for example [ Within a preset temperature search range, a unique solution is sought that makes the output value of the heat index calculation model equal to the corrected daily maximum heat index. The unique solution is determined as the equivalent daily maximum temperature, and the obtained Fahrenheit is... Converted to Celsius and output. In this embodiment, the equivalent daily maximum temperature refers to the numerical representation of the multidimensional thermal environment pressure, which includes the air heat accumulation effect, building radiation effect, and high humidity effect, mapped back to a single temperature dimension. This value eliminates the lack of representation of the cumulative effect of continuous high temperature in traditional temperature data, and more accurately reflects the actual perceived heat on the power grid load side. It can directly replace the original temperature input in the existing load forecasting model.

[0085] Example 2: A heat index correction system based on accumulated temperature effect, running a heat index correction method based on accumulated temperature effect as described in Example 1, includes: The data acquisition module is used to acquire historical data of the target area within a preset time period, including residential electricity load data, meteorological characteristic data, and non-meteorological characteristic data. The behavior profile construction module is used to construct a behavior feature vector based on the residential electricity load data; based on the behavior feature vector, a clustering algorithm is used to divide each day within the preset time period into several behavior profiles, and a correspondence between the non-meteorological feature data and the behavior profiles is established. The basic index calculation module is used to establish a heat index calculation model based on the meteorological characteristic data and calculate the maximum daily heat index within the preset time period. The parameter matrix generation module is used to construct a dual correction model that includes the air thermal accumulated temperature effect and the building radiation accumulated temperature effect. For each behavior profile obtained by the behavior profile construction module, the parameters of the dual correction model are solved using a subset of historical data belonging to that behavior profile, generating a behavior profile-accumulated temperature effect parameter matrix containing differentiated parameter groups under different profiles. The index correction calculation module is used to identify the behavioral profile of the target day to be corrected based on its non-meteorological characteristic data, call the corresponding differentiated parameter group from the behavioral profile-accumulated temperature effect parameter matrix, and substitute it into the dual correction model to calculate the corrected daily maximum heat index.

[0086] Example 3: A terminal, comprising a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method described in the first part.

[0087] Part four is a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in Part one.

[0088] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0089] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0090] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0091] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for correcting the heat index based on the accumulated temperature effect, characterized in that, include: S1: Obtain historical data of the target area within a preset time period, including residential electricity load data, meteorological characteristic data, and non-meteorological characteristic data; S2: Construct a behavioral feature vector based on the residential electricity load data; use a clustering algorithm based on the behavioral feature vector to divide each day within the preset time period into several behavioral profiles, and establish a correspondence between the non-meteorological feature data and the behavioral profiles; S3: Based on the meteorological characteristic data, establish a heat index calculation model to calculate the maximum daily heat index within the preset time period; S4: Construct a dual correction model that includes air heat accumulation effect and building radiation accumulation effect; for each behavior profile obtained in S2, use a subset of historical data belonging to that behavior profile to maximize the correlation between the corrected daily maximum heat index sequence and the residential electricity load data sequence in the historical data subset as the optimization objective, solve the parameters of the dual correction model, and generate a behavior profile-accumulation effect parameter matrix containing differentiated parameter groups under different profiles; S5: For the target day to be corrected, identify its behavioral profile based on its non-meteorological feature data, call the corresponding differentiated parameter group from the behavioral profile-accumulated temperature effect parameter matrix, and substitute it into the dual correction model to calculate the corrected daily maximum heat index.

2. The heat index correction method based on accumulated temperature effect according to claim 1, characterized in that, In S1, the historical data includes: The residential electricity load data includes daily maximum electricity load data; The meteorological data includes daily maximum temperature, daily maximum humidity, and daily solar radiation intensity. The non-meteorological feature data includes date type labels, electricity price type labels, and electricity demand-side management event information.

3. The method for correcting the heat index based on accumulated temperature effect according to claim 2, characterized in that, In S2, constructing a behavioral feature vector based on the residential electricity load data includes: The date type label and the electricity price type label are converted using unique thermal encoding, and the electricity demand-side management event information is converted into binary values ​​to obtain non-meteorological numerical features; The non-meteorological numerical features and the residential electricity load data are normalized, and the normalized data are combined to construct the behavioral feature vector.

4. The method for correcting the heat index based on the accumulated temperature effect according to claim 3, characterized in that, In S2, based on the behavioral feature vector, a clustering algorithm is used to divide each day within the preset time period into several behavioral profiles, including: The behavior feature vectors are clustered using a clustering algorithm. The number of clusters is determined according to a preset clustering evaluation index, forming several behavior pattern clusters. The behavior pattern cluster to which each day belongs within the preset time period is also determined. The behavior pattern clusters are defined as corresponding behavior profiles, and the distribution patterns of non-meteorological feature data corresponding to each behavior profile are statistically analyzed to construct the correspondence between non-meteorological feature data and behavior profiles.

5. The heat index correction method based on accumulated temperature effect according to claim 1, characterized in that, In S4, the dual correction model is constructed in the following ways: Calculate the air heat accumulated temperature correction and the building radiation accumulated temperature correction separately, and then add the air heat accumulated temperature correction and the building radiation accumulated temperature correction to the daily maximum heat index to obtain the corrected daily maximum heat index. The calculation of the air heat accumulated temperature correction includes: within the effective air heat backtracking days, determining the corresponding discrete air heat accumulated temperature effect coefficient based on the numerical range of the daily maximum heat index and the number of lag days; calculating the difference between the daily maximum heat index and the boundary air heat index; multiplying the difference by the discrete air heat accumulated temperature effect coefficient and summing them to obtain the air heat accumulated temperature correction. The calculation of the building radiation accumulated temperature correction includes: within the effective building radiation backtracking days, determining the corresponding discretized building radiation accumulated temperature effect coefficient based on the numerical range of the daily solar radiation intensity and the number of lag days; calculating the difference between the daily solar radiation intensity and the limit building radiation index; multiplying the difference by the discretized building radiation accumulated temperature effect coefficient and summing them to obtain the building radiation accumulated temperature correction. The determination of the effective air heat backtracking days includes: taking the smaller value between the theoretical maximum air heat backtracking days and the first consecutive days, where the first consecutive days refers to the number of days prior to day i where the daily maximum heat index is greater than the threshold air heat index; the determination of the effective building radiation backtracking days includes: taking the smaller value between the theoretical maximum building radiation backtracking days and the second consecutive days, where the second consecutive days refers to the number of days prior to day i where the daily solar radiation intensity is greater than the threshold building radiation index.

6. The heat index correction method based on accumulated temperature effect according to claim 5, characterized in that, In S4, the parameters of the dual-correction model are solved using the corresponding subset of historical data, specifically including: A constrained optimization model is constructed, the objective function of which is to maximize the Pearson correlation coefficient between the corrected daily maximum heat index sequence calculated by the dual correction model and the residential electricity load data sequence in the historical data subset. The constraints of the constrained optimization model include: the discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient are both within a preset value range, and within the same value range, the discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient decrease or remain unchanged as the number of lag days increases. A heuristic global optimization algorithm is used to perform global collaborative optimization on the constrained optimization model. The discretized air thermal accumulated temperature effect coefficient and the discretized building radiation accumulated temperature effect coefficient are encoded as a unified individual and iteratively solved to obtain the behavior profile-accumulated temperature effect parameter matrix.

7. The method for correcting the heat index based on accumulated temperature effect according to claim 1, characterized in that, The method further includes: Based on the revised daily maximum heat index, the equivalent daily maximum temperature is calculated by reverse calculation, specifically including: A numerical solution equation for the temperature to be determined is constructed. The numerical solution equation takes the temperature to be determined and the daily maximum humidity of the target day as input to the heat index calculation model, and takes the corrected daily maximum heat index as the target output value. A numerical approximation algorithm is used to solve for the temperature to be determined. Within a preset temperature search interval, a unique solution is found that makes the output value of the heat index calculation model equal to the corrected daily maximum heat index. The unique solution is then determined as the equivalent daily maximum temperature.

8. A heat index correction system based on accumulated temperature effect, comprising a heat index correction method based on accumulated temperature effect as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire historical data of the target area within a preset time period, including residential electricity load data, meteorological characteristic data, and non-meteorological characteristic data. The behavior profile construction module is used to construct a behavior feature vector based on the residential electricity load data; based on the behavior feature vector, a clustering algorithm is used to divide each day within the preset time period into several behavior profiles, and a correspondence between the non-meteorological feature data and the behavior profiles is established. The basic index calculation module is used to establish a heat index calculation model based on the meteorological characteristic data and calculate the maximum daily heat index within the preset time period. The parameter matrix generation module is used to construct a dual-correction model that includes both air thermal accumulated temperature effect and building radiation accumulated temperature effect; For each behavior profile obtained by the behavior profile construction module, a subset of historical data belonging to that behavior profile is used to optimize the correlation between the corrected daily maximum heat index sequence and the residential electricity load data sequence in the historical data subset. The parameters of the dual correction model are solved to generate a behavior profile-accumulated temperature effect parameter matrix containing differentiated parameter groups under different profiles. The index correction calculation module is used to identify the behavioral profile of the target day to be corrected based on its non-meteorological feature data, call the corresponding differentiated parameter group from the behavioral profile-accumulated temperature effect parameter matrix, and substitute it into the dual correction model to calculate the corrected daily maximum heat index.

9. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-7.

Images